The food-entrainable oscillator

Restriction of food availability to a narrow time window each day results in profound reorganisation of behaviour and physiology⁽ Reference Stephan ^{58 ,} Reference Mistlberger ^{59 )}. Such temporal restriction induces a bout of activity in advance of food availability, termed ‘food anticipatory activity’ (FAA). This phenomenon was originally observed in rats, but has since been reported in multiple vertebrate and invertebrate species. More detailed analysis in rodents reveals that this FAA is also accompanied by physiological changes including increased core body temperature and serum glucocorticoid concentration.

Interestingly, FAA exhibits properties that are consistent with it being controlled by endogenous circadian clock(s), rather than being merely a food-driven phenomenon. For example, if the temporal window of food availability is abruptly delayed, the onset of FAA takes multiple 24 h cycles to resynchronise to the new feeding time. Moreover, if an animal is completely food deprived, FAA persists at approximately the same time every 24 h for as long as the food deprivation can be maintained⁽ Reference Stephan ^{58 ,} Reference Mistlberger ^{59 )}.

The underlying circadian basis of FAA has led to the postulation that animals contain a food-entrainable oscillator (FEO). Although there are reported differences in FAA in mice lacking genetic components of the circadian clock⁽ Reference Challet, Mendoza and Dardente ^{60 )}, these mice do retain the ability to display FAA⁽ Reference Storch and Weitz ^{61 )}. It is therefore believed that the genetic control of the FEO differs from other circadian processes. In keeping with this idea, food anticipatory responses persist in SCN-lesioned animals⁽ Reference Krieger, Hauser and Krey ^{62 ,} Reference Stephan ^{63 )}, indicating that the FEO resides in tissue(s) outside of the SCN, the master circadian clock. Some studies have suggested that the FEO may be closely linked to the dorsomedial hypothalamic nuclei, a brain region known to be involved in the homeostatic regulation of feeding⁽ Reference Mieda, Williams and Richardson ^{64 ,} Reference Fuller, Lu and Saper ^{65 )}. Other work has highlighted the potential role of extra-hypothalamic brain regions in FAA⁽ Reference Mendoza, Pevet and Felder-Schmittbuhl ^{66 )}. The anatomical localisation of the FEO remains a controversial topic, however⁽ Reference Mistlberger, Buijs and Challet ^{67 –} Reference Landry, Kent and Patton ^{69 )}. Indeed the FEO may lie outside of the brain or require the interplay between multiple tissues.

Regulation of rhythms in peripheral tissues

One potential mechanism underlying food-entrainable rhythmicity is the effect of feeding time on peripheral tissue clocks. In normal physiological conditions, the timing of behavioural rhythms, such as feeding, is driven by the SCN and thus represents a mechanism through which the SCN can synchronise rhythms in the periphery. The powerful nature of timed feeding as a circadian signal becomes apparent when food availability is divorced from SCN rhythms.

A typical protocol restricts food availability to nocturnal rodents, so that they can only eat during the light period of a light–dark cycle. This inverts the phase of clock gene rhythms in peripheral tissues, such as the liver, kidney, heart, pancreas, lung, gastrointestinal tract, and brown and white adipose tissue⁽ Reference Zvonic, Ptitsyn and Conrad ^{33 ,} Reference Damiola, Le Minh and Preitner ^{70 –} Reference Hoogerwerf, Hellmich and Cornelissen ^{72 )}. Full entrainment of the liver rhythms appears to occur within 2–3 d, whereas the other peripheral tissues may require up to 1 week before they exhibit the maximal phase shift. A later mouse study involved removal of food access for the first 6 h of the dark period of a light–dark cycle, with mild energy restriction. After 4 d in this protocol, mice exhibited delayed rhythms of hepatic and plasma TAG concentration, together with delayed rhythms of lipogenic and clock gene expression in both liver and adipose tissue⁽ Reference Yoshida, Shikata and Seki ^{73 )}. When animals are able to eat ad libitum quantities of food during temporal restriction paradigms, SCN rhythms remain locked to the light–dark cycle⁽ Reference Damiola, Le Minh and Preitner ^{70 ,} Reference Stokkan, Yamazaki and Tei ^{71 )}. However, the combination of temporal food restriction and hypoenergetic food availability does induce reorganisation of rodent SCN rhythms⁽ Reference Mendoza, Graff and Dardente ^{74 ,} Reference Mendoza, Gourmelen and Dumont ^{75 )}. Indeed, hypoenergetic feeding of nocturnal rodents without restricting food availability to the light period alters the phase of SCN-driven rhythms⁽ Reference Mendoza, Drevet and Pevet ^{76 )} as well as gene expression in peripheral tissues⁽ Reference Kuroda, Tahara and Saito ^{77 )}. Thus the overall effect of feeding on circadian organisation appears to involve an interaction between both the timing and the quantity of food intake (Fig. 1).

Fig. 1

Regulation of the circadian timing system by light and food. Under normal conditions of ad libitum food, light synchronises the master clock, the suprachiasmatic nuclei (SCN), which then synchronises peripheral clocks via neuronal and endocrine pathways, together with control over behavioural activity and thus feeding time. When feeding time (but not energy availability) is restricted, light remains the dominant synchroniser of the SCN, but peripheral clocks are synchronised to feeding time. Under conditions of temporal and energy food restriction, both the SCN and peripheral clocks are synchronised to the feeding time. (A colour version of this figure can be found online at http://www.journals.cambridge.org/nrr)

An important caveat when extrapolating the studies described above to human society is the relevance to standard human meal patterns. A common theme in human society is a feeding pattern of three meals per d. In contrast, most animal studies to date have utilised prolonged ad libitum feeding opportunities restricted to certain phases of the 24 h day. Some studies, however, have developed a ‘humanised’ meal protocol for rodents. When rats are only given their daily energy intake over ‘lunch’ and ‘dinner’ their gene expression rhythms in liver, heart and white adipose tissue are delayed compared with a group receiving the same total energy intake spread over three meals⁽ Reference Wu, Sun and ZhuGe ^{78 )}. Consistent with this finding, studies in mice using a range of mealtime combinations indicate that the first meal following a long fasting period provides an important synchronising signal to peripheral clocks⁽ Reference Kuroda, Tahara and Saito ^{77 ,} Reference Hirao, Nagahama and Tsuboi ^{79 )}. To date there are no comparable molecular data from human studies. One rare investigation of timed feeding on human circadian physiology is an experiment in which subjects were fed a single daily carbohydrate-rich meal for 3 d; morning consumption of this meal advanced core body temperature and heart rate, but not melatonin, rhythms compared with evening meal timing⁽ Reference Krauchi, Cajochen and Werth ^{80 )}. Furthermore, a delay in the timing of three daily meals within a fixed light–dark cycle is known to delay the phase of plasma leptin rhythms⁽ Reference Schoeller, Cella and Sinha ^{81 )}, which may be at least partially due to changes in adipose tissue clocks.

Although redundancy of signalling pathways providing input to the relevant tissues hinders elucidation of mechanistic insight into food entrainment, some progress has been made. At the nutritional level, phase shifting of the liver clock seems greater when the starch component of a mixed diet provides a large postprandial glucose concentration⁽ Reference Itokawa, Hirao and Nagahama ^{82 )}. Interestingly, however, ingestion of 100 % glucose, sucrose or maize starch is insufficient to alter the phase of liver rhythms, indicating that mixed macronutrient content may be necessary for food entrainment in the liver⁽ Reference Hirao, Tahara and Kimura ^{83 )}. At the physiological level, temporal food restriction more rapidly resynchronises peripheral clock gene rhythms in mice that have been adrenalectomised, compared with sham-operated controls⁽ Reference Le Minh, Damiola and Tronche ^{84 )}. This finding suggests that glucocorticoid signalling, which is believed to be an endocrine link between the SCN and peripheral clocks⁽ Reference Balsalobre, Brown and Marcacci ^{85 )}, may inhibit or delay the impact of temporal food availability. At the molecular level, it has been demonstrated that rhythmic phosphorylation of key transcription factors CREB and Akt (cAMP response element-binding protein and protein kinase B) in the mouse liver is driven by temporal feeding patterns⁽ Reference Vollmers, Gill and DiTacchio ^{86 )}. In keeping with this result, the expression of multiple genes targeted by molecular nutrient and stress sensors was similarly dependent on feeding time⁽ Reference Vollmers, Gill and DiTacchio ^{86 )}. Finally, mice deficient for the gene Parp1 (poly [ADP-ribose] polymerase 1) or the neurone-specific γ form of protein kinase C (PKCγ) exhibit impaired synchronisation to timed feeding⁽ Reference Asher, Reinke and Altmeyer ^{87 ,} Reference Zhang, Abraham and Lin ^{88 )}. Despite these advances, the mechanisms that mediate the effects of food on the circadian system are poorly understood and will doubtless be the subject of multiple further studies.

Diurnal changes

Diurnal rhythmicity refers to 24 h changes that occur in individuals kept in a varying environment, for example, a 24 h light–dark cycle. Although such rhythms are often relevant to a real-life scenario, there is the possibility that they are driven by environmental fluctuations rather than endogenous processes per se. As a result, they are not considered to be truly circadian.

Arguably the best-characterised daily metabolic rhythms in humans relate to changes in glucose homeostasis. Diurnal changes in glucose tolerance have been recognised in human subjects for many years⁽ Reference Van Cauter, Polonsky and Scheen ^{89 )}. Sensitivity to elevated glucose concentration is greatest in the early morning and then declines over the course of the day, leading to a phenomenon that has been termed ‘afternoon diabetes’. This daily change is not dependent upon changes in gastrointestinal function, but instead appears to be the result of altered glucose utilisation and insulin sensitivity, with maximal insulin sensitivity occurring in the early morning and decreasing throughout the day⁽ Reference Van Cauter, Polonsky and Scheen ^{89 )}.

In addition to glucose homeostasis, the regulation of plasma lipids is also subject to daily variation. Not only are basal concentrations of TAG elevated at night, but also there are diurnal changes in the postprandial TAG response. Ingestion of a meal at night results in increased plasma TAG that remains elevated for longer than the response to the same meal given during the day⁽ Reference Sopowski, Hampton and Ribeiro ^{90 )}. A study of postprandial responses to breakfast and lunch reported approximately 50 % less change in plasma TAG concentration following lunch than breakfast, despite the plasma TAG fraction showing no differences in concentration of [¹³C]palmitic acid that was included in each meal⁽ Reference Burdge, Jones and Frye ^{91 )}. The physiological basis for temporal differences in postprandial TAG response may therefore be independent of absorption or mobilisation of meal-derived lipids from the gut⁽ Reference Burdge, Jones and Frye ^{91 )}.

A number of groups have studied temporal variation of adipokines, which are adipose-derived hormones that regulate metabolic physiology in the brain and multiple peripheral tissues⁽ Reference Trujillo and Scherer ^{92 ,} Reference Galic, Oakhill and Steinberg ^{93 )}. Diurnal rhythms have been reported for many of these hormones, including leptin, adiponectin, chemerin, lipocalin and visfatin⁽ Reference Sinha, Ohannesian and Heiman ^{94 –} Reference Benedict, Shostak and Lange ^{98 )}. Although the secretion of these hormones is likely to be governed by multiple factors such as feeding and sleep, detailed analyses of leptin secretion suggest that there is likely to be an underlying circadian component to adipokine rhythmicity⁽ Reference Shea, Hilton and Orlova ^{99 ,} Reference Otway, Frost and Johnston ^{100 )}. Furthermore, given the functional roles of adipokines, their rhythmic secretion may make important contributions towards the daily changes in glucose and lipid homeostasis described above.

Identification of endogenous circadian rhythms

In order to unmask truly endogenous circadian rhythms from temporal changes in the environment, a number of different laboratory protocols have been developed. The most widely used of these are the constant routine and forced desynchrony protocols⁽ Reference Duffy and Dijk ^{101 ,} Reference Blatter and Cajochen ^{102 )}. In a constant routine, subjects are kept awake in a supine posture in constant dim light, with identical regular (for example, hourly) snacks. Although this protocol effectively removes environmental rhythms, it does result in the development of sleep debt due to the necessity to keep subjects awake. One solution to this problem is to allow subjects to sleep during the dark phase of a light–dark cycle that is sufficiently different from 24 h to permit entrainment of subjects' circadian rhythms; this is the basis of a forced desynchrony protocol. For example, a commonly used variant of this protocol employs a 28 h light–dark cycle that therefore allows subjects to sleep every 28 h, while their circadian rhythms occur (‘free run’) with a frequency of approximately 24 h.

Various research groups have utilised the above protocols to investigate the contribution of the endogenous circadian system to daily rhythms of glucose and lipid metabolism. One study in which 4-hourly meals were administered over a constant routine revealed elevation of both postprandial glucose and TAG during the biological night, especially after high-fat meal intervention over the week preceding the constant routine⁽ Reference Holmback, Forslund and Forslund ^{103 )}. Consistent with this finding, analysis of postprandial responses during a forced desynchrony of 27-h days revealed effects of both circadian time and length of prior wakefulness on glucose and TAG concentration⁽ Reference Morgan, Arendt and Owens ^{104 )}. By contrast, postprandial insulin responses were regulated by circadian time, but not length of wakefulness. A more recent forced desynchrony protocol kept volunteers on 28-h days, each of which contained four meals: breakfast, lunch, dinner and a snack shortly before bedtime⁽ Reference Scheer, Hilton and Mantzoros ^{105 )}. It was found that the times during which subjects were awake and eating during their biological night resulted in multiple cardiometabolic changes, including decreased plasma leptin concentration and increased concentrations of both plasma glucose and insulin. In fact, the postprandial responses of some of these healthy subjects during the biological night were equivalent to the responses of a pre-diabetic individual⁽ Reference Scheer, Hilton and Mantzoros ^{105 )}. In a separate cross-over study using forced desynchrony, volunteers had daily metabolic profiles assessed after both three 21-h days and also three 27-h days. Although there were some differences in response to the 21-h v. 27-h days, both schedules disrupted glucose–insulin metabolism, increased carbohydrate oxidation and reduced protein oxidation, but had little or no effect on appetite or energy balance⁽ Reference Gonnissen, Rutters and Mazuy ^{106 )}. The human circadian system therefore exerts clear influence over key aspects of metabolic physiology.

Dietary regulation of body weight

Evidence is rapidly accumulating to support an important role of meal times in the long-term regulation of body weight. Proof-of-concept studies in animal models have utilised different timed feeding paradigms. Mice housed in a light–dark cycle and fed a high-fat diet gain more weight when the food is available only throughout the light phase, when they would usually be resting, than when it is provided throughout the dark phase⁽ Reference Arble, Bass and Laposky ^{133 )}. This body-weight effect becomes statistically significant within 2 weeks and despite trends towards increased energy intake and reduced activity in the light-fed animals, there were no statistically significant changes in these parameters. In a refinement of the protocol, mice were fed a high- or low-fat diet that was available either ad libitum throughout the day, or during a 4 h period in the middle of the light phase⁽ Reference Sherman, Genzer and Cohen ^{134 )}. Remarkably, temporal food restriction resulted in reduced body weight on both high- and low-fat diets, to the extent that mice on a restricted high-fat diet weighed less than those provided with a low-fat diet ad libitum. This occurred despite no difference in energy consumption (as expressed relative to body weight) between the restricted high-fat-diet group and the two low-fat-diet groups, although mice under restricted feeding did exhibit elevated total daily activity compared with ad libitum controls⁽ Reference Sherman, Genzer and Cohen ^{134 )}. Comparing these two studies, it seems that the duration of restricted food availability has profound effects on body-weight regulation, but the mechanisms underlying this phenomenon are not yet clear.

In humans, there is an increasing interest in the effects of meal timing per se on metabolism and body weight. Work in this area has understandably focused to a large degree on meals taken at the start and end of the day, for instance the study of breakfast consumption and individuals with night eating disorders. Evidence suggests a role for regular breakfast consumption in the maintenance of healthy body weight, although the issue has rarely been approached from a chronobiological perspective and questions remain about the causative mechanisms involved⁽ Reference de la Hunty and Gibson ^{135 ,} Reference Casazza, Fontaine and Astrup ^{136 )}. Night eating syndrome has recently been included in the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5). It is broadly characterised by recurrent episodes of nocturnal eating that cannot be better accounted for by other behavioural and psychiatric disorders, and is discussed in detail elsewhere⁽ Reference Berner and Allison ^{137 )}. The relationship between night eating syndrome and body weight is complex, with some variable findings reported in the literature. Despite this, the overall evidence provides compelling associations between night eating and obesity, with a consistent finding that night eating syndrome is more prevalent in overweight and obese groups⁽ Reference Gallant, Lundgren and Drapeau ^{138 )}. The importance of evening meals is further highlighted in studies of subjects without night eating disorder. For example, energy consumption after 20.00 hours has been associated with BMI independently of age, sleep timing and sleep duration⁽ Reference Baron, Reid and Kern ^{139 )}. This group later reported that protein intake within 4 h of sleep onset is associated with elevated BMI after controlling for age, sex, sleep timing and sleep duration⁽ Reference Baron, Reid and Horn ^{140 )}.

The link between food timing and body weight is also apparent in dietary weight-loss studies. A group of 420 individuals undergoing a 20-week weight-loss programme were categorised according to the time at which they ate lunch, which was their main daily meal. Those in the late group lost less weight and at a slower rate than the early group, with no difference in energy intake, energy expenditure, dietary composition or sleep duration⁽ Reference Garaulet, Gomez-Abellan and Alburquerque-Bejar ^{141 )}. In a separate experiment, obese/overweight women consumed energy-restricted diets that differed in the proportion of energy distributed between breakfast and dinner. The women eating more energy at breakfast than dinner not only lost more weight but also exhibited an improved metabolic profile in insulin sensitivity and TAG concentration⁽ Reference Jakubowicz, Barnea and Wainstein ^{142 )}. Together these data support the hypothesis that the timing of food intake is important for body-weight regulation.

Shift work and jet-lag

Since industrialisation, humans have gained the ability to regulate environmental conditions and subsequently alter temporal patterns of behaviour. Indeed the phrase ‘24/7 society’ is now in common usage to describe the constant presence of industrial and social activity. In order to cope with the demands of this modern aspect of society, varied work schedules are now commonplace, with approximately 20 % of the European workforce engaged in night shifts⁽ Reference Costa ^{143 )}. This clearly indicates that a large section of the population experiences regular misalignment of their behaviour with the solar day. A second common cause of abrupt circadian misalignment is jet-lag, the rapid travel across time zones. Although this experience is a rare and transient phenomenon for most people, it nonetheless affects a substantial number of individuals and in some cases (for example, airline crew) can be a regular event. A related phenomenon that is common in society is a weekly change between widely differing sleep times on work and free days. This has been termed ‘social jet-lag’⁽ Reference Wittmann, Dinich and Merrow ^{144 )} and is associated with elevated BMI⁽ Reference Roenneberg, Allebrandt and Merrow ^{145 )}.

Multiple health problems are associated with shift work, including increased risk of cardiovascular and metabolic disease⁽ Reference Tucker, Marquie and Folkard ^{146 )}. Understanding the aetiology of shift work-related morbidity is complex due to diverse contributory factors such as sleep disturbance, altered social pressure and patterns of food intake⁽ Reference Esquirol, Bongard and Mabile ^{147 ,} Reference Lowden, Moreno and Holmback ^{148 )}. For example, although shift workers often report normal total energy intake, there is commonly an altered temporal distribution of feeding characterised by more irregular eating times, more snacking and fewer substantial meals⁽ Reference Lowden, Moreno and Holmback ^{148 )}. Alongside these lifestyle changes it is likely that disrupted circadian physiology is a major contributor to the pathophysiological consequences of shift work. Indeed it is generally recognised that many shift workers in temperate regions poorly adapt circadian rhythms to their work conditions⁽ Reference Folkard ^{149 )}. As a result, these individuals experience prolonged durations of misalignment between their circadian biology and behavioural patterns.

As described previously, postprandial profiles of glucose and TAG concentration vary over the day. Such studies clearly imply that shift workers eat a substantial proportion of their meals during the time of suboptimal glucose and lipid tolerance. This prediction is strengthened by studies of simulated shift work where subjects are subjected to an abrupt shift of typically 6–10 h in their daily routine. Interestingly, the postprandial response in such protocols is altered by the preceding diet. In comparable experiments, the exaggerated postprandial response following a test meal in shifted subjects was reduced for glucose and insulin but increased for TAG following a low-fat pre-meal⁽ Reference Ribeiro, Hampton and Morgan ^{150 )} rather than a high-fat pre-meal⁽ Reference Hampton, Morgan and Lawrence ^{151 )}. Furthermore, there are reported sex differences in postprandial response, with a more pronounced elevation of TAG in the first night of a simulated night shift in men than in women⁽ Reference Sopowski, Hampton and Ribeiro ^{90 )}.

In real shift workers, there are also postprandial data describing the relative insulin resistance and lipid intolerance following abrupt shift changes⁽ Reference Lund, Arendt and Hampton ^{152 )}. Given the large number of individuals undertaking shift work in modern society, there is a clear need to improve circadian alignment in these individuals. One possible intervention is the manipulation of light, which is able to reset circadian rhythms and also directly improve alertness⁽ Reference Vandewalle, Maquet and Dijk ^{153 )} even following short exposure times⁽ Reference Chang, Santhi and St Hilaire ^{154 )}. However, timed food may also be a powerful method for resetting rhythms to a new phase. Of particular interest is the ability of timed food to reset peripheral tissue rhythms.

Animal models support the idea that timed food intake could be a valuable intervention to minimise adverse effects of shift work. Mice subjected to an artificial shift work-like environmental schedule exhibit circadian desynchrony and metabolic disturbance⁽ Reference Barclay, Husse and Bode ^{155 )}. Similar findings have been reported in rats exposed to an artificial shift work protocol, although the body-weight increase and metabolic disturbances experienced were attenuated when food availability was restricted to the normal activity phase⁽ Reference Salgado-Delgado, Angeles-Castellanos and Saderi ^{156 )}. Further development of these animal models will permit detailed molecular analysis to complement human studies of shift work and its adverse effects.