Introduction The master clock, which, in mammals, resides in the hypothalamic suprachiasmatic nucleus (SCN), is thought to synchronize multiple peripheral oscillators to ensure temporal coordination of behavior and metabolism. Peripheral clocks amplify or dampen central rhythms or exhibit autonomous behavior to facilitate local adaptive responses (Hastings et al. 2003). The central clock may communicate to modulate or entrain rhythms in the periphery via hormones (McNamara et al. 2001) or hemodynamic cues. Asynchronous environmental cues, such as eating, also influence the autonomous behavior of peripheral clocks (Damiola et al. 2000; Stokkan et al. 2001). The variation in sleep and wakefulness (activity) is perhaps the most well-known circadian rhythm. Surgical ablation of the SCN in mice (Ibuka et al. 1980; Welsh et al. 1988) and rats (Ibuka et al. 1977; Mosko and Moore 1979) abolishes the nocturnal burst in locomotor activity. Similarly, disruption and/or mutation of Bmal1 (Bunger et al. 2000) or Clock (Vitaterna et al. 1994), transcription factors that compose the positive limb of an autoregulatory feedback loop in the core molecular clock (Young and Kay 2001; Reppert and Weaver 2002), also impairs circadian behavior. Bmal1 and Clock may influence behavioral rhythms by regulating the firing rate of SCN neurons (Herzog et al. 1998; Deboer et al. 2003). Genes relevant to the molecular clock are also expressed in peripheral tissues (Akhtar et al. 2002; Kita et al. 2002; Panda et al. 2002; Storch et al. 2002; Oishi et al. 2003) where approximately 5%–10% of the transcriptome is subject to circadian oscillation (Albrecht and Eichele 2003). Although the precise role of peripheral clocks and the mechanisms that link them to the SCN remain largely obscure, genetic mutation or deletion has implicated peripheral clocks in the regulation of some aspects of cellular function, including division (Matsuo et al. 2003), estrous cyclicity (Miller et al. 2004), and phospholipid metabolism (Marquez et al. 2004). Glucose and lipid homeostasis are also known to exhibit circadian variation (Seaman et al. 1965; Malherbe et al. 1969; Gagliardino and Hernandez 1971; Schlierf and Dorow 1973). Surgical ablation of the SCN impairs the control of glucose homeostasis (la Fleur et al. 2001). However, the proximity of satiety centres to the SCN has potentially confounded interpretation of these results. Indeed, there is no direct evidence implicating the molecular clock in the regulation of glycaemia or insulin sensitivity (Si). Our studies revealed a profound role for core clock genes—Bmal1 and Clock—in regulating recovery from insulin-induced hypoglycaemia. Furthermore, the impact of a high-fat diet (HF) was to amplify the diurnal variation in glucose tolerance and Si in a manner dependent on the Clock gene. These studies suggest that the temporal distribution of a caloric load may influence the response to insulin and that circadian variability in glucose homeostasis may be subject to modulation by asynchronous dietary cues. Results We examined the role of the molecular clock in glucose homeostasis by using mice in which core clock genes are impaired (Clockmut) or deficient (Bmal1−/−). Both plasma glucose and triglycerides were subject to circadian variation in wild-type (WT) mice, peaking at approximately circadian time point 4 (CT4) and CT28 (where CT0 is subjective day beginning at 7 AM, and CT12 is subjective night beginning at 7 PM) (Figure 1A and 1B), as reported previously (Seaman et al. 1965; Schlierf and Dorow 1973). We also observed that corticosterone (Figure 1C), which stimulates gluconeogenesis during hypoglycaemia (Cryer 1993), and adiponectin (Figure 1D), which has been associated with insulin resistance (Yamauchi et al. 2001; Maeda et al. 2002), oscillated significantly, but out of phase with the glucose and triglyceride rhythms. Diurnal variation in glucose and triglycerides, but not in corticosterone, was disrupted in the mutant mice (Table 1). Although there was no clear rhythm in the hypoglyacemic response to insulin, recovery of blood glucose exhibited a robust circadian variation (Figure 2A), with an excessive rebound from the effects of insulin evident at subjective dawn (CT19 and CT25) (Figure 2A). Insulin caused a profound hypoglyacemic response, independent of clock time, in both Bmal1 −/− and Clockmut mice (Figure 2B). This response was more pronounced in the former, consistent with the comparative severity of the molecular and behavioral phenotypes between the Bmal1−/− and Clockmut animals (King et al. 1997; Bunger et al. 2000). Despite exacerbation of the hypoglycaemic response to insulin in the mutants, the counterregulatory responses of both corticosterone and glucagon were retained (Figure 2C and 2D). Gluconeogenesis also contributes to restoration of blood glucose after insulin-induced hypoglycaemia. Consistent with this observation, conversion of exogenously administered pyruvate to glucose, which reflects gluconeogenesis (Miyake et al. 2002), was impaired in the Clockmut animals. This impairment was most marked in Bmal1−/− mice, while Bmal1+/− and Clockmut mice exhibited an intermediate phenotype when compared with WT littermate controls (Figure 3A). Furthermore, activity of the key rate-limiting enzyme of gluconeogenesis, phosphoenolpyruvate carboxykinase (PEPCK), exhibited diurnal variation in the liver and aorta that was blunted in Clockmut mice (Figure 3B). PEPCK activity in kidney was antiphasic to the rhythm in aorta and liver and was unimpaired in Clockmut mice (Figure 3B), suggesting tissue-specific regulation of enzyme activity. The frequent sampling intravenous glucose tolerance test (FSIGT) was performed to assess more precisely the impact of the molecular clock on sensitivity to insulin. This test provides an estimate of Si, consistent with that obtained by the euglycaemic clamp (Pacini et al. 2001). Additionally, data modeling provides estimates of glucose-mediated glucose disposal (Sg), insulin secretion, and Si. Si and insulin secretion, but not Sg, exhibited a diurnal variation in WT mice fed a regular chow diet (RC) (Table 2). Circadian variation of glucose and lipid homeostasis might condition the metabolic response to asynchronous environmental cues, such as diet, that impinge on Si. Dyslipidemia coincides with insulin resistance in the metabolic syndrome (Brotman and Girod 2002), and a diet high in fat impairs Si (Grundleger and Thenen 1982; Coulston et al. 1983). Both HF-fed WT and HF-fed Clockmut mice increased body weight significantly and to a similar degree in comparison to their age-matched, RC-fed controls (Table 3). Body fat composition averaged 17.6% of lean body mass in RC-fed WT mice, rising to 27.7% (p Clockmut). Thus, the most parsimonious interpretation is that the observed metabolic deficiencies in the Bmal1 −/− and Clockmut mice are due to their roles in the circadian clock, rather than to “off-clock” effects. We observed that the impact of HF on glucose homeostasis was apparently to emphasize the role of the molecular clock. Diet has previously been shown to interact with peripheral clocks. Changes in feeding shift the circadian pattern of gene expression in the liver, but not in the master clock in the SCN (Damiola et al. 2000), demonstrating the importance of food as a cue to circadian control. Individual constituents of food could also provide discrete stimuli. For example, glucose alone can induce rhythmic gene expression in isolated fibroblasts (Hirota et al. 2002). Thus, dietary composition, the size and timing of a feed might all be expected to interact differentially with an underlying circadian regulation of metabolic control. Alterations in dietary content, the availability of “fast food,” inactivity, and sociocultural factors have all been implicated in the emergence of the metabolic syndrome as a major challenge to the public health (Zimmet et al. 2001). However, while mechanistic integration of the diverse elements of the syndrome has proven elusive, our studies suggest that timing may influence the functional consequences of ingesting a caloric load. Materials and Methods Animals Mice were acclimatized for 2 wk in 12 h light–12 h dark cycles before being subjected to a 36-h period of constant darkness followed by experimentation in darkness. Experimental chronology is measured in CT, subjective day beginning at 7 AM (CT0), and subjective night beginning at 7 PM (CT12). Diet WT and Clockmut mice were placed on an HF (Teklad, TD02435) and compared to age-matched WT mice on a regular chow diet (RC). Mice were on RC for 8 wk except for those subjected to FSIGT where they received RC for 11 mo. Body mass composition was measured by dual energy X-ray absorptiometry at 10 mo. Intraperitoneal tolerance Tests were performed as described (Klaman et al. 2000) with a diminution in the glucose bolus (0.1 g/kg). Intravenous glucose tolerance test and minimal modeling The tolerance test was performed as described (Pacini et al. 2001) in unanesthetized mice, and the minimal model of Bergman et al. (1979) was applied to the data using MINMOD software (Boston et al. 2003). The derived values were Si, Sg, and acute insulin response to glucose, which measures insulin secretion. Si is the ratio of insulin delivery rate to the interstitium to insulin extraction rate from the interstitium. Long-term feeding of HF to WT mice resulted in imperceptibly small insulin sensitivity values. This could be the consequence of impaired delivery of insulin to the interstitium, exacerbated extraction rate, or a combination of both factors. Insulin secretion is derived from area under the insulin curve, above basal, from 0 to 10 min after glucose infusion; and disposition index, which equals the product of insulin sensitivity multiplied by insulin secretion and measures the degree to which insulin sensitivity can be compensated for by elevated insulin secretion (Pacini et al. 2001). Assay methods Insulin, leptin, corticosterone, and glucagon levels were measured by immunoassays from Crystalchem (Downers Grove, Illinois, United States), ICN Biochemicals (Costa Mesa, California, United States), and Linco Research (St. Charles, Missouri, United States). Plasma glucose was measured by the glucose oxidase method using a glucose analyzer machine for FSIGT and by glucometer for the intraperitoneal tolerance test. PEPCK activity was quantitated by a bioluminescent method (Wimmer 1988). Statistical analysis The significance of differences amongst the tolerance test curves was assessed by distribution-free two-way ANOVA with a Bonferroni correction. FSIGT data were tested by one-way ANOVA with the Kruskal-Wallis test. Paired Student's t-tests were used to perform comparisons of corticosterone levels before and after insulin injection in Bmal−/− mice and between WT and Clockmut mice. Plasma samples for glucagon analysis were pooled and were thus not compared by a formal statistical analysis. Results are presented as mean ± standard error of the mean (SEM), except for the FSIGT data (Table 2), presented as mean ± fractional standard deviation. Differences were considered significant when p < 0.05.