Coordination of cardiac rhythmic output and circadian metabolic regulation in the heart

In vivo data, however, are not available, as they will be subjected to a change in load, which also varies in a diurnal fashion. Blood pressure (BP) shows a day–night cycle, with a peak in the mid-morning and a trough at night. Normal ambulating humans have a 10% variation between the peak and the trough [12]. The diurnal variation in BP results from both cyclic physical activity and other endogenous factors such as autonomic tone and humoral factors, including the renin–angiotensin–aldosterone axis and cortisol [41, 42]. Heart rate varies in mammals as well; mathematical modeling of heart rate fluctuations across the diurnal cycle has demonstrated scale-invariant patterns in rats and humans, implying an intrinsic control mechanism common in mammals [43]. Additionally, both CCM mice and CBK mice demonstrated a slowed heart rate and loss of circadian rhythmicity; in particular, in CCM mice, a larger decrease in heart rate during the dark phase contributed to the loss of circadian rhythmicity [38, 44]. The diurnal variability in contractility necessitates a plastic metabolic program that can match its demand (Fig. 2).

The accurate timing of the onset of myocardial infarction (MI) is challenging in human studies. Although the onset of MI is demonstrably circadian, the pathobiology underlying circadian occurrence of vulnerable plaque rupture remains complex and likely involves circadian rhythmicity of several factors including BP, platelet aggregation, and fibrinolysis. A large study of 5645 individuals from 8 populations over a median follow-up of 11.4 years found that high morning systolic blood pressure surge could be a useful prognostic factor in stratifying cardiovascular risk [45]. This is consistent with the observation that a high rate of variability in systolic BP is correlated with increased medial hypertrophy of the large arteries [46]. Additionally, the morning blood pressure surge may directly contribute to destabilizing plaques; compared with plaques from individuals with hypertension but lacking a morning BP surge, plaques from individuals with a morning BP surge displayed an unstable plaque phenotype, with increased infiltration by immune cells (macrophages, T lymphocytes), increased ubiquitin–proteasome activity, nuclear factor-κB, matrix metalloproteinase-9, and less collagen content [47].

The influence of circadian variation in inflammatory response on circadian MI deserves special mention. Ly6C^hi inflammatory monocytes demonstrate diurnal variation in trafficking to sites of inflammation, a phenotype abolished by deletion of Bmal1, and myeloid-specific deletion of Bmal1 exacerbates diet-induced obesity and insulin resistance in mice, presumably due to alterations in chronic inflammatory responses [48]. External circadian signals from the autonomic nervous system also influence variable expression of ICAM-1 and CCL2 in the endothelium of skeletal muscle and VCAM-1 and CXCL12 in the bone marrow to influence leukocyte emigration to tissues [49]. Additionally, tumor necrosis factor α and interleukin-6 secretion after lipopolysaccharide stimulation is circadian, observations in line with the finding that macrophages have an endogenous clock with about 8% of transcripts showing rhythmic expression (including components involved in LPS-induced responses such as nuclear factor κB, JUN, and FOS), suggesting that beyond circadian recruitment of inflammatory cells, local inflammatory responses once they have arrived remain rhythmic [50]. Indeed, Schloss et al. show in a recent study that cardiac neutrophil recruitment, myeloid progenitor production, expression of cardiac adhesion molecule, and chemokine levels are all increased specifically during the active phase, and that mice subjected to left anterior descending coronary artery ligation during this phase develop larger infarcts and worsened cardiac function compared to control mice operated upon during the rest phase. Blockade of CXCR2 by antibody selectively reduced cardiac infiltration by neutrophils in the active phase but not the rest phase [51].

Platelet aggregability, circulating platelet aggregates, and fibrinolytic activity also vary in a circadian fashion, associating with a well-known early morning hypercoagulability and elevations in tissue factor pathway inhibitor and activated factor VII levels [52–55]. Platelet aggregability in the morning has been associated with risk of MI and sudden cardiac death [56]. In contrast, fibrinolytic activity is highest in the afternoon and lowest in the morning, in accordance with observations that tissue plasminogen activator inhibitor 1 levels are highest in the early morning, leading to lower tissue-type plasminogen activator [52, 57, 58]. Low morning fibrinolytic activity has been proposed as an independent risk factor for first acute MI in men and women [59]. Together, these studies provide evidence that there is a complex interplay of circadian factors underlying the time-of-day dependence of MI.

The first study to demonstrate an early morning clustering of MI used the time of the first elevation in the plasma creatine kinase MB level to estimate the time of the onset [60]. Subsequent studies, including data from the National Cardiovascular Data Registry, TIMI III registry and TIMI IIIB trial chose to use the onset of symptoms and confirmed the excess of STEMI, unstable angina, and non-Q MI in the early morning hours [61, 62]. Further, it has been shown that this chronological effect can be ameliorated by either a beta blocker or aspirin, suggesting diurnal variation in adrenergic activity and platelet aggregation play important roles (Fig. 2) [60, 63].

Although it is generally accepted that the timing of the MI has a circadian rhythmicity, there is some controversy regarding the size of the MI. Durgan et al. demonstrated that wild-type mouse hearts subjected to ischemic/reperfusion (IR) injury at the sleep to wake transition (ZT12) have a 3.5-fold increase in infarct size compared to the wake to sleep transition, and this variation was abolished in CCM mice [64]. This provided the first evidence that there is a time-of-day-dependent susceptibility to IR injury that is cell autonomous to the cardiomyocytes. However, Eckle et al. found a reduction in infarct size at ZT12 and ZT18 compared to ZT0. The same group also reported that mPer2 (dominant negative) mutant mice have larger infarct sizes and loss of ischemic cardioprotection when compared to the wild type. They demonstrated that PER2 is a downstream target of adenosine receptor A2b, which leads to stabilization of PER2 during myocardial ischemia and subsequent stabilization of hypoxia-inducible factor-1α and induction of glycolysis. Stabilization of PER2 by intense light can bypass A2b signaling and induce glycolysis and cardioprotection during ischemia [65]. The exact same mutant mice mPer2 were also studied by Virag et al., who showed that mPer2 mice have reduced infarct sizes after non-reperfused MI, after IR, and after preconditioned IR. More studies by different groups are required to elucidate the reasons for the discrepancy [66, 67]. Recently, Rotter et al. demonstrated IR tolerance was greatest in hearts at the transition from wake to rest, and that this tolerance was dependent on protein regulator of calcineurin 1 (Rcan1), whose protein levels oscillate in a circadian fashion and peak at this transition [68]. Nonetheless, the variable susceptibility to ischemic injury suggests a dynamic cellular metabolic and reduction/oxidation environment (Fig. 2). Several initial studies in humans investigating the relationship between time of symptom onset and infarct size in STEMI achieved varying results and were the subject of debate [69–72]. However, more recent studies, including a large study of 6223 STEMI patients treated with primary angioplasty in Switzerland using peak creatine kinase as a measure of infarct size, support the conclusion that myocardial infarct size as well as in-hospital mortality exhibit circadian variation [73–75].

Zhang et al. recently looked at the top 100 best-selling drugs in the US and found that 56 targets (directly bound and affected by the drug) are encoded by a circadian gene, which oscillates at the transcript level [76]. 23 of these 56 drugs have a plasma half-life of less than 6 h, suggesting the time of administration is critical to allow effective interaction with the targets. However, the importance of the chronality of very few of these medications is appreciated. In the list, several widely used cardiovascular medications are included, such as Diovan and Toprol XL. Brand names such as “XL” are used to suggest “extended release” and “long acting”, which may be misleading. Interestingly, the primary target for aspirin, Ptgs1 (COX1, cyclooxygenase-1) oscillates in the heart, lung, and the kidney. Aspirin as a standard therapy for myocardial infarction has a biological half-life of 2–3 h; therefore, given the circadian rhythm of the onset of heart attacks, the optimal window for cardioprotection is fairly narrow. This window of benefits also coincides with previously reported clinical studies that nighttime aspirin is more effective in reducing ambulatory blood pressures [77, 78]. Studies by Hermida et al. showed that the effectiveness of antihypertensives also varies depending on time of day. They found a strong association between reduced BP during sleep, or “dipping”, and ramipril 5 mg daily taken at bedtime as opposed to taking the drug upon awakening, and an increased proportion (from 43 to 65%, P = 0.019) of patients with controlled ambulatory BP at 6 weeks [79]. In the Heart Outcomes Prevention Evaluation (HOPE) study, ramipril administered at night provided benefit three times greater than predicted based solely on BP reduction [80]. A separate study in 2156 individuals with a mean follow-up of 5.6 years, the Ambulatory Blood Pressure Monitoring for Prediction of Cardiovascular Events (MAPEC), concluded from comparing morning and nighttime doses that individuals on nighttime dosing had overall better BP control and that this was associated with lower risk of total cardiovascular disease events (relative risk 0.39, confidence interval 0.29–0.51, P = 0.001) [81, 82].

Although most of the pharmacokinetics occur outside the cardiovascular system itself, our understanding of the chrono-metabolism of these drugs and the chronality of their targets will further improve the efficacy and reduce side effects of these already widely used medications.

The core clock machinery is essentially the same in every cell, yet cells are able to limit rhythmicity at the gene expression level to a particular subset (3–10%) of the transcriptome relevant to the distinct function of the organ in which they reside. This was first demonstrated by the pioneering work of Panda et al. [83]. A recent study by Zhang et al. further substantiated this observation by high-chronological-resolution transcriptomic studies across 12 mouse organs, where they demonstrated that the steady-state RNA level of 43% of mouse protein-coding genes cycle in at least one of the 12 organs examined [76]. This allowed them to suggest that more than half of the protein-coding genome is rhythmic somewhere in the mammalian body at the transcript level.

Panda et al. have predicted that such tissue-specific temporal organization demands the existence of “slave oscillators”, or factors whose rhythmicity is driven by the cell-autonomous core clock and whose expression or activity is tissue specific, allowing the clock in disparate peripheral areas to selectively confer or suppress rhythmicity to the necessary cellular functions unique to that tissue’s function [83]. Although this concept has been proposed for more than a decade, understanding in “slave oscillators” is still poor. Further, although studies have been conducted to search for modifiers of the circadian regulation, most attention has been focused on the “positive” regulators, which promote oscillation, change the duration of the period or increase amplitude. Little efforts have been bestowed upon the “negative” regulators, which inhibit oscillation or reduce amplitude. Conceptually, negative regulators are equally important in attaining this tissue-specific landscape of temporal transcriptomic control.

Similarly, transcriptomic studies in the heart have provided crucial insights into understanding circadian regulation. Since the earliest study by Storch et al., multiple groups have used oligo microarray and RNAseq to characterize the oscillating genes in the heart [76, 84, 85]. It is generally agreed that like other well-characterized tissues, ~10% of the genes expressed in the heart oscillate. Young et al. reported microarray analysis of CBK mice, which allowed identification of BMAL1-dependent oscillating genes and subsequent comparisons to pre-existing system-level transcriptional regulatory databases including gene expression, presence of a putative BMAL1/CLOCK binding site, and prior confirmed BMAL1 binding in the regulatory region in mouse livers via ChIPseq (chromatin immunoprecipitation and deep sequencing), to identify direct BMAL1 targets in the heart and found 19 putative BMAL1 target genes of which 15 of these were also differentially expressed in CCM mice, suggesting they are highly likely to be bona fide targets for BMAL1. These targets include several relevant metabolic genes such as Nampt and Dgat2, highlighting the direct regulation of core clock machinery on metabolic targets [38, 86].

More recently, we demonstrated the mechanistic principle of “positive” and “negative” regulation in circadian gene regulation by studying the cardiac-specific null mice of transcription factor KLF15 (Kruppel-like factor 15), which is a direct target of BMAL1 in the core clock machinery [85, 87]. Using transcriptomic approaches in the heart in a circadian fashion, we showed that Kruppel-like factor 15 (KLF15) directly regulates large numbers of oscillatory genes in the heart (75% of all genes). Importantly, this study provided the first evidence for negative regulation by KLF15 through engaging the circadian repressor REV-ERBα [85]. This prevents aberrant fluctuation of genes, which may oscillate in other tissues or organs. KLF15 is likely a bona fide “slave oscillator” in the heart, for which it not only promotes oscillation but also suppresses oscillation. This corroborates and extends the original concept of the “slave oscillator”.

Among the metabolic enzymes that were found to be oscillating, several of them share an interesting connection with nicotinamide adenine dinucleotide (NAD+), which include the rate-limiting enzyme in the NAD(+) biosynthesis salvage pathway, nicotinamide phosphoribosyltransferase (NAMPT), the mitochondria oxidation–phosphorylation enzyme, succinate dehydrogenase (SDHa), and the mitochondria tricarboxylic acid cycle enzyme, isocitrate dehydrogenase (IDH3g). Regulated by the core clock machinery, cellular levels of NAD(+) oscillate in a circadian fashion; indeed, CLOCK binds rhythmically to the Nampt promoter to influence its expression [88, 89]. NAD(+) is a critical co-enzyme of many cellular enzymes and its levels in turn influence many cellular processes. Sirtuins are NAD(+)-dependent protein deacetylases, which are involved in a broad array of cellular activities, including cellular metabolism. In the liver, Sirtuin 1 (SIRT1) is required for the high-magnitude oscillation of several core clock genes by promoting the deacetylation and degradation of PER2 and BMAL1, thus coupling energy metabolism with the circadian clock [90, 91]. SIRT1 can further influence metabolism through modulation of PPARα, FOXO3, and PGC1α [92–94]. In addition, SIRT6 controls circadian chromatin recruitment of SREBP-1 in the liver resulting in cyclic expression of genes involved in fatty acid and cholesterol metabolism, whilst SIRT3 regulates the rhythmic acetylation and activity of oxidative enzymes and respiration in mitochondria [95, 96].

Although most of the findings are in metabolic organs other than the heart, such as the liver and the white adipose tissue, NAD+ levels likely play similar roles in rhythmic metabolic changes in the heart, as NAD levels also oscillate in cardiomyocytes, and this oscillation is attenuated in CCM mice [97, 98]. Sirtuins also have roles in regulation of metabolism in the heart; loss of SIRT1 abolished the effects of PPARα on cardiomyocyte fatty acid oxidation, and SIRT3 interacts with acetyl-CoA synthetase 2, a cardiac-enriched mitochondrial enzyme promoting entry of acetyl-CoA into the TCA cycle [99, 100].

NAD+ as a cofactor serves as a potential link between the circadian clock and numerous metabolic enzymes in the heart. SDH and IDH3 are two intriguing examples. In addition to their cofactor levels, they also oscillate at the transcript level. SDH and IDH3 are critical enzymes in the TCA cycle and SDH is also complex II in the oxidative phosphorylation chain. Their rhythmicities highlight the importance of circadian regulation in oxidative metabolism in the heart; however, the exact evolutionary benefits of this additional layer of regulation are yet to be discovered. Furthermore, SIRT3 post-translationally modifies SDH, providing evidence of extensive cross-regulation among metabolic pathways; indeed, none of the above mechanisms ought to be considered in isolation [101].

Metabolism is a recurrent theme in circadian regulation. Elegant studies by Wang et al. have elucidated one logic of circadian regulation as a strategy to economize costly gene expression. The oscillatory genes tend to be highly expressed genes of the given tissue and through upregulating them at the required time and offsetting the peak with the trough, the overall energetic cost remains relatively low [102]. Thus, cyclic gene expression is intrinsically connected to energy utilization. As translation is costlier than transcription, it makes sense to exert control at the level of transcription. Direct comparison of the transcriptomic and proteomic dataset is required to further clarify this hypothesis.

Recent studies in mouse liver and drosophila head have demonstrated that only ~30% of the steady-state “oscillation” was a direct result of transcription [103, 104]. In the near future, it is imperative to use novel techniques such as nascent transcript RNAseq and RNA half-life measurements to further dissect the mechanisms by which precise circadian gene regulation is achieved.

Proteomic studies are still limited by sensitivity; for example, in a transcriptomic study usually several thousand genes are identified as oscillatory, but only a few hundred may be identified in a proteomic study of the same tissue [105, 106]. However, proteomics offers information closer to functionality, as only about half of the oscillatory genes overlap when comparing transcriptomic and proteomic studies. Podobed et al. used two-dimensional gel electrophoresis and mass spectrometry to investigate the circadian proteomics of the heart and identified 8% (90/1147) of the soluble cardiac proteome as “oscillatory”. Importantly, they demonstrated that the CCM hearts had 4% (56/1471) of the soluble proteome which was differentially expressed from the wild type, including several key enzymes regulating multiple catabolic pathways such as mitochondrial pyruvate dehydrogenase E1a, aspartate aminotransferase, mitochondrial dihydrolipoyllysine succinyltransferase of 2-oxoglutarate dehydrogenase, aldehyde dehydrogenase, l-lactate dehydrogenase, enoyl-CoA hydratase, and d-β-hydroxybutyrate dehydrogenase [40]. This further strengthens previous observations from the transcriptomic studies suggesting the clock regulates catabolism to optimize active phase energy utilization.

Organelle level proteomics has greatly increased the sensitivity in quantifying low-abundance proteins. In the liver, almost 40% of the mitochondrial proteome and 10% of the nuclear proteome were found to oscillate in a circadian fashion [107, 108]. We look forward to similar studies in the heart.

Protein O-linked β-N-acetylglucosamine (O-GlcNAc) is a reversible covalent post-translational modification catalyzed by O-GlcNAc transferase (OGT) using fructose 6-phosphate and glutamine as substrates and removed/hydrolyzed by O-GlcNAcase (OGA). It, therefore, may serve as a nutrient sensor for glucose and amino acid availability. Like phosphorylation, O-GlcNAcylation affects every aspect of its target protein, including subcellular localization, half-life, and activity [109–111]. Importantly, cardiac O-GlcNAc levels as well as O-GlcNac transferase OGT levels are circadian, observations which are not found in CCM hearts. Many important metabolic regulators are subjected to O-GlcNAc modification, such as AMPK, AKT, and GSK3β. Further, BMAL1 can be modified with O-GlcNAc and pharmacologically increasing cellular protein O-GlcNAcylation leads to induced BMAL1 mRNA and reduced PER2 protein, although the precise mechanism is yet to be elucidated [112].

Other nutrient-sensing transcriptional regulatory pathways and protein kinases are also potential candidates to bridge cellular metabolism to the core clock machinery. For example, 5′ AMP-activated protein kinase (AMPK) responds to AMP:ATP ratio and thus is a sensitive gauge for cellular energy status. AMPK activity in mouse livers is circadian and AMPK phosphorylates CRY1 to rhythmically alter CRY1 protein levels, providing a direct link between AMP/ATP ratio and the circadian clock [113]. In the heart, it has been observed that inhibition of hormone-sensitive lipase through AMPK-dependent phosphorylation has a diurnal variation, which was reduced in CCM mice, suggesting that AMP/ATP ratio remains intimately linked to the cardiomyocyte peripheral clock [98]. These molecular changes likely mediate the circadian variations in triglyceride turnover and lipolysis observed in WT mice and which are suppressed in CCM mice.

A recent report of phosphoproteomics in the liver demonstrated that at least a quarter of the phosphoproteome and more than 40% of the phosphoproteins are regulated in a circadian manner with large amplitude, which enables rapid and economic temporal regulation [114]. It would not be surprising if circadian regulation of the phosphoproteome also plays a significant role in the heart.