Genomics of circadian rhythms in health and disease

This review focuses on the most recent advances in mammalian circadian rhythms research, highlighting novel techniques and explaining the importance and implications of these research findings for human disease, translational research, and medicine. We discuss a number of modern genomics approaches to the study of circadian rhythms, such as the evaluation of chromatin dynamics and gene regulation. Owing to the circadian functions that are common to these diseases, another factor that we highlight is the opportunity to intervene using the timed administration of drugs (chrono-pharmacology) or by targeting clock components. Indeed, as we discuss throughout this review, there may be great benefits to considering circadian timing in the treatment of metabolic disorders, cardiovascular disease, and cancer [53, 57, 58].

Recently, studies in mouse tissues have greatly enhanced our understanding of the circadian regulatory mechanisms for rhythmic transcription [43–45, 68, 69]. Sobel et al. [68] characterized the chromatin accessibility landscape by mapping DNase I hypersensitive sites (DHSs) in mouse liver across 24 h. DHS sites reflect open chromatin and their occupation of transcription start sites (TSSs), enhancers, and silencers mean that they are hallmarks of regulatory DNA. In this study, the authors found that 8% of 65,000 DHSs cycled with a 24-h period, in phase with Pol II binding and histone 3 lysine 27 acetylation (H3K27ac) marks, suggesting that regulatory elements within DHSs control rhythmic transcription [68]. Two additional studies have further advanced our understanding of chromatin interactions [43, 44]. Mermet et al. [43] utilized circular chromosome conformation capture sequencing (4C-seq) to explore the three-dimensional chromatin interactions of a locus of interest with other genomic regions (one-to-all). They examined the TSSs of the clock repressor gene Cryptochrome 1 (Cry1) and of a liver-specific clock-controlled gene, Gys2 (Glycogen synthetase 2), which encodes the rate-limiting enzyme in hepatic glycogen synthesis. These genes show rhythmic transcription with opposite phases, allowing the authors to correlate their chromatin interaction profiles with their gene transcription regulation. The authors found that chromatin contact with such regions increases at the time of the day when the corresponding gene has its peak expression. Strikingly, abrogation of an enhancer that is rhythmically recruited to the Cry1 promoter leads to a shortened period of locomotor activity, suggesting that such interacting loops are necessary for the modulation of rhythmic behaviors [43]. Together, these studies show that rhythmic modulation of chromatin conformation adds an important layer of control over circadian gene transcription (Fig. 2).

Despite these genome-wide advances, our understanding of circadian regulation at the protein level is much more limited, mostly because of the difficulty of quantitative assessment of the proteome [70, 71]. Recent technological advances have allowed quantification of the circadian proteome, the nuclear proteome [72], and the phospho-proteome [73]. These studies revealed the rhythmic presence of about 500 proteins (~ 10%) in the nucleus that are components of nuclear complexes involved in transcriptional regulation, ribosome biogenesis, DNA repair, and the cell cycle [72]. Strikingly, more than 5000 (~ 25%) phosphorylation sites are rhythmic, far exceeding rhythms in protein abundance (phosphorylation is an example of post-translational modification (PTM); Fig. 2). Overall, recent studies have vastly enhanced our understanding of the genome-wide reach of the molecular clock and how it is regulated.

Driven by the circadian clock, a regular daily pattern of eating and fasting maintains normal circadian physiology. However, recurrent disruption of daily activity–rest rhythms, and thus feeding patterns (as occurs in shift workers), is associated with metabolic syndrome [91]. Genetic disruption of the circadian clock also predisposes rodents to metabolic disease [32, 33]. The clock controls metabolism directly by driving transcriptional programs for certain metabolic pathways. For example, CRY1 suppresses hepatic gluconeogenesis during fasting through the regulation of cAMP/CREB signaling, the rhythmic repression of the glucocorticoid receptor gene, and the suppression of nuclear FOXO1 that, in turn, downregulates gluconeogenesis [92–94]. Another clock repressor, PER2, controls lipid metabolism by direct regulation of peroxisome proliferator-activated receptor gamma (PPARγ) and mitochondrial rate-limiting enzymes [95, 96]. The nuclear hormone receptors, REV-ERBs, also directly regulate the transcription of several key rate-limiting enzymes for fatty acid and cholesterol metabolism [97] (reviewed in [98]). Disruption of CLOCK and BMAL1 has also been associated with obesity, hyperinsulinemia, and diabetes [32, 33, 99, 100]. The circadian posttranscriptional regulator Nocturnin also controls lipid and cholesterol metabolism [101]. Recently, an atlas of circadian metabolic profiles across eight tissues revealed temporal cohesion among tissues, whereas nutritional challenge (a high-fat diet) impacted each tissue differentially [102]. In addition to direct modulation of mammalian metabolism, indirect control by the clock occurs through its regulation of behavior, food intake, and the oscillation of hormones such as insulin, glucagon, peptide YY, glucagon-like peptide 1, corticosterone, leptin, and ghrelin (reviewed in [103]). Although we know much about the circadian clock’s control of metabolism, the mechanisms behind this control are far from understood [104]. How nutritional challenges dysregulate the clock and how clock disruption increases adipogenesis remain open questions in the field. However, recent studies have contributed to our understanding of such complex phenomena.

Dramatic temporal variation in sensitivity to endotoxins between morning and evening was first discovered in the 1960s [13]; but only in the past decade have major inroads been made in our understanding of clock control over the immune system (Fig. 1). Circadian clock control impinges upon many aspects of the immune response, from the trafficking of immune cells, to the activation of innate and adaptive immunity, to host–pathogen interactions. There have been thorough reviews of these topics [121], so instead we highlight the most recent findings.

Cardiovascular complications have higher incidence in the morning. Many different studies have connected the clock with cardiovascular function, including daily variation in blood pressure, and even response to aspirin [82, 145, 146]. Some studies suggest that pharmacological targeting of REV-ERB decreases atherosclerotic plaque burden in mice [147]. On the other hand, other studies suggest that deletion of Bmal1 in myeloid cells increased monocyte recruitment and atherosclerosis lesion size [148]. A recent study has shed light on a mechanism that may contribute to this phenomenon. The adherence of myeloid cells to microvascular beds peaks during the early active phase, which appears to be a consequence of peak cell recruitment to atherosclerotic lesions 12 h earlier [57]. Winter et al. [57] showed that both the upregulation of cell adhesion molecules during the active phase by endothelial cells and the presence of immobilized chemokines (emitted by either endothelial cells or myeloid cells) on arterial vessels attract leukocytes into atherosclerotic lesions. Thus, the chemokine CCL2 (C-C motif chemokine ligand 2) and its receptor CCR2 (C-C motif chemokine receptor 2) are at the core of this daily pattern of leukocyte migration and adhesion to the lesions. Importantly, the authors found that timed pharmacological CCR2 neutralization caused inhibition of atherosclerosis without disturbing microvascular recruitment, providing a proof-of-principle treatment schedule for chrono-pharmacological intervention in atherosclerosis (Fig. 3).

Loss of Bmal1 results in an acceleration of aging and a shortened life span in mice [84]. The cardiovascular system is among the systems affected by aging, with Bmal1^−/− mice being predisposed to developing atherosclerosis. Using an inducible knockout (iKO), Yang et al. [149] tested whether these age-related phenotypes remained if mice lost BMAL1 as adults. They found that both Bmal1^−/− and iKO models exhibit markers consistent with accelerated aging (ocular abnormalities and brain astrogliosis), behavioral disruption, and transcriptional dysregulation. This is consistent with the fact that conditional ablation of the pancreatic clock still causes diabetes mellitus [99]. However, some other biomarkers of aging, including premature death in Bmal1^−/− mice, were not replicated in the iKOs [149]. Among those, the predisposition for atherosclerosis appears to be reversed in iKOs [149]. These data suggest that some of the cardiovascular phenotypes associated with Bmal1 depletion may result from Bmal1 function during development. Although it is clear that there is a link between the circadian clock and atherosclerosis, further dissection of the importance of BMAL1 and other clock proteins in this disease is warranted.

Circadian rhythms in the suprachiasmatic nucleus (SCN) have been the focus of many years of research; but how the SCN imposes rhythmicity throughout the body (or even locally in the brain) is not fully understood. Recent studies have broadened the focus from neurons to astrocytes, demonstrating the important role of these glial cells in maintaining circadian rhythmicity [150–152]. A recent circadian atlas of non-human primates includes 64 tissues in the body, including 22 different regions in the brain [153]. The authors found genes cycling across the day in all brain regions, providing a comprehensive view of the reach of the circadian clock throughout the CNS of baboons [153]. While further studies are needed to understand fully the impact of rhythms in the nervous system and all their potential functions, the following studies are a step in that direction.

Epidemiological studies have linked circadian disruption to increased cancer susceptibility in all key organ systems [158–160]. Compelling evidence has shown that polymorphisms in the core circadian genes Per1, Per2, and Per3 are frequently found in human cancers, resulting in decreased expression of these genes [158], and that oncogenic MYC suppresses the clock [161]. Genetic loss of Per2 or Bmal1 promotes lung tumorigenesis in mice, leading to increased c-Myc expression, enhanced proliferation, and metabolic dysregulation [50]. Similarly, hepatocellular carcinoma (HCC) is induced by chronic jet lag in mice in a manner similar to that observed in obese humans: beginning with non-alcoholic fatty liver disease (NAFLD), then progressing to steatohepatitis and fibrosis and, ultimately, to HCC [49] (Fig.3). Thus, these two studies have convincingly shown a mechanistic connection between clock disruption and cancer development [49, 50]. In addition, microRNA miR-211, which suppresses Clock and Bmal1, also promotes tumor progression [162]. Targeting REV-ERBs is an effective strategy for combating cancer without altering the viability of normal cells or tissues. Using anticancer agonists of REV-ERBs (SR9009 and SR9011), Sulli et al. [58] were able to interfere with at least two cancer hallmarks: de novo lipogenesis and autophagy, which are important in meeting the metabolic demands of cancer cells.

Low oxygen levels in solid tumors stabilize hypoxia-inducible factors (HIFs), which are transcription factors that acidify the tumor microenvironment. Recent research has shown that HIFs are capable of influencing various clock transcripts [163–165]. Furthermore, Walton et al. [166] showed that acidfication of the tumor microenvironment by hypoxic cells disrupts the circadian clock and rhythmic transcriptome. They showed that low pH suppresses mTORC1 (Mammalian target of rapamycin complex 1) signaling, causing inhibition of translation. The authors further found that restoring mTORC1 signaling, either by buffering against acidification or by inhibiting lactic acid production, fully rescues translation and clock oscillations [166]. Overall, recent research on circadian rhythms and cancer has given major insights into disease mechanisms, which will hopefully allow for improved treatments, possibly including circadian considerations.

Circadian rhythms seem to decline with age [167, 168], with neuronal activity rhythms displaying an age-dependent decline in the master clock in the SCN [169]. In addition, the disruption of circadian rhythms through the ablation of Bmal1 leads to premature aging in mice [84]. Recent studies of aged stem cells and liver suggest that circadian transcriptional profiles in aging cells are rewired. However, in contrast with what was predicted, aging does not simply cause dampened circadian rhythmicity in the expression of genes that cycle when animals are young. Instead, a new set of genes begin to cycle in aged mice [83, 170]. Aged epidermal and skeletal muscle stem cells show reprogramming of gene expression towards a stress response, inflammation, and DNA damage, with core clock genes maintaining their rhythms [83]. Thus, this study supports the idea that aged stem cells retain a functional clock, but that this clock redirects the cell into new circadian functions with age. Perhaps this reprogramming is associated with the differential DNA methylation that occurs with aging [171] (see below). The key pathways or molecules that lead to this rewiring of the circadian transcriptome with aging remain unknown.

Additional studies have brought to light extra layers of circadian regulation that appear to decline with age. Polyamines modulate multiple cellular functions, and altered polyamine metabolism is associated with aging. Zwighaft et al. [55] linked polyamine metabolism, the clock, and aging, showing that the circadian clock controls polyamine levels and, in turn, that polyamines regulate circadian period. Polyamines exert their effects by impacting the interaction between the circadian repressors PER2 and CRY1. Interestingly, the longer circadian period of aged mice can be shortened with polyamine supplementation in the drinking water [55]. Another layer of circadian regulation appears to be in the modification of cytosines in DNA. De novo DNA methylation is established by the DNA methyltransferases DNMT3A and DNMT3B, which transfer a methyl group from S-adenosylmethionine to a cytosine at a cytosine guanine (CpG) site. On the other hand, cytosine methylation marks can be removed through an active demethylation pathway involving oxidation performed by TET (ten eleven translocation) enzymes [171]. DNA methylation might affect gene regulation by changing nucleosome stability and altering the nucleosome structure. Recently, Oh et al. [172] reported that a large proportion of cytosines show a circadian pattern of methylation in mice, and that the mRNA levels of nearby genes are positively correlated with corresponding oscillations in DNA methylation in liver and lung tissues. Consistent with the decrease of circadian oscillation of certain transcripts with age, oscillatory cytosine modifications (and DNA methylation, in general) also appear to decrease in older animals [172].

Alzheimer’s disease (AD) patients frequently experience increased daytime sleep and night-time wakefulness [54]. AD is associated with the production and deposition of the β-amyloid (Aβ) peptide, and soluble Aβ levels exhibit robust daily oscillations in mouse hippocampal interstitial fluid [78, 173]. However, little is known about how circadian rhythms may influence AD [174]. In a recent study trying to address the role of the circadian clock in determining Aβ levels, Kress et al. [175] showed that Aβ rhythms are normal when Bmal1 is deleted in the brain and retained only in the SCN. Nevertheless, whole-brain Bmal1 deletion causes loss of Aβ interstitial fluid rhythms in the hippocampus and markedly increases amyloid plaque burden. In addition to Aβ oscillations, tau levels also fluctuate in the brain interstitial fluid of mice and in the cerebral spinal fluid (CSF) of humans [54]. Tau levels appear to be higher during the animal’s active period and to increase when animals are subjected to sleep deprivation. Similarly, human CSF tau levels also increased by over 50% during sleep deprivation [54]. Finally, an interesting human cross-sectional study revealed an association between pre-clinical AD and disruption of the activity–rest rhythms. Specifically, pre-clinical amyloid plaques or higher CSF phosphorylated-tau to Aβ-42 ratios were associated with increased variability in daily behaviors, indicating fragmentation of activity–rest rhythms. The presence of abnormalities in circadian rhythms in pre-clinical AD suggests that circadian dysfunction could contribute to early pathogenesis or could serve as a biomarker of AD [176]. Together, these studies suggest that we should investigate the importance of a healthy sleep–wake cycle as an intervention for preventing AD and other tauopathies.

Circadian research, in particular the concept of chrono-pharmacology, is increasingly shaping our view of future research and medicine [177, 178]. It has introduced a time component to our view of metabolism, inflammation, and host–pathogen interactions (among other interactions), and has shown that targeting genes that are cycling at specific times of day may be advantageous [179–181]. Recent characterizations of the circadian transcriptional profiles of non-human primates [153] and humans [46] across multiple tissues have complemented the circadian atlas previously obtained for mice [181]. These reports have strengthened an important conclusion from the rodent data—the potential for chrono-pharmacological treatment of multiple diseases. Most of the protein-coding genes that have been found to be oscillating in primates encode proteins that are identified as druggable targets by the US Food and Drug Administration.

Regarding infectious diseases, treatments, and vaccinations could be more effective when administered at specific times of day. Indeed, influenza vaccine administration in the morning has been shown to improve antibody response over afternoon vaccination response in people over 65 years old [182]. This shows the potential to align the timing of external interventions, such as drug treatment or vaccinations, to the phase of our internal defenses. A further aspect to take into consideration is the potential for the pathogen itself to have circadian rhythms, as is the case for the sleeping sickness parasite, Trypanosoma brucei. We recently showed that this parasite has intrinsic circadian rhythms that affect its sensitivity to suramin treatment [183]. This may be a common feature of pathogens, although this remains to be determined.

Pharmacological modulation of the circadian machinery may also be an effective therapy for cancer [58] and potentially for sleep and anxiety [184]. Our own studies on parasite–host interactions may help to identify factors that alter the period of the circadian clock [138]. If so, molecule(s) could potentially be used to accelerate the rhythms of both central and/or peripheral clocks, helping people to overcome jet lag or even improving symptoms in patients with DSPD. The fact that physiology is intimately linked with circadian rhythmicity raises the question of when to intervene in all human diseases, and if there is a particular time of the day when treatment would be more effective or whether modulating a key clock protein function could alleviate the pathology.

The past few years have been very exciting for circadian research, making it clear that circadian biology is at the core of animal physiology. A multitude of additional layers of circadian clock regulatory mechanisms have been demonstrated recently. Such additional layers of regulation of the circadian clock machinery include chromatin conformation and interactions [43, 56], polyamines [55], NADP⁺:NADPH redox ratio [185], cytosine modifications [172], and even autophagy [120]. Among those, the genomics of circadian rhythms have expanded our understanding of daily physiological rhythms in health [43, 88, 112] and disease [53, 162].

In addition to circadian rhythms, there are also biological rhythms with shorter (ultradian) periods. Clusters of ultradian genes that cycle with a 12-h period have been identified in several peripheral tissues in mice [181, 186], many of which respond to feeding [187]. Recently, it was proposed that the mechanism behind these 12-h rhythms is a cell-autonomous 12-h pacemaker that is important for maintaining metabolic homeostasis [188]. In the future, it will be interesting to see what other aspects of physiology are influenced by ultradian rhythms and how they integrate with circadian physiology.

Overall, we believe that the growing body of evidence in mammalian circadian rhythms research is revealing an undisputable link between circadian rhythms and human health. Nevertheless, we are far from understanding the complexity of circadian biology and medicine. Exciting new aspects continue to emerge in terms of health and lifespan, including dietary influences [189], as well as differences between genders [190]. Circadian medicine is clearly an interdisciplinary field that requires complementary expertise [57, 138, 175]. Advances in technology have shaped circadian research in recent years [43, 73, 112] and will continue to be crucial going forward. Integrating the temporal axis into human physiology and medicine offers an opportunity to optimize the alignment of our internal rhythms to the environment, which will provide new opportunities for lifestyle and pharmacological interventions to treat diseases and promote health.