Obesity has become a serious and growing public

Transcription

1 REVIEW Metabolism and Circadian Rhythms Implications for Obesity Oren Froy Institute of Biochemistry, Food Science, and Nutrition, Robert H. Smith Faculty of Agriculture, Food and Environment, The Hebrew University of Jerusalem, Rehovot 76100, Israel I. Introduction II. Circadian Rhythms A. Circadian rhythms in mammals B. Circadian rhythms, well-being, and life span III. The Biological Clock A. The location of the mammalian biological clock B. The biological clock at the molecular level C. Circadian rhythms in peripheral clocks IV. The Biological Clock and Energy Homeostasis A. SCN efferents B. SCN afferents C. Effect of neuropeptides and hormones on SCN D. Circadian rhythms and metabolism E. Effect of metabolism on circadian rhythms F. Effect of feeding regimens on circadian rhythms V. The Biological Clock and Obesity A. SCN connection to adipose tissue B. Effect of circadian rhythms on lipid metabolism C. Effect of high-fat diet on circadian rhythms D. Circadian rhythms and body weight E. Clock mutations and metabolic disorders F. Sleep, shift work, and obesity VI. Summary and Conclusions ISSN Print X ISSN Online Printed in U.S.A. Copyright 2010 by The Endocrine Society doi: /er Received April 16, Accepted August 24, First Published Online October 23, 2009 Obesity has become a serious public health problem and a major risk factor for the development of illnesses, such as insulin resistance and hypertension. Human homeostatic systems have adapted to daily changes in light and dark in a way that the body anticipates the sleep and activity periods. Mammals have developed an endogenous circadian clock located in the suprachiasmatic nuclei of the anterior hypothalamus that responds to the environmental light-dark cycle. Similar clocks have been found in peripheral tissues, such as the liver, intestine, and adipose tissue, regulating cellular and physiological functions. The circadian clock has been reported to regulate metabolism and energy homeostasis in the liver and other peripheral tissues. This is achieved by mediating the expression and/or activity of certain metabolic enzymes and transport systems. In return, key metabolic enzymes and transcription activators interact with and affect the core clock mechanism. In addition, the core clock mechanism has been shown to be linked with lipogenic and adipogenic pathways. Animals with mutations in clock genes that disrupt cellular rhythmicity have provided evidence for the relationship between the circadian clock and metabolic homeostasis. In addition, clinical studies in shift workers and obese patients accentuate the link between the circadian clock and metabolism. This review will focus on the interconnection between the circadian clock and metabolism, with implications for obesity and how the circadian clock is influenced by hormones, nutrients, and timed meals. (Endocrine Reviews 31: 1 24, 2010) I. Introduction Obesity has become a serious and growing public health problem (1). Previous ways to combat obesity have failed, and new approaches need to be taken. The biological clock regulates the expression and/or activity of enzymes and hormones involved in metabolism. However, recently, there is a growing body of evidence that metabolism, food consumption, timed meals, and some nutrients feed back to entrain circadian clocks. Moreover, disruption of circadian rhythms leads to metabolic disorders. This review will summarize recent findings concern- Abbreviations: ACS1, Fatty acyl-coenzyme A synthetase 1; ADRP, adipocyte differentiation-related protein; AgRP, agouti-related protein; AMPK, AMP-activated protein kinase; ap2, adipocyte fatty acid-binding protein 2; ARC, arcuate nucleus; BMAL1, brain and muscle-arnt-like 1; BMI, body mass index; C/EBP, CCAAT enhancer binding protein ; CKI, casein kinase I ; CLOCK, circadian locomotor output cycles kaput; CR, calorie restriction; CRY, cryptochrome 1; DMH, dorsomedial hypothalamus; FATP1, fatty acid transport protein 1; FFA, free fatty acid; IF, intermittent fasting; MCH, melanin-concentrating hormone; MC4R, melanocortin 4 receptor; MPOA, medial preoptic area; NAD, nicotinamide adenine dinucleotide; NPAS2, neuronal PAS domain protein 2; NREM, nonrapid eye movement; NPY, neuropeptide Y; PAS, PER, ARNT, SIM; PER, Period; PGC-1, peroxisome proliferator-activated receptor-coactivator 1 ; POMC, proopiomelanocortin; PPAR, peroxisome proliferator-activated receptor; PPRE, peroxisome-proliferator response element; PVN, paraventricular nucleus; REV-ERB, reverse erythroblastosis virus ; RF, restricted feeding; RHT, retinohypothalamic tract; ROR, retinoic acid receptor-related orphan receptor; RORE, ROR response element; SCN, suprachiasmatic nuclei; SPZ, subparaventricular zone; SREBP-1a, sterol regulatory element-binding protein 1a; vmarc, ventromedial ARC; VMH, ventromedial hypothalamus; WAT, white adipose tissue. Endocrine Reviews, February 2010, 31(1):1 24 edrv.endojournals.org 1

2 2 Froy Circadian Rhythms, Metabolism, and Obesity Endocrine Reviews, February 2010, 31(1):1 24 ing the relationship between metabolism and circadian rhythms with implications for obesity. II. Circadian Rhythms A. Circadian rhythms in mammals The rotation of earth around its axis imparts light and dark cycles of 24 h. Organisms on earth developed the ability to predict these cycles and evolved to restrict their activity to the night or day, being nocturnal or diurnal, respectively. By developing an endogenous circadian circa (about) and dies (day) clock that is entrained to external stimuli, animals and plants ensure that physiological processes are performed at the appropriate, optimal time of day or night (2). In mammals, the circadian clock influences nearly all aspects of physiology and behavior, including sleep-wake cycles, cardiovascular activity, endocrine system, body temperature, renal activity, physiology of the gastrointestinal tract, and hepatic metabolism (2, 3). B. Circadian rhythms, well-being, and life span The control of the circadian clock over human pathophysiology is demonstrated in epidemiological studies. Clinical epidemiology in humans indicates that myocardial infarction, pulmonary edema, hypertensive crises, and asthma and allergic rhinitis attacks all peak at certain times during the day (4 6). A number of other disorders, e.g., psychological and sleep disorders, are associated with irregular or pathological functioning of the central biological clock (3). In this day and age, there is a need to extend wakefulness or repeatedly invert the normal sleepwake cycle. As a result, travelers experience the condition known as jet lag, with its associated symptoms of fatigue, disorientation, and insomnia. Similarly, shift workers exhibit altered nighttime melatonin levels and reproductive hormone profiles that could increase the risk of hormonerelated diseases (7). These findings reveal the negative effect of disruption of circadian rhythms on physiology. Disruption of circadian coordination has also been found to accelerate cancer proneness and malignant growth, suggesting that the circadian clock controls tumor progression (7 9). Also, chronic reversal of the external light-dark cycle at weekly intervals resulted in a significant decrease in the survival time of cardiomyopathic hamsters (10). Circadian rhythms change with normal aging, including a shift in the phase and decrease in amplitude (11 13). Indeed, longevity in hamsters is decreased with a disruption of rhythmicity and is increased in older animals given fetal suprachiasmatic implants that restore higher amplitude rhythms (14). Thus, disruption of circadian coordination may be manifested by hormone imbalance, psychological and sleep disorders, cancer proneness, and reduced life span (3, 7 10, 15). In contrast, resetting of circadian rhythms has led to well-being and increased longevity (14, 16). These findings reveal the prominent influence of the circadian clock on human physiology and pathophysiology. III. The Biological Clock A. The location of the mammalian biological clock In mammals, the central circadian clock is located in the suprachiasmatic nuclei (SCN) of the anterior hypothalamus in the brain. The SCN clock is composed of multiple, single-cell circadian oscillators, which, when synchronized, generate coordinated circadian outputs that regulate overt rhythms (17 20). Similar clock oscillators have been found in peripheral tissues, such as the liver, intestine, heart, and retina (3, 21 23) (Fig. 1). SCN oscillation is not exactly 24 h; therefore, it is necessary to entrain the circadian pacemaker each day to the external light-dark cycle to prevent drifting (or free-running) out of phase. Light is the most potent synchronizer for the SCN (24). Light is perceived by the retina, and the signal is transmitted via the retinohypothalamic tract (RHT) to the SCN (3, 25, 26). As a result, vasoactive intestinal polypeptide, an intrinsic SCN factor, acutely activates and synchronizes SCN neurons and coordinates behavioral rhythms (27, 28). The SCN sends signals to peripheral oscillators to prevent the dampening of circadian rhythms in these tissues. The SCN accomplishes this task via neuronal connections or circulating humoral factors (29) (Fig. 1), although the mechanisms are not fully understood. Several humoral factors expressed cyclically by the SCN, such as TGF (30), prokineticin 2 (31), and cardiotrophin-like Light Food, Feeding regimens SCN Adipose tissue Liver Muscle Hormones, Metabolic pathways Autonomic innervation, Humoral factors Locomotor activity, Sleep-wake cycle, Blood pressure Fig. 1. Resetting signals of the central and peripheral clocks. Light is absorbed through the retina and is transmitted to the SCN via the RHT. The SCN then dictates the entrainment of peripheral oscillators via humoral factors or autonomic innervation. As a result, tissue-specific hormone expression and secretion and metabolic pathways exhibit circadian oscillation. In addition, the SCN dictates rhythms of locomotor activity, sleep-wake cycle, blood pressure, and body temperature. Food and feeding regimens affect either peripheral clocks or the central clock in the SCN.

3 Endocrine Reviews, February 2010, 31(1):1 24 edrv.endojournals.org 3 cytokine (32), have been shown to affect peripheral clocks because their intracerebroventricular injection inhibited nocturnal locomotor activity. In turn, SCN rhythms can be altered by neuronal and endocrine inputs (33) (see Section IV.B). Complete destruction of SCN neurons abolishes overall circadian rhythmicity in the periphery, because of loss of synchrony among individual cells and damping of the rhythm at the population level. However, at the cellular level each cell oscillates, but with a different phase (34, 35). Thus, the central circadian clock (often termed the master clock) is located within the SCN (36, 37), whereas peripheral clocks (often termed slave clocks) are found within non-scn cells of the organism, including other regions of the central nervous system (3, 21 23, 38). B. The biological clock at the molecular level In mammals, the clock is an intracellular mechanism sharing the same molecular components in SCN neurons and peripheral cells (39). Generation of circadian rhythms is dependent on the concerted coexpression of specific clock genes. Transcriptional-translational feedback loops lie at the very heart of the core clock mechanism. Many clock gene products function as transcription factors that possess PAS (PER, ARNT, SIM) and basic helix-loop-helix domains involved in protein-protein and protein-dna interactions, respectively. These factors ultimately activate or repress their own expression and, thus, constitute a self-sustained transcriptional feedback loop. Changes in concentration, subcellular localization, posttranslational modifications (phosphorylation, acetylation, deacetylation, SUMOylation), and delays between transcription and translation lead to the approximately 24-h cycle (2, 3, 40, 41). The genes encoding the core clock mechanisms include circadian locomotor output cycles kaput (Clock), brain and muscle-arnt-like 1 (Bmal1), Period1 (Per1), Period2 (Per2), Period3 (Per3), Cryptochrome1 (Cry1), and Cryptochrome2 (Cry2). In the mouse, the first clock gene identified encodes the transcription factor CLOCK (42), which dimerizes with BMAL1 to activate transcription (Fig. 2). CLOCK and BMAL1, two basic helix-loop-helix-pas transcription factors, are capable of activating transcription upon binding to E-box (5 - CACGTG-3 ) and E-boxlike promoter sequences (3). BMAL1 can also dimerize with other CLOCK homologs, such as neuronal PAS domain protein 2 (NPAS2), to activate transcription and sustain rhythmicity (43, 44). PERIOD (PER1, PER2, and PER3) and two CRYPTOCHROME (CRY1 and CRY2) proteins operate as negative regulators (20, 45, 46) (Fig. 2). Thus, CLOCK:BMAL1 heterodimers bind to E-box sequences and mediate transcription of a large number of genes, including those of the negative feedback loop Pers and Crys. When PERs and CRYs are produced in the cytoplasm, they oligomerize and translocate to the nucleus to inhibit CLOCK:BMAL1-mediated transcription (Fig. 2). Pers and Bmal1 have robust oscillation in opposite phases correlating with their opposing functions (38). All the aforementioned clock genes exhibit a 24-h rhythm in SCN cells and peripheral tissues, except for Clock, which has been shown not to oscillate in the SCN (40). Recent studies have demonstrated that CLOCK has intrinsic histone acetyltransferase activity, suggesting that rhythmic activation of chromatin remodeling may underlie the clock transcriptional network (47, 48). Indeed, cyclic histone acetylation and methylation have been observed on the promoters of several clock genes (48 53). Several other players appear to be important to sustain clock function. Casein kinase I epsilon (CKI ) is thought to phosphorylate the PER proteins and, thereby, enhance their instability and degradation (40, 54 56). CKI also phosphorylates and partially activates the transcription factor BMAL1 (57). Bmal1 expression is negatively regulated by the transcription factor reverse erythroblastosis virus (REV-ERB ) (58), which recruits histone deacetylase complexes (59). Bmal1 expression is positively regulated by retinoic acid receptor-related orphan receptor (ROR ) and ROR (60) via the ROR response element (RORE) (61). Thus, Bmal1 oscillation is driven by a rhythmic change in RORE occupancy by RORs and REV-ERB. This alternating promoter occupancy occurs because REV-ERB levels are robustly rhythmic, a result of direct transcriptional activation of the Rev-erb gene by the heterodimer CLOCK:BMAL1 (58) (Fig. 2). C. Circadian rhythms in peripheral clocks The fraction of cyclically expressed transcripts in each peripheral tissue ranges between 5 and 20% of the total population, and the vast majority of these genes are tissuespecific (2, 23, 62 69). These findings emphasize the circadian control over a large portion of the transcriptomes in peripheral tissues. Considering the circadian gene expression in peripheral tissues, it is difficult to determine whether the SCN clock drives these rhythmic patterns directly or indirectly by driving rhythmic feeding, activity, and/or body temperature, which, in turn, contribute to driving rhythms in gene expression in the periphery. In an elegant study, it was shown that tetracycline-responsive, hepatocyte-specific overexpression of REV-ERB, which caused constitutive repression of Bmal1 transcription (see Section III.B) when the tetracycline analog doxycycline was absent, resulted in the loss of clock function. Microarray analysis of livers after removal of doxycycline from the food revealed that the great majority of the 351 rhythmic genes was abolished, indicating that these genes are normally driven by the local liver clock and not by rhythmic

4 4 Froy Circadian Rhythms, Metabolism, and Obesity Endocrine Reviews, February 2010, 31(1):1 24 CLOCK BMAL1 PERs PERs CRYs CRYs PPRE systemic signals. Interestingly, 31 genes, among which was the core clock gene mper2, still exhibited robust circadian patterns of expression, suggesting that this small subset of rhythmic liver genes is driven by systemic signal independent of the liver clock (70). Thus, for a peripheral tissue, such as the liver, signals from the central SCN clock or the local endogenous clock may control rhythmic gene expression (70, 71). Indeed, as mentioned in Section III.A, peripheral rhythms persist in the periphery at the cellular level even without SCN control (34, 35). IV. The Biological Clock and Energy Homeostasis PPARα CLOCK BMAL1 E-box CLOCK BMAL1 E-box CLOCK BMAL1 E-box CLOCK BMAL1 E-box Rev-erbα Rorα ~ REV-ERBs A. SCN efferents The SCN provides its most intense output to the subparaventricular zone (SPZ) and dorsomedial hypothalamus (DMH) (72, 73). SCN fibers have also been shown to terminate in and around the arcuate nucleus (ARC) in the ~ Per1-3 Cry1-2 CLOCK BMAL1 RORs E-box ~ ~ Clock ~ Bmal1 PPRE RORE ~ Pparα ~ PERs CRYs Nucleus Cytoplasm Fig. 2. The core mechanism of the mammalian circadian clock and its link to energy metabolism. The cellular oscillator is composed of a positive limb (CLOCK and BMAL1) and a negative limb (CRYs and PERs). CLOCK and BMAL1 dimerize in the cytoplasm and translocate to the nucleus. The CLOCK:BMAL1 heterodimer then binds to enhancer E-box sequences located in the promoter region of Per and Cry genes to activate their transcription. After translation, PERs and CRYs undergo nuclear translocation and inhibit CLOCK:BMAL1, resulting in decreased transcription of their own genes. The autoregulatory transcription translation loop comprising CLOCK:BMAL1 and PER CRY constitutes the core clock and generates 24-h rhythms of gene expression. CLOCK:BMAL1 heterodimer also induces the transcription of Rev-erb and Ror. ROR and REV-ERB regulate lipid metabolism and adipogenesis, and also participate in the regulation of Bmal1 expression. ROR stimulates and REV-ERB inhibits Bmal1 transcription, acting through ROR elements (RORE). CLOCK:BMAL1 heterodimer also mediates the transcription of Ppar, a nuclear receptor involved in glucose and lipid metabolism. PPAR activates transcription of Rev-erb by binding to a PPRE. PPAR also induces Bmal1 expression, acting through PPRE located in its promoter. ventromedial hypothalamus (VMH) and in the ventral part of the lateral hypothalamus, suggesting an interaction with areas involved in food intake and organization of activity (74) (Fig. 3). The SPZ and DMH, which are innervated by the SCN, have many outputs to other regions in the brain, including the paraventricular nucleus (PVN), the lateral hypothalamus, ventrolateral preoptic nucleus, and medial preoptic area (MPOA), that regulate corticosteroid release, wakefulness/feeding, sleep, and thermoregulation, respectively (Fig. 3). The role of the SPZ and DMH in regulating circadian rhythms was determined using lesion studies (75, 76). Destruction of the ventral SPZ reduced circadian rhythms of sleep-wakefulness and locomotor activity but had little effect on circadian regulation of body temperature (76). Conversely, degeneration of the dorsal SPZ disrupted circadian regulation of body temperature with minimal effect on sleep-wakefulness and locomotor activity (76). Thus, ventral SPZ regulates sleepwakefulness, whereas dorsal SPZ regulates body temper-

5 Endocrine Reviews, February 2010, 31(1):1 24 edrv.endojournals.org 5 VLPO RHT MPOA Ghrelin, Insulin, Leptin, Glucose PVN SPZ SCN DMH VMH NPY/AgRP POMC ARC Metabolism, Hepatic glucose production, Hormone secretion IGL ature (72). Ablation of DMH cell bodies, which receive inputs from both the SCN and the SPZ, resulted in severe impairment of circadian-regulated sleep-wakefulness, locomotor activity, corticosteroid secretion, and feeding (75). Thus, DMH and VMH constitute a gateway between the master pacemaker neurons of the SCN and cell bodies located within brain centers in the hypothalamus (77) (Fig. 3). The ventral and dorsal borders of the PVN, the location of preautonomic nervous system neurons, are selectively innervated by fibers of the SCN (78). The SCN can control energy homeostasis by providing its output to preautonomic neurons in the hypothalamus that are connected to the parasympathetic and sympathetic systems (79). Studies have shown that many interneurons, which project from the SCN to the PVN, contain -aminobutyric acid as neurotransmitter and inhibit the PVN (80, 81). Thus, it seems that the SCN is capable of controlling peripheral tissues not only by the secretion of humoral signals but also by affecting the two branches of the autonomic nervous system, i.e., the sympathetic and parasympathetic systems (Fig. 3). B. SCN afferents There are two major ways by which metabolic information may reach the SCN: 1) the sympathetic and parasympathetic branches of the autonomic nervous system; ORX LH Raphe MCH and 2) hormones or nutrients, such as glucose, that cross the blood-brain barrier. From the sites where visceral sympathetic information enters the brain (the dorsal horn) and visceral parasympathetic information enters the brain (the nucleus of the tractus solitarius), no projections are known to the SCN. Thus, it is possible that autonomic information is transmitted first to the PVN and then to the SCN (Fig. 3). This seems to be different for information circulating in the bloodstream. Areas free of the blood-brain barrier, where metabolites and hormones in the bloodstream can directly reach receptors on neurons, are circumventricular organs. Injection of the neuronal tracer cholera toxin B into the ventromedial ARC (vmarc) resulted in the visualization of an elaborate network of projections to many targets within and outside the hypothalamus (74). The vmarc is considered the site where information from the circulation can reach the hypothalamus, either via its connection with the circumventricular median eminence or through hormones that cross the blood-brain barrier and bind to its membrane-bound receptors. The dense reciprocal interaction between the vmarc and the SCN provides the anatomical basis for the link between circulating metabolic information and the SCN (74). This anatomical connection between vmarc and SCN may form the basis upon which the SCN is informed about circulating hormones and the vmarc about the time of the day (Fig. 3). Gut-derived polypeptides have been shown to affect circadian rhythms. For example, gastrinreleasing peptide, a mediator of both feeding and locomotor activity, mediates light-like resetting of the SCN (82); peptide tyrosine-tyrosine has also been shown to correlate with alterations to wakefulness and sleep architecture (83). However, the effect of peptide tyrosinetyrosine and gastrin-releasing peptide is presumably mediated via vagal afferents that travel through the autonomic nervous system to the SCN rather than directly from the vmarc. Autonomic nervous system Gut-derived hormones, Nutrients, Abdominal distension Fig. 3. SCN afferents and efferents. The SCN can be entrained by light, hormones and nutrients, and neuronal connections (green arrows). The SCN sends neuronal connections to the ARC, MPOA, PVN, and SPZ (blue arrows). Hormones and nutrients may affect the ARC directly. The ARC controls expression of orexins and MCH in LH. The SPZ innervates the DMH, which, in turn, innervates PVN, VLPO, and LH, regulating corticosteroid production, sleep, and feeding, respectively. MPOA, PVN, and DMH through the autonomic nervous system regulate adipose tissue, liver, and other peripheral tissues (red arrows). In turn, gut-derived hormones, nutrients, and abdominal distension signal to the brain through the autonomic nervous system. LH, Lateral hypothalamus; ORX, orexins; Raphe, brainstem raphe nuclei; VLPO, ventrolateralpreoptic area. C. Effect of neuropeptides and hormones on SCN Several brain regions have been associated with energy homeostasis. These regions are the VMH, PVN, DMH, and ARC located at the mediobasal hypothalamus. Destruction of VMH, PVN, and DMH results in obesity, whereas ablation of the lateral hypothalamus results in anorexia (84). At the molecular level, the melanocortin system plays a major role in the neural control of energy homeostasis (85, 86). Leptin, a satiety signal, stimulates

6 6 Froy Circadian Rhythms, Metabolism, and Obesity Endocrine Reviews, February 2010, 31(1):1 24 proopiomelanocortin (POMC)- and cocaine- and amphetamine-regulated transcript-expressing neurons within the ARC to produce -melanocyte-stimulating hormone ( - MSH), which subsequently activates the melanocortin 4 receptor (MC4R) and results in decreased food intake and increased energy expenditure (87, 88). In humans, mutations in the POMC and MC4R genes are associated with morbid obesity (89 94). In parallel, leptin suppresses a distinct set of neuropeptide Y (NPY)- and agouti-related protein (AgRP)-expressing neurons within the ARC. In the absence of leptin, such as during fasted state, NPY/AgRPexpressing neurons release AgRP, an antagonist of the MC4R (95 97) causing decreased energy expenditure and increased appetite (98 102). Thus, agonists ( -melanocyte-stimulating hormone) and antagonists (AgRP) of the MC4R determine the weight-regulating effects of leptin in the central nervous system. Leptin can be the bridge between energy homeostasis and circadian control, due to its circadian oscillation (see Section IV.D) and expression of its receptor in several hypothalamic regions. Receptors for leptin and ghrelin are present on SCN cells (74, 103, 104), so it is possible that these hormones bind directly to SCN neurons, similarly to their effect on NPY/AgRP-neurons. Activation of vmarc neurons by systemic administration of the ghrelin mimetic GH-releasing peptide-6 combined with SCN tracing showed that vmarc neurons transmit feeding-related signals to the SCN (74). This injection induced Fos in the vmarc and resulted in attenuation of light-induced phase delay in mice and light-induced Fos expression in the SCN in rats (105). Administration of ghrelin to SCN slices or SCN explants in vitro caused phase shifts in Per2::luc reporter gene expression. However, administration of ghrelin to wild-type mice only caused phase shifts after 30 h of food deprivation, whereas ip injection of ghrelin did not cause phase shifts in wildtype mice fed ad libitum (106). Thus, it seems that ghrelin and leptin may affect the SCN directly or through their effect on the ARC, which is then relayed to the SCN (Fig. 3). Hypothalamic neuropeptides, such as NPY, AgRP, and POMC, are also expressed according to a pronounced diurnal rhythm, although the extent to which these oscillations are entrained by feeding, light, or nutrient signaling remains uncertain (107). In addition to its being a clockcontrolled output gene, NPY is involved in communicating nonphotic signals to the SCN via the intergeniculate leaflet (108) (Fig. 3). In addition, serotonergic signaling pathways from the raphe nuclei that influence feeding and energy metabolism in the hypothalamus have been shown to modulate both SCN oscillations and sleep (109) (Fig. 3). Therefore, these signaling systems appear to be involved in a feedback loop that links feeding and metabolic state to the SCN. Recently, the NPY/AgRP system has been shown to be further connected to the SCN clock. Disruption of the brain-specific homeobox factor Bsx, encoding a key regulator of hypothalamic NPY/AgRP neuron development (110), yielded mice that were resistant to obesity when crossed with leptin-deficient obese (ob) mice and displayed attenuated onset and dampened amplitude of nocturnal locomotor activity. Further analysis of these mutant mice may shed light on the linkage between the SCN and NPY/AgRP neurons. ARC neurons project to multiple nuclei involved in feeding behavior (99, ), such as the lateral hypothalamus, which produces the hunger-stimulating neuropeptides melanin-concentrating hormone(mch), orexin A, and orexin B ( ) (Fig. 3). Targeted deletion studies of MCH resulted in hypophagic lean mice with a high metabolic rate and demonstrated that MCH acts downstream of leptin and the melanocortin system (117). Orexins A and B are two neuropeptides generated from a single transcript that display a circadian rhythm of expression and are strongly induced by fasting (118, 119). Indeed, mutant mice, in which Clock function is impaired, exhibit significantly higher energy intake and almost complete ablation of rhythmic expression in Cart and Orexin (120). Intracerebroventricular injection of orexin A stimulates food intake acutely in rats, in part through excitation of NPY in the ARC (118, 121). Orexins also play a role in the regulation of sleep-wake rhythms because mutations in the orexin B receptor (119, 122) and deletion of the orexin gene (123) caused narcolepsy and obesity (124, 125). All these findings demonstrate the intricate relationships between circadian rhythms and energy homeostasis. D. Circadian rhythms and metabolism Many hormones involved in metabolism, such as insulin (126), glucagon (127), adiponectin (128), corticosterone (129), leptin, and ghrelin (130, 131), have been shown to exhibit circadian oscillation. Leptin, an adipocyte-derived circulating hormone that acts at specific receptors in the hypothalamus to suppress appetite and increase metabolism, is extremely important in obesity. Leptin exhibits striking circadian patterns in both gene expression and protein secretion, with peaks during the sleep phase in humans (132). Neither feeding time nor adrenalectomy affected the rhythmicity of leptin release. However, ablation of the SCN has been shown to eliminate leptin circadian rhythmicity in rodents, suggesting that the central circadian clock regulates leptin expression (133). In addition, SCN-lesioned rats, as opposed to intact animals, showed no elevation in plasma free fatty acids (FFAs) after ip administration of leptin, suggesting a role for SCN in leptin function (134). In addition to the endocrine control, the circadian clock has been reported to regulate metabolism and energy ho-

7 Endocrine Reviews, February 2010, 31(1):1 24 edrv.endojournals.org 7 meostasis in peripheral tissues (135, 136). This is achieved by mediating the expression and/or activity of certain metabolic enzymes and transport systems (137, 138) involved in cholesterol metabolism, amino acid regulation, drug and toxin metabolism, the citric acid cycle, and glycogen and glucose metabolism (77, 126, ). Some examples are glycogen phosphorylase (142), cytochrome oxidase (143), lactate dehydrogenase (144), acetyl-coa carboxylase (145, 146), malic enzyme, fatty acid synthase, glucose-6-phosphate dehydrogenase (145), and many more. Moreover, lesion of rat SCN abolishes diurnal variations in whole body glucose homeostasis (147), altering not only rhythms in glucose utilization rates but also endogenous hepatic glucose production. Indeed, the SCN projects to the preautonomic PVN neurons to control hepatic glucose production (148). Similarly, glucose uptake and the concentration of the primary cellular metabolic currency ATP in the brain and peripheral tissues have been found to fluctuate around the circadian cycle (139, 148, 149). In addition, a large number of nuclear receptors involved in lipid and glucose metabolism has been found to exhibit circadian expression (150). A B NAD(P)H CLOCK BMAL1 Ac E-box ADP ATP CRYs PERs AMPK Ac AMPK PERs CRYs CLOCK BMAL1 Ac E-box E. Effect of metabolism on circadian rhythms The effect of metabolism on the master or peripheral clocks could arise from feeding, food metabolites, or hormones whose secretion is controlled by food or its absence. Several studies have identified single nutrients capable of resetting or phase-shifting circadian rhythms, such as glucose ( ), amino acids (154), sodium (155, 156), ethanol (157, 158), caffeine (159), thiamine (160, 161), and retinoic acid (162, 163). In addition to nutrients, hormones that regulate metabolism can also induce or reset circadian rhythms through regulation of clock gene expression. For example, in Rat-1 fibroblast cultures, insulin causes an acute induction of Per1 mrna production (164). Glucocorticoids were shown to induce circadian gene expression in cultured rat-1 fibroblasts and transiently change the phase of circadian gene expression in liver, kidney, and heart (165, 166). Interestingly, it was recently reported that leptin causes up-regulation of Per2 and Clock gene expression in mouse osteoblasts that exhibit endogenous circadian rhythms (167). Recent experiments have suggested that cellular energy levels are capable of influencing rhythms (168). CLOCK and its homolog NPAS2 can bind efficiently to BMAL1 and consequently to E-box sequences in the presence of reduced nicotinamide adenine dinucleotides (NADH and NADPH) (Fig. 4A). On the other hand, the oxidized forms of the nicotinamide adenine dinucleotides (NAD and NADP ) inhibit DNA binding of CLOCK:BMAL1 or x AMP AMP NAD + NAD + PERs CRYs CLOCK BMAL1 E-box SIRT1 x NAM NAM Ac ADP-r PPRE Ac P PGC-1α REV-ERBs RORE x PPARα PPRE RORs RORE P PGC-1α SIRT1 Ac ADP-r BMAL1 REV-ERBs Fig. 4. The role of NAD /NAD(P)H levels in the core clock mechanism. A, High NAD(P)H levels promote CLOCK:BMAL1 binding to E-box sequences leading to the acetylation of BMAL1 and expression of Pers, Crys, and other clock-controlled genes. The negative feedback loop, PERs: CRYs binds to CLOCK:BMAL1, and consequently PERs are acetylated. When NAD levels rise as a result of AMPK activation by high AMP levels, SIRT1 deacetylates PERs and BMAL1, relieving PERs:CRYs repression and another cycle begins. B, Expression of Bmal1 and Rev-erb genes are controlled by PPAR and binding of RORs to RORE sequences. RORs need a coactivator, PGC-1, which is phosphorylated by AMPK. In parallel, AMPK activation leads to an increase in NAD levels, which, in turn activate SIRT1. SIRT1 activation leads to PGC-1 deacetylation and activation. Ac-ADP-r, Acetyl adenosine diphosphate ribose; NAM, nicotinamide.

8 8 Froy Circadian Rhythms, Metabolism, and Obesity Endocrine Reviews, February 2010, 31(1):1 24 NPAS2:BMAL1 (168, 169). Because the NAD(P) / NAD(P)H redox equilibrium depends on the metabolic state of the cell, this ratio could dictate the binding of CLOCK/NPAS2:BMAL1 to E-boxes and result in phaseshifting of cyclic gene expression (137, 168, 169). However, the role of redox on circadian rhythms needs to be investigated in vivo. In addition to NAD, AMP is another cellular indicator of low energy. Interestingly, AMP-activated protein kinase (AMPK), an important nutrient sensor, has been found to phosphorylate Ser-389 of CKI, resulting in increased CKI activity and degradation of mper2. mper2 degradation leads to a phase advance in the circadian expression pattern of clock genes in wild-type mice (170). In addition, the expression profile of clock-related genes, such as Per1 and Cry2, in skeletal muscle in response to 5-amino-4-imidazole-carboxamide riboside, an AMPK activator, as well as the diurnal shift in energy utilization, is impaired in AMPK 3 subunit knockout mice (171). Because AMPK has been implicated in feeding regulation (172) and it serves as an energy sensor, it could be one of the links to integrate the circadian clock with metabolism (Fig. 4). Another protein, recently found to link metabolism with the circadian clock, is SIRT1. SIRT1 is the mammalian homolog of yeast Sir2, an NAD -dependent histone deacetylase involved in transcriptional silencing and genome stability and a key factor in the longevity response to caloric restriction (173, 174). Recent studies show that SIRT1 interacts directly with CLOCK and deacetylates BMAL1 and PER2 (175, 176) (Fig. 4A). It seems that after binding to E-box, CLOCK and CBP/p300 acetylate histones H3 and H4 (47) and BMAL1 (177). BMAL1 acetylation potentiates its binding by the repressive PER/CRY complex (177), and, as a result, PER2 is acetylated (175). When acetylated, PER2 (175) and possibly BMAL1 (176) are more stable (Fig. 4A). SIRT1 then becomes activated and starts deacetylating BMAL1, PER2, and histones (178). Deacetylated PER2 is further phosphorylated and degraded, and a new cycle begins. It has also been shown that CLOCK:BMAL1 heterodimer regulates the circadian expression of nicotinamide phosphoribosyltransferase, a rate-limiting enzyme in the NAD salvage pathway. SIRT1 is recruited to the Nampt promoter and contributes to the circadian synthesis of its own coenzyme (179). Most recently, it has been shown that AMPK enhances SIRT1 activity by increasing cellular NAD levels, resulting in the deacetylation and modulation of the activity of downstream SIRT1 targets (180) (Fig. 4A). The levels of NAD together with the cycling of SIRT1 determine the activity and robustness of transcription. F. Effect of feeding regimens on circadian rhythms Similarly to the control of the circadian clock on metabolism, feeding is a very potent synchronizer (zeitgeber) for peripheral clocks (Fig. 1). Limiting the time and duration of food availability with no calorie reduction is termed restricted feeding (RF) (39, 137, 181, 182). Animals that receive food ad libitum everyday at the same time for only a few hours adjust to the feeding period within a few days and consume their daily food intake during that limited time (38, 183, 184). Restricting food to a particular time of day has profound effects on the behavior and physiology of animals. Two to 4 h before the meal, the animals display food anticipatory behavior, which is demonstrated by an increase in locomotor activity, body temperature, corticosterone secretion, gastrointestinal motility, and activity of digestive enzymes (181, 183, 185, 186), all of which are known output systems of the biological clock. RF is dominant over the SCN and drives rhythms in arrhythmic and clock mutant mice and animals with lesioned SCN, regardless of the lighting conditions (181, ). In most incidents, RF affects circadian oscillators in peripheral tissues, such as liver, kidney, heart, and pancreas, with no effect on the central pacemaker in the SCN (39, 137, 182, 189, 190, 192, 193). Thus, RF uncouples the SCN from the periphery, suggesting that nutritional regulation of clock oscillators in peripheral tissues may play a direct role in coordinating metabolic oscillations (194). Many physiological activities that are normally dictated by the SCN master clock, such as hepatic P450 activity, body temperature, locomotor activity, and heart rate, are phase-shifted by RF to the time of food availability (141, 188, 189, 195, 196). As soon as food availability returns to normal, the SCN clock, whose phase remains unaffected, resets the peripheral oscillators (192). The location of this food-entrainable oscillator has been elusive. Lesions in the dorsomedial hypothalamic nucleus (DMH) ( ), the brainstem parabrachial nuclei (198, 201), and the core and shell regions of nucleus accumbens (202, 203) revealed that these brain regions may be involved in food-entrainable oscillator output, but they cannot fully account for the oscillation (204). Neither vagal signals nor leptin are critical for the entrainment (205, 206). CLOCK (207) or BMAL1 (208) and other clock genes (209) have been shown not to be necessary for food anticipatory activity. However, it has recently been demonstrated that mper2 mutant mice did not exhibit wheel-running food anticipation (210, 211). Thus, the effect of RF on circadian rhythms warrants further study. Calorie restriction (CR) refers to a dietary regimen low in calories without malnutrition. CR restricts the amount of calories derived from carbohydrates, fats, or proteins to 60 75% of ad libitum-fed animals (212). It has been doc-

9 Endocrine Reviews, February 2010, 31(1):1 24 edrv.endojournals.org 9 umented that calorie restriction significantly extends the life span of rodents by up to 50% (213, 214). In addition to the increase in life span, CR also delays the occurrence of age-associated pathophysiological changes, such as cancer, diabetes, kidney disease, and cataracts ( ). Theories on how CR modulates aging and longevity abound, but the exact mechanism is still unknown (214). As opposed to RF, CR entrains the clock in the SCN ( ), indicating that calorie reduction could affect the central oscillator. CR during the daytime affects the temporal organization of the SCN clockwork and circadian outputs in mice under light-dark cycle. In addition, CR affects photic responses of the circadian system, indicating that energy metabolism modulates gating of photic inputs in mammals (220). These findings suggest that synchronization of peripheral oscillators during CR could be achieved directly due to the temporal eating, as has been reported for RF (189, 192, 193), or by synchronizing the SCN ( ), which in turn sends humoral or neuronal signals to entrain the peripheral tissues (38, 222). During intermittent fasting (IF), food is available ad libitum every other day. IF-treated mice eat on the days they have access to food approximately twice as much as those having continuous access to food (223, 224). Similarly to calorically restricted animals (214), IF-fed animals exhibit increased life span in comparison with the ad libitumfed control (225) as well as improved glucose metabolism, cardio-protection, neuro-protection (223, ), and increased resistance to cancer (224). The IF-induced beneficial effects are thought to occur independently of the overall caloric intake, but the underlying mechanisms are still unknown. One suggested mechanism is stimulation of cellular stress pathways induced by the IF regimen (223, 231, 232). Recently, it has been shown that when food was introduced during the light period, mice exhibited almost arrhythmicity in clock gene expression in the liver. Unlike daytime feeding, nighttime feeding yielded rhythms similar to those generated during ad libitum feeding (233). The fact that IF can affect circadian rhythms differently depending on the timing of food availability suggests that this regimen affects the SCN clock, similarly to CR. SCN resetting by IF and CR could be involved in the health benefits conferred by these regimens (222). V. The Biological Clock and Obesity A. SCN connection to adipose tissue The daily rhythm in adipose leptin production strongly suggests a direct control of adipose tissue activity by the biological clock (133). Indeed, injection of the pseudorabies virus into white adipose tissue (WAT) led to labeling in SCN neurons (234, 235). WAT is innervated by the sympathetic nervous system leading to the mobilization of lipid stores (236, 237) and by the parasympathetic nervous system resulting in anabolism (238). To test whether the SCN uses its projections to preautonomic PVN neurons to control the mobilization of lipid stores, similarly to its control over hepatic glucose production (see Section IV.A), the -aminobutyric acid-antagonist bicucilline was infused into the PVN, and plasma glucose, leptin, and FFA levels were measured (148). Contrary to plasma glucose concentrations, plasma FFA and plasma leptin concentrations were not affected by bicucilline treatment in the PVN. Also, PVN lesions did not attenuate fasting-induced lipid mobilization (148). Viral tracing from WAT, besides the PVN, was found especially in the MPOA, an area implicated in lipid metabolism, the DMH, and the ARC (234, 235). Thus, it seems that the SCN uses different outputs to control glucose (via the PVN) and lipid (via the MPOA) metabolism (148). B. Effect of circadian rhythms on lipid metabolism Circadian clocks have been shown to be present in inguinal WAT, epididymal WAT, and brown adipose tissue (67, 239, 240). Diurnal variations in the sensitivity of adipose tissue to adrenaline-induced lipolysis persist ex vivo, suggesting that the intrinsic nature of the adipocyte exhibits a diurnal variation (241). Recent transcriptome studies revealed rhythmic expression of clock and adipokine genes, such as resistin, adiponectin, and visfatin, in visceral fat tissue (128). The expression of these mediators is blunted in obese patients (133, 242, 243). Fatty acid transport protein 1 (Fatp1), fatty acyl-coa synthetase 1 (Acs1), and adipocyte differentiation-related protein (Adrp) exhibit diurnal variations in expression, suggesting that nocturnal expression of FATP1, ACS1, and ADRP will promote higher rates of fatty acid uptake and storage of triglyceride in rodents (244). Recent molecular studies established the involvement of BMAL1 activity in the control of adipogenesis and lipid metabolism in mature adipocytes. Embryonic fibroblasts from Bmal1 / knockout mice failed to differentiate into adipocytes. Loss of BMAL1 expression led to a significant decrease in adipogenesis and gene expression of several key adipogenic/lipogenic factors peroxisome proliferator-activated receptor (PPAR) 2, adipocyte fatty acidbinding protein 2 (ap2), CCAAT enhancer binding protein (C/EBP ), C/EBP, sterol regulatory element-binding protein 1a (SREBP-1a), phosphoenolpyruvate carboxykinase, fatty acid synthase. Furthermore, overexpression of BMAL1 in adipocytes increased lipid synthesis activity. These results indicate that BMAL1, a master regulator of circadian rhythm, also plays important roles in the regulation of adipose differentiation and lipogenesis in mature adipocytes (245).

10 10 Froy Circadian Rhythms, Metabolism, and Obesity Endocrine Reviews, February 2010, 31(1):1 24 Recently, a comprehensive survey of nuclear receptor mrna profiles in white and brown adipose tissue, liver, and skeletal muscle in mice revealed that approximately 50% of the known nuclear receptors exhibit rhythmic expression (150). Because these receptors sense various lipids, vitamins, and fat-soluble hormones, they serve as direct links between nutrient-sensing pathways and the circadian control of gene expression. The circadian rhythmicity of a nuclear receptor family member, PPAR, provides an example of a reciprocal link between circadian and lipid metabolic processes. The CLOCK:BMAL heterodimer mediates transcription of PPAR, which subsequently binds to the peroxisome-proliferator response element (PPRE) and activates transcription of Bmal1 ( ) (Figs. 1 and 4B). Bmal1 has also been shown to be regulated by PPAR in cells of the aorta (249). PPAR regulates the transcription of genes involved in lipid and glucose metabolism upon binding of endogenous FFAs. Thus, PPAR may play a unique role at the intersection of circadian and lipid metabolic pathways. Another example for the relationship between nuclear receptors and the biological clock is seen with retinoic acid. Retinoic acid has been shown to up-regulate Per1 and Per2 expression in an E-box-dependent manner in mouse fibroblast NIH3T3 cells (163). Similarly, retinoic acid can phase-shift Per2 expression in vivo and in seruminduced smooth muscle cells in vitro (162). However, when retinoic acid is administered to cells expressing the retinoic acid receptors RAR or RXR, the ligand-receptor complex competes with BMAL1 for binding to CLOCK or NPAS2 in vascular cells. These interactions negatively regulate CLOCK/NPAS2:BMAL1-mediated transcriptional activation of clock gene expression (162, 163). An important candidate to link between the circadian clock and lipid metabolism is REV-ERB. This proadipogenic transcription factor, whose levels increase dramatically during adipocyte differentiation (250), exhibits striking diurnal variations in expression in murine adipose tissue (244) and rat liver (251). During adipocyte differentiation, REV-ERB has been shown to act downstream of the differentiation factor PPAR by facilitating gene expression of PPAR target genes, including Ap2 and C/ebp, but it has no effect on C/ebp or SREBP-1 gene expression (252, 253). Ectopic REV-ERB expression in 3T3L1 preadipocytes promotes their differentiation into mature adipocytes (252). In addition to its role in lipid metabolism and adipocyte differentiation, REV-ERB is a negative regulator of Bmal1 expression (58), as mentioned above (Figs. 1 and 4B). In contrast, ROR, which regulates lipogenesis and lipid storage in skeletal muscle, is a positive regulator of Bmal1 expression (60, 254, 255) (Figs. 1 and 4B). Interestingly, CLOCK:BMAL1 heterodimer regulates the expression of both Rev-erb and Ror (58, 60, 256) (Figs. 1 and 4B). Mice deficient in ROR or REV-ERB have impaired circadian rhythms of locomotor activity and clock gene expression (58, 60). The PPAR coactivator, PGC-1 (peroxisome proliferator-activated receptor-coactivator 1a), a transcriptional coactivator that regulates energy metabolism, is rhythmically expressed in the liver and skeletal muscle of mice. PGC-1 stimulates the expression of Bmal1 and Rev-erb through coactivation of the ROR family of orphan nuclear receptors (257, 258) (Fig. 4B). Mice lacking PGC-1 show abnormal diurnal rhythms of activity, body temperature, and metabolic rate due to aberrant expression of clock genes and those involved in energy metabolism. Analyses of PGC-1 -deficient fibroblasts and mice with liver-specific knockdown of PGC-1 indicate that it is required for cell-autonomous clock function (257). Acetylated PGC-1 is also a substrate for SIRT1 (see Section IV.E) (180) (Fig. 4B). Thus, PPAR, PPAR, REV-ERB, ROR, and PGC-1 are key components of the circadian oscillator that integrate the mammalian clock and lipid metabolism. The interconnection between the clock core mechanism and lipogenic and adipogenic pathways emphasizes why clock disruption leads to metabolic disorders (see Section V.E). C. Effect of high-fat diet on circadian rhythms Few studies show that a high-fat diet leads to minimal effects on the rhythmic expression of clock genes in visceral adipose tissue and liver (259, 260). However, recent studies have shown that introduction of a high-fat diet to animals leads to rapid changes in both the period of locomotor activity in constant darkness and to increased food intake during the normal rest period under light-dark conditions (261). These changes in behavioral rhythmicity correlated with disrupted clock gene expression within hypothalamus, liver, and adipose tissue, as well as with altered cycling of hormones and nuclear hormone receptors involved in fuel utilization, such as leptin, TSH, and testosterone in mice, rats, and humans ( ). Furthermore, a high-fat diet modulates carbohydrate metabolism by amplifying circadian variation in glucose tolerance and insulin sensitivity (267). In addition to the disruption of clock gene expression, high-fat diet induced a phase delay in clock and clockcontrolled genes (265). Recently, AMPK has been found to phosphorylate Ser-389 of CKI, resulting in increased CKI activity and degradation of mper2. mper2 degradation leads to a phase advance in the circadian expression pattern of clock genes in wild-type mice (170) (see Section IV.E). As the levels of mampk decline under a high-fat diet (265), it is plausible that the changes seen in the expression phase of genes under a high-fat diet are mediated by changes in AMPK levels. In addition to its effect on gene