Systemic oscillator-driven and nutrient-responsive hormonal regulation of daily expression rhythms for gluconeogenic enzyme genes in the mouse liver

ABSTRACT Gluconeogenesis is de novo glucose synthesis from substrates such as amino acids and is vital when glucose is lacking in the diurnal nutritional fluctuation. Accordingly, genes for hepatic gluconeogenic enzymes exhibit daily expression rhythms, whose detailed regulations under nutritional variations remain elusive. As a first step, we performed general systematic characterization of daily expression profiles of gluconeogenic enzyme genes for phosphoenolpyruvate carboxykinase (PEPCK), cytosolic form (Pck1), glucose-6-phosphatase (G6Pase), catalytic subunit (G6pc), and tyrosine aminotransferase (TAT) (Tat) in the mouse liver. On a standard diet fed ad libitum, mRNA levels of these genes showed robust daily rhythms with a peak or an elevation phase during the late sleep-fasting period in the diurnal feeding/fasting (wake/sleep) cycle. The rhythmicity was preserved in constant darkness, modulated with prolonged fasting, attenuated by Clock mutation, and entrained to varied photoperiods and time-restricted feedings. These results are concordant with the notion that gluconeogenic enzyme genes are under the control of the intrinsic circadian oscillator, which is entrained by the light/dark cycle, and which in turn entrains the feeding/fasting cycle and also drives systemic signaling pathways such as the hypothalamic-pituitary-adrenal axis. On the other hand, time-restricted feedings also showed that the ingestion schedule, when separated from the light/dark cycle, can serve as an independent entrainer to daily expression rhythms of gluconeogenic enzyme genes. Moreover, nutritional changes dramatically modified expression profiles of the genes. In addition to prolonged fasting, a high-fat diet and a high-carbohydrate (no-protein) diet caused modification of daily expression rhythms of the genes, with characteristic changes in profiles of glucoregulatory hormones such as corticosterone, glucagon, and insulin, as well as their modulators including ghrelin, leptin, resistin, glucose-dependent insulinotropic polypeptide (GIP), and glucagon-like peptide-1 (GLP-1). Remarkably, high-protein (60% casein or soy-protein) diets activated the gluconeogenic enzyme genes atypically during the wake-feeding period, with paradoxical up-regulation of glucagon, which frequently formed correlation networks with other humoral factors. Based on these results, we propose that daily expression rhythms of gluconeogenic enzyme genes are under the control of systemic oscillator-driven and nutrient-responsive hormones.


Introduction
Under the nutritional fluctuation in the daily feeding/fasting (wake/sleep) cycle, blood glucose levels are strictly maintained within a narrow normal range. The brain and red blood cells depend largely on glucose for their energy source, and hypoglycemia in the sleep-fasting period can be life threatening. On the other hand, prolonged hyperglycemia in diabetes causes various organ disorders including nephropathy, neuropathy, and retinopathy. During the wake-feeding period under the condition of ad libitum access to standard diets, extra blood glucose is converted to glycogen and lipids in the liver. During the sleep-fasting period, blood glucose is supplied by immediate early glycogenolysis, and succeeding gluconeogenesis from substrates such as amino acids, with energy obtained mainly by lipolysis (Wahren and Ekberg 2007).
Regulation of gluconeogenesis is largely conducted through transcription of genes for gluconeogenic enzymes such as the glucose-6-phosphatase (G6Pase), catalytic subunit (G6pc) (Hutton and O'Brien 2009), phosphoenolpyruvate carboxykinase (PEPCK), cytosolic form (Pck1) (Croniger et al. 2002), as well as for amino acidmetabolizing enzymes, including tyrosine aminotransferase (TAT) (Tat) Nitsch et al. 1993) which provides carbon skeletons essential for gluconeogenesis. Daily rhythms in mRNA levels for Pck1 and Tat were first detected by in situ hybridization and RNA blot analyses (Bartels et al. 1990), and for G6pc by microarray analysis (Storch et al. 2002). To supply glucose adequately, the expression rhythms of gluconeogenic enzyme genes must be precisely regulated in response to nutritional changes such as prolonged fasting and day-to-day variations in dietary nutrient composition.
Candidates for mediators between nutritional changes and gene regulation in the context of daily rhythmicity are hepatic clock proteins . In particular, CRYs are implicated in the repression of Pck1 and G6pc by several different mechanisms: inactivating the glucocorticoid receptor (Lamia et al. 2011); inhibiting the glucagon signaling pathway (Zhang et al. 2010); and degrading a transcription factor FoxO1 in response to insulin (Jang et al. 2016). The high-fat diet accelerates degradation of CRY1 by autophagy, and stimulates gluconeogenesis with activation of Pck1 and G6pc in the late sleep-fasting period, while nuclear levels of other clock proteins are not severely affected (Toledo et al. 2018). Besides CRY1 and CRY2, recent genome-wide and site-specific chromatin immunoprecipitation analyses revealed that Pck1, G6pc, and Tat loci bind with clock proteins BMAL1, CLOCK, NPAS2, PER1, PER2, REV-ERBα, and REV-ERBβ in the liver and hepatic cells (Cho et al. 2012;Koike et al. 2012;Lamia et al. 2011;Rey et al. 2011;Schmutz et al. 2010;Yin et al. 2007). In spite of these interactions of clock proteins with regulatory machineries for gluconeogenic enzyme genes, liver-specific disruption of Bmal1 (Lamia et al. 2008), which is a nonredundant essential clock component (Bunger et al. 2000), only moderately affected hepatic Pck1 expression rhythms, while whole-body disruption of clock genes had varying results (Doi et al. 2010;Kennaway et al. 2007Kennaway et al. , 2013Le Martelot et al. 2009;Marcheva et al. 2010;Schmutz et al. 2010;Zani et al. 2013).
In addition, our group (Ishihara et al. 2007;Matsumoto et al. 2010) and others repeatedly noted that effects of nutritional variations on daily expression rhythms of clock genes in the liver are limited. Prolonged fasting, a high-fat diet, and a high-protein diet, but not a highcarbohydrate diet, caused marginal phaseadvancement of expression rhythms of Rev-Erbα and/or Dbp (Ishihara et al. 2007;Matsumoto et al. 2010). In another study (Oishi et al. 2012), a highprotein diet caused 2.4-2.8-h phase-advancement for expression rhythms of Bmal1, Npas2, and Cry1, but not significantly of other clock genes. It was also reported that high-fat diets caused weak, if any, phase-advancement and/or slight dampening of clock gene expression rhythms, in a number of studies (Eckel-Mahan et al. 2013;Guan et al. 2018;Kohsaka et al. 2007;Yanagihara et al. 2006), as well as in the study described above (Toledo et al. 2018). In general, as we postulated previously (Matsumoto et al. 2010), the relative stability of hepatic clock gene expression against nutrient changes is favorable to keep time for strictly feeding-associated liver functions, such as bile acid synthesis (Duez et al. 2008;Le Martelot et al. 2009). On the other hand, the expression rhythms of genes for metabolic adaptation, such as gluconeogenesis, must be tuned precisely in response to nutrient variations, sometimes even separate from entrainment to the feeding period.
Besides hepatic clock proteins, whole-body rhythmic cues such as hormones, autonomic innervation, metabolites, and body temperature (Kornmann et al. 2007;Lamia et al. 2008) can drive rhythmic gene expression in the liver, and their respective roles should be clarified. Here, we focus on hormonal regulation, since it is wellknown that gluconeogenic enzyme genes are under dynamic regulation by glucoregulatory hormones (e.g. activation by glucocorticoids and glucagon, and repression by insulin). Glucocorticoids (mainly cortisol in humans and corticosterone in mice) are secreted from the adrenal cortex with circadian rhythmicity (Atkinson et al. 2006;Van Cauter et al. 1991) under the control of the suprachiasmatic nucleus (SCN), the master pacemaker of the systemic clock oscillator, through the hypothalamic-pituitary-adrenal (HPA) axis and the autonomic nervous system (Dickmeis et al. 2013;Kalsbeek et al. 2012), and activate gluconeogenic enzyme genes through binding to the cognate nuclear receptor. Glucagon and insulin are secreted from pancreatic islet αand β-cells, respectively, in response to increased and decreased blood glucose, and bind to cognate G-protein-coupled receptor and tyrosine kinase receptor, leading to activation and repression of gluconeogenic enzyme genes. We, at first, characterized general features of daily expression rhythms of Pck1, G6pc, and Tat, and then comprehensively investigated their changes under various nutritional conditions, to examine whether these rhythmicity changes of gene expression for gluconeogenic enzymes associate with those of blood hormones.

Animals
Male C57BL/6J mice were purchased from Clea Japan (Tokyo, Japan). Clock mutant C57BL/6J mice were purchased from Jackson Laboratory (Stock No. 002923; Bar Harbor, ME, USA). Mice were housed at 24 ± 1°C under a 12-h light/12-h dark cycle (12:12LD) working as a time-giving cue "Zeitgeber", and given food and water ad libitum, unless otherwise noted. Lights-on time was assigned Zeitgeber time (ZT) 0. After entrainment under this condition at least for 2 weeks, constant darkness experiments were performed. The date of transferring mice into constant darkness was defined as Day 1, and the beginning of a subjective day was assigned circadian time (CT) 0.
For prolonged fasting in 12:12LD, mice were food-deprived at ZT9, and maintained for 42 h to 62 h. To adopt this range of fasting, we referred to a previous study that reported rhythmic expression of clock genes during 71-h fasting (Kobayashi et al. 2004) and studies that reported body weight loss within 30% after 72-h fasting (Furner and Feller 1971;Jensen et al. 2013;Sokolovic et al. 2007), these indices being met in the present study. For the prolonged fasting under constant darkness, mice were fooddeprived at CT11 of Day 1, and maintained for 16 h to 36 h. This range of fasting is shorter than that in 12:12LD, and was chosen to minimize the effect of individual differences in free-running periods under constant darkness.
To assess effects of varied photoperiod lengths and restricted feeding, mice were entrained to the shifted conditions for at least 2 weeks and 10 days, respectively.
All experimental procedures in this study were approved by the Animal Experiment Committee of Chiba University and the Experimental Animal Welfare Committee of Waseda University, and were conducted in accordance with the Guidelines for Proper Conduct of Animal Experiments of the Science Council of Japan.

Northern blot analysis
Total RNA was isolated from the liver of mice by the acid-guanidine thiocyanate-phenol/chloroform extraction procedure (Chomczynski and Sacchi 1987) or with TRIzol reagent (Invitrogen, Carlsbad, CA, USA). RNAs were electrophoresed in denaturing formaldehydeagarose (1%) gels. After 28S and 18S rRNAs were visualized by ethidium bromide staining, their fluorescence images were recorded and quantified by using the Light Capture system (Atto, Tokyo, Japan). RNAs were blotted onto Nytran N membranes (Whatman, Sanford, ME, USA), and were subjected to hybridization. Digoxigenin-labeled RNA probes were synthesized using a transcription kit (Roche Diagnostics, Mannheim, Germany) from cDNA sequences for Pck1 (accession NM_198780.3 nt. 1544-2623, a gift from D. K. Granner), G6pc (accession NM_008061.3 nt. 106-937), and Tat (accession NM_146214.3 nt. 1-973, a gift from G. Schütz). Hybridization, washing, and chemiluminescence detection of mRNAs were done as recommended by Roche Diagnostics. Recording and quantification of images were performed by using the Light Capture system (Atto). The relative levels of mRNAs were calculated as their chemiluminescence intensities divided by the fluorescence intensities of rRNAs.

Data analyses
Measured values are expressed as mean ± SEM. Statistical analyses were performed using the software SPSS Statistics version 23.0 (IBM Japan, Tokyo, Japan). Rhythmicity/fluctuation was tested by one-way analysis of variance (ANOVA). Significant differences among varied lighting, dietary, and genetic groups were analyzed by two-way ANOVA followed by pairwise comparisons with Bonferroni correction for multiplicity. Correlation was tested by Pearson's linear regression analysis. Values of 0.05 or less were considered to be statistically significant. Correlation networks were graphically represented with the software Cytoscape (Shannon et al. 2003) version 3.2.1.

Results
Daily and circadian expression rhythms of Pck1, G6pc, and Tat in the mouse liver Male C57BL/6 mice were fed a standard diet ad libitum in 12:12LD, and total RNA was extracted from the liver. As shown in Figure 1a, Pck1 mRNA levels exhibited a daily rhythm with a sharp peak at ZT11 (see also Supplementary Figure S1A and refer to Supplementary Table S1 also when required below) in the late phase of the light period (the sleepfasting period for nocturnal mice), and both G6pc and Tat mRNA levels showed daily rhythms with elevation at ZT11 and a peak around ZT19 in the middle phase of the dark period (the wake-feeding period). These daily expression profiles of each gene were largely concordant with those found in databases CircaDB (http://circadb.hogeneschlab.org) (Pizarro et al. 2013) and CircadiOmics (http://circa diomics.igb.uci.edu) (Eckel-Mahan et al. 2013) for transcriptome rhythmicities investigated mainly by microarray and RNA-seq analyses, in case when significant rhythmicities were detected for the genes in these databases (data not shown). Prolonged fasting for 42-62 h ( Figure 1b) caused general elevation of the mRNA levels with a shift of each peak phase (Supplementary Figure S1B), being reminiscent of previous reports for the effects of fasting on the oscillation of gluconeogenic enzyme activities (Kato and Saito 1980;Mlekusch et al. 1981;Phillips and Berry 1970). The daily expression rhythm of the clock gene Bmal1 (Supplementary Figure S2A), examined as a reference of intrinsic clockwork, was moderately damped and peak phaseadvanced by the prolonged fasting (Supplementary Figure S2B), resembling previous observations (Kawamoto et al. 2006;Shavlakadze et al. 2013).
To examine if the rhythmicity persists in the absence of environmental time cues (i.e. in a circadian manner), we housed mice in constant darkness and observed that rhythmic Pck1, G6pc,  . Daily rhythms of mRNA levels for gluconeogenic enzyme genes Pck1, G6pc, and Tat in the mouse liver on feeding ad libitum (a) and prolonged fasting (b). Male wild-type C57BL/6 mice were housed in a 12-h light/12-h dark cycle (12:12LD, light/dark periods are represented by open/solid boxes) with feeding ad libitum for at least 2 weeks until Day 1 (experimental schedule is shown below). For the prolonged fasting, mice were food-deprived at ZT9 of Day 1. Livers were excised at 4-h intervals on Day 3 from 3 mice at each time point. Total RNA was prepared and subjected to Northern blot analysis. A chemiluminogram for detection of each mRNA indicated is shown along with ethidium bromide staining for 28S and 18S rRNAs. Below, quantified results for each mRNA level relative to the maximum value (100%) in a are represented as mean ± SEM. p values for rhythmicity were assessed by one-way ANOVA. The broken lines in b represent the results reproduced from a. *p < 0.05, **p < 0.01 in comparison with a in each time point (two-way ANOVA followed by pairwise comparisons with Bonferroni correction).

Figure 2.
Circadian rhythms of mRNA levels for gluconeogenic enzyme genes in constant darkness on feeding ad libitum (a) and prolonged fasting (b). After the housing in 12:12LD for at least 2 weeks, male wild-type C57BL/6 mice were transferred to constant darkness on Day 1 (experimental schedule is shown below). The subjective day is represented by the hatched box. For the prolonged fasting, mice were food-deprived at CT11 of Day 1. Livers were excised on Day 2 and subjected to RNA extraction to examine the rhythmicity of Pck1, G6pc, and Tat mRNA levels, as described in the legend of Figure Figure S3B). It is difficult to infer the exact cause of this difference because of varied experimental conditions in both the fasting duration and the lighting schedule. On the other hand, rhythmic expression of gluconeogenic enzyme genes under different fasting conditions suggests its generally robust feature. Especially, persistence of the rhythmicity during the prolonged fasting in constant darkness is concordant with the notion that Pck1, G6pc, and Tat expressions are under the control of an intrinsic circadian oscillator. The Bmal1 expression rhythm in constant darkness (Supplementary Figure S4A) showed no significant difference compared to that in 12:12LD (Supplementary Figure  S2A) under the ad libitum feeding condition, and its peak phase was apparently advanced by the prolonged fasting (Supplementary Figure S4B), being reminiscent of a previous report (Barnea et al. 2009).

Alterations in Clock mutant mice
To test involvement of the circadian oscillation generator in the Pck1, G6pc, and Tat expression rhythmicity, we examined mice harboring a Clock mutation yielding a splice variant for a dominantnegative form of the CLOCK protein (King et al. 1997). Under the ad libitum feeding condition (Figure 3a), the daily rhythm of Pck1 mRNA levels exhibited a large shift of the peak phase (Supplementary Figure S5A), whereas G6pc and Tat mRNA levels lost rhythmicity ( Figure 3a). These results are largely concordant with those of previous reports for expression rhythms of Pck1 (Doi et al. 2010;Kennaway et al. 2007;Marcheva et al. 2010) and G6pc (Doi et al. 2010;Marcheva et al. 2010) in Clock mutants. Paradoxically and most interestingly, with prolonged fasting (Figure 3b), G6pc and Tat expressions recovered rhythmicity with a peak around ZT11 (Supplementary Figure  S5B), when the Pck1 expression also showed a peak. Therefore, Pck1, G6pc, and Tat expression rhythms exhibited apparent synchronization in fasted Clock mutant mice. A plausible explanation for these results is that disappearance of day/night difference of food intake (Meyer-Kovac et al. 2017;Turek et al. 2005) in Clock mutant mice fed ad libitum disturbed daily rhythms of Pck1, G6pc, and Tat expression, and prolonged fasting eliminated this disturbance and unmasked the effects of other oscillatory regulators. A candidate for such regulators is corticosterone, blood levels of which are still rhythmic in Clock mutant mice (Oishi et al. 2006;Rudic et al. 2004), but the corticosterone rhythmicity in prolonged fasting remains to be confirmed.
The daily expression rhythm of Bmal1 was damped in Clock mutant mice (Supplementary Figure S6A) compared to wild-type mice (Supplementary Figure S2A) under the ad libitum feeding condition, and was lost by the prolonged fasting (Supplementary Figure S6B). Therefore, effects of Clock mutation on daily expression profiles were substantially different between Bmal1 and gluconeogenic enzyme genes, and, taken together the results of Figures 1-3 and Supplementary Figures S1-S6, effects of fasting on the expression profiles were also different between Bmal1 and gluconeogenic enzyme genes.

Entrainment to varied photoperiods and time-restricted feedings
We examined whether Pck1, G6pc, and Tat expressions are entrained to two varied photoperiods, the 6-h light/18-h dark cycle (6:18LD) and 18-h light/ 6-h dark cycle (18:6LD). In 6:18LD (Figure 4a), Pck1, G6pc, and Tat expression rhythms were phaseadvanced compared with those in 12:12LD. On the other hand, in 18:6LD (Figure 4b), the expression rhythms were damped, but Pck1 trended to and G6pc and Tat reached statistical significance for the rhythmicity, with apparent phase-delays. Therefore, expression rhythms of gluconeogenic enzyme genes were entrained to the varied photoperiods. We previously demonstrated that varied photoperiods entrain the wake/sleep cycle and then feeding/fasting cycle (Matsumoto et al. 2010), which here in turn likely entrained the expression of gluconeogenic enzyme genes. Elongated fasting in the longer light period of 18:6LD may cause multi-time activation of gluconeogenic enzyme genes, thus attenuating their daily expression rhythms.
To examine the effects of the feeding desynchronized from the light/dark cycle, we carried out time-restricted feeding experiments, which uncouple oscillation of clock genes in peripheral tissues from that in the SCN, the light-entrained master pacemaker (Damiola et al. 2000;Hara et al. 2001;Stokkan et al. 2001). When we supplied mice with chow only during the central 4 h (ZT4-8) of the light period, Pck1, G6pc, and Tat expression rhythms were phase-advanced (Figure 5a). With restricted feeding during the central 4 h (ZT16-20) of the dark period (Figure 5b), Pck1 showed a phase-delay with a peak around ZT15, while G6pc also showed a phase-delay with two apparent peaks around ZT15 and 23, and Tat showed a narrowed peak phase at ZT15. As a common feature, mRNA levels of these genes exhibited sharp elevations from ZT11 to 15, just prior to initiation of the restricted feeding. Therefore, both light-and dark-phase restricted feedings entrained expression rhythms of gluconeogenic enzyme genes.

Effects of a high-fat diet
A large number of studies have reported nutritional regulation of gluconeogenic enzymes, prompting us to examine whether gene expression rhythms are affected by variations in the composition of three major nutrients. First, we tested a high-fat diet (34% fat by weight) on which mice were fed ad libitum for 7 days (Figure 6a). Compared with the standard diet (5% fat), the high-fat diet elevated Pck1, G6pc, and Tat mRNA levels in the middle light (sleep-fasting) period at ZT7, attenuating the rhythmicities, yet still reaching statistical significance.
Effects of a high-carbohydrate (no-protein) diet As shown in Figure 6b, each of the gluconeogenic enzyme genes displayed characteristic features in their daily expression profiles when a highcarbohydrate (81% carbohydrate, 0% protein) diet was given for 7 days. Compared with the standard diet (55% carbohydrate, 28% protein), Pck1 mRNA levels were generally lowered, with a damping of the rhythmicity. While glucose represses Pck1 (Kahn et al. 1989;Meyer et al. 1991), mice fed on the high-carbohydrate (no-protein) diet were rather hypoglycemic, as described below (Figure 7a). Figure 5. Effects of time-restricted feeding on daily expression rhythms of gluconeogenic enzyme genes. Male wild-type C57BL/6 mice were housed in 12:12LD with feeding restricted to the middle of the light period at ZT4-8 (a) or the dark period at ZT16-20 (b) for at least 10 days, and examined for the rhythmicity of Pck1, G6pc, and Tat mRNA levels, as described in the legend of Figure 1. The broken lines represent the results reproduced from Figure 1a for feeding ad libitum. *p < 0.05, **p < 0.01 in comparison with Figure  1a in each time point (two-way ANOVA followed by pairwise comparisons with Bonferroni correction).
Another candidate factor repressing Pck1 is hypoproteinosis, which results from the protein-free diet and causes a lack of amino acids, a major source for gluconeogenesis.
G6pc mRNA levels were elevated in the dark (wake-feeding) period compared with the standard diet. Because G6Pase is involved in glucose production not only via gluconeogenesis but also via glycogenolysis, G6pc is likely activated, rather than repressed, by ingestion of the high-carbohydrate (no-protein) diet, which lacks gluconeogenic amino acids, and which may accelerate glycogen turnover during the short interprandial phase of the wakefeeding period as well as during the long interprandial phase of the sleep-fasting period. While glucose activates G6pc (Arden et al. 2012;Argaud et al. 1997;Massillon et al. 1996), its underlying mechanism is unknown, either similar or not to that for the present high-carbohydrate (no-protein) diet causing hypoglycemia.
The Tat expression rhythm was phase-advanced compared with the standard diet, with a plateau at ZT7-15, when especially the high-carbohydrate (noprotein) diet may necessitate degradation of body Figure 6. Effects of varied dietary nutrients on daily expression rhythms of gluconeogenic enzyme genes. Male wild-type C57BL/6 mice were housed in 12:12LD with feeding ad libitum on the high-fat (a), high-carbohydrate (no-protein) (b), 60% casein (c), 15% casein (d), 60% soy-protein (e), and 15% soy-protein diets (f) for 7 days, and examined for the rhythmicity of Pck1, G6pc, and Tat mRNA levels, as described in the legend of Figure 1. The broken lines represent the results reproduced from Figure 1a for the standard diet. *p < 0.05, **p < 0.01 in comparison with Figure 1a in each time point; # p < 0.05, ## p < 0.01 for 60% casein diet (c) versus 15% casein diet (d), and for 60% soy-protein diet (e) versus 15% soy-protein diet (f) in each time point (two-way ANOVA followed by pairwise comparisons with Bonferroni correction). Figure 7a. Daily profiles of blood humoral factors on various nutritional conditions. Male wild-type C57BL/6 mice were housed in 12:12LD on the nutritional conditions indicated: from left to right, the standard diet, prolonged fasting, the high-fat diet, the highcarbohydrate (no-protein) diet, the casein (60% or 15%) diet, and the soy-protein (60% or 15%) diet with ad libitum feeding for 7 days, except for the prolonged fasting condition identical to that of Figure 1b, and whole blood or plasma was subjected to measurement of blood glucose (a), plasma PAI-1 (b), corticosterone (c), ghrelin (d), leptin (e), resistin (f), glucagon (g), insulin (h), glucagon/insulin ratio (i), GIP (j), and GLP-1 (k). Concentrations of the humoral factors are represented as mean ± SEM. p values for rhythmicity were assessed by one-way ANOVA. *p < 0.05, **p < 0.01 in comparison with the standard diet in each time point; # p < 0.05, ## p < 0.01 for 60% casein diet versus 15% casein diet, and for 60% soy-protein diet versus 15% soy-protein diet in each time point (two-way ANOVA followed by pairwise comparisons with Bonferroni correction). The insets are for scale change and overlap resolution.
protein to supply amino acids including tyrosine, the substrate for TAT, to produce sources for gluconeogenesis.

Effects of high-and low-protein (casein or soy-protein) diets
Mice were fed on a high-(60%) or low-(15%) protein diet containing either casein or soy protein for 7 days. On the 60% casein diet (Figure 6c), compared with that on the 15% casein diet ( Figure 6d) and/or the standard diet (Figure 1a), Pck1, G6pc, and Tat mRNA levels rose mainly during the dark (wake-feeding) period. It was also the case on the 60% soy-protein diet (Figure 6e), compared with that on 15% soy-protein diet (Figure 6f) and/or the standard diet (Figure 1a). This may seem paradoxical for gluconeogenesis to be activated typically during the light (sleep-fasting) period, but is likely required to combat hypoglycemia (see  7a) caused by the high-protein (lowcarbohydrate) diets. A previous study (Oishi et al. 2012) reported general elevation of the mRNA levels of Pck1 and G6pc on the high-protein (74.1%) diet for 14 days (Oishi et al. 2012).
As for the factor(s) causing hepatic activation of the genes in response to high-protein diets, amino acids themselves are candidates. However, in a previous study (Azzout-Marniche et al. 2007), high concentrations of amino acids failed to activate Pck1 and G6pc in primary-cultured hepatocytes, while hepatic mTOR and GCN2 transduction pathways can sense increases in amino acid concentration (Chotechuang et al. 2009). In contrast, it is highly plausible that hormonal changes, such as glucagon up-regulation, in response to highprotein diets lead to the activation of gluconeogenic enzyme genes, as described below.

Alteration of daily rhythms of blood humoral factors by nutritional variations
We examined effects of nutritional changes on daily profiles of blood glucose and plasma humoral factors (Figure 7a). Blood glucose level (Figure 7a) was relatively constant across the day and was non-rhythmic on the standard diet. The nonrhythmicity was also observed with prolonged fasting as well as on the high-fat, high-carbohydrate (no-protein), 60% casein, 15% casein, and 15% soy-protein diets, with the exception being the 60% soy-protein diet. The prolonged fasting lowered glucose levels severely throughout the day, and the high-carbohydrate (no-protein) diet did so moderately in both light and dark periods. The 60% casein diet lowered glucose levels at ZT23 in the late dark (wake-feeding) period compared with the 15% casein diet. As mentioned above, the 60% soy-protein diet did cause daily oscillation of blood glucose levels, which were lowered in the early to middle dark (wakefeeding) period at ZT15-19 compared with the standard diet. Therefore, the high-protein diets led to dark (wake-feeding) period-specific hypoglycemia.
As shown in Figure 7b, PAI-1 levels exhibited robust daily oscillation with a peak/plateau at ZT11-15 around the light-dark transition under most nutritional conditions, except for the high-fat diet, which abolished the rhythmicity of the PAI-1 levels. The high-carbohydrate diet generally raised PAI-1 levels. PAI-1 is the primary inhibitor of thrombolysis. As recently reported, obesity elevates plasma PAI-1 levels, which are associated with complications such as cardiovascular disorders (Dellas and Loskutoff 2005;Van de Craen et al. 2012). PAI-1 is synthesized in most tissues and organs, including heart, vascular endothelium, lung, adipose, and liver, and is induced by various stimuli such as stresses, hormones, and cytokines (Dellas and Loskutoff 2005;Van de Craen et al. 2012). PAI-1 has been repeatedly noted for its robust daily rhythm (Andreotti and Kluft 1991;Kudo et al. 2004;Scheer and Shea 2014), and can serve as a standard oscillating marker in the present plasma sampling.
Corticosterone levels (Figure 7c) also exhibited significant daily rhythms under most dietary conditions, except for prolonged fasting (but with rhythmic tendency, p= 0.074) and the high-fat diet. When rhythmic, corticosterone reached a peak/plateau around ZT11-19, and a trough at ZT23. Relatively stable circadian rhythms of corticosterone levels likely underlie the expression rhythms of gluconeogenic enzyme genes in the various nutritional conditions. Prolonged fasting and the high-carbohydrate (no-protein) diet generally raised corticosterone levels, seemingly, at least in part, through hypoglycemia ( Figure 7a). Elevated ghrelin levels ( Figure 7d) and lowered leptin levels ( Figure 7e) may also contribute to raise corticosterone levels, as described below. On the high-fat diet, the dampened corticosterone rhythm likely leads to altered expression rhythms of gluconeogenic enzyme genes (Figure 6a).
Ghrelin rhythmicity (Figure 7d) was observed on the standard diet, prolonged fasting, the highcarbohydrate (no-protein) diet, and the 60% soyprotein diet, with a peak/plateau at ZT7, 15-23, 3, and 3-7, respectively. Prolonged fasting and the high-carbohydrate (no-protein) diet generally raised ghrelin levels compared with the standard diet. These marked elevations may contribute to raise corticosterone levels (Figure 7c) by directly stimulating adrenocorticotropic hormone release from the anterior pituitary in the HPA axis (Spencer et al. 2015). Variations of ghrelin rhythms in response to dietary changes likely reflect complicated regulation of plasma ghrelin levels by multiple factors such as intrinsic clockwork, ingestion schedules, nutrients, hormones, and body mass index, as reviewed in (Iwakura et al. 2015).
Leptin rhythmicity (Figure 7e) was detected only on the 15% soy-protein diet, with a peak at ZT11. Compared with the standard diet, the high-fat and 15% casein diets generally raised leptin levels, while prolonged fasting and the high-carbohydrate (noprotein) diet lowered them at all and 3 time points, respectively. Rhythmic tendency (p= 0.067) on the prolonged fasting might reflect the circadian nature of the rhythmicity. Several previous reports have attributed the diurnal leptin patterns to diet schedules and/or systemic circadian oscillation (Kettner et al. 2015;Scheer et al. 2009;Shea et al. 2005). The lowering of leptin levels on the prolonged fasting and the high-carbohydrate (no-protein) diets may contribute to elevation of corticosterone levels ( Figure 7c) by ceasing to repress the HPA axis and then increasing the secretion of corticotropinreleasing factor from the hypothalamic paraventricular nucleus and corticosterone from the adrenal cortex (Roubos et al. 2012).
Resistin (Figure 7f), an adipokine as with leptin, exhibited daily rhythms on the high-fat, 60% soyprotein, and 15% soy-protein diets, with a peak/ plateau at ZT3, 15-23, and 23-3, respectively. Previously, daily resistin rhythm in rat serum (Oliver et al. 2006), but not in human plasma (Sanchez-de-la-Torre et al. 2014), was detected. Like leptin levels, the high-fat diet generally raised, while prolonged fasting lowered, resistin levels, concordant with previous reports (Rajala et al. 2002;Steppan et al. 2001). Since a study with gene targeting (Banerjee et al. 2004) postulated that resistin is responsible for maintenance of blood glucose and for hepatic activation of Pck1 and G6pc during prolonged fasting, generally lowered resistin levels with prolonged fasting may still contribute to activation of the gluconeogenic enzyme genes.
Glucagon (Figure 7g) did not show significant rhythmicity in any dietary condition, but showed rhythmic tendency (p= 0.075) on the 60% soyprotein diet with a plateau at ZT15-19. The prolonged fasting exhibited no apparent effect, though controversial in previous reports that, by fasting, glucagon levels were slightly lowered in rats (Ruiter et al. 2003) and mice , and slightly elevated in rats (Mlekusch et al. 1981). The high-fat diet raised glucagon levels at ZT11 and 23, compared with the standard diet. The high-carbohydrate (no-protein) diet lowered them in the early to middle dark (wake-feeding) period at ZT15-19. Contrarily, high-protein (60% casein or soy-protein) diets raised glucagon levels in the wake-feeding period. The 60% casein diet raised glucagon levels at ZT19-23 in the middle to late dark period, compared with the 15% casein and standard diets. The 60% soy-protein diet also raised glucagon levels at ZT15-19 in the early to middle dark period, compared with the 15% soy-protein and standard diets. We concluded that the highprotein (60% casein or soy-protein) diets raised plasma glucagon levels, especially in the dark (wake-feeding) period, and hypothesize that this contributes to the high-protein diet-induced elevation of mRNA levels for gluconeogenic enzymes (Figure 6c and e).
Insulin (Figure 7h) exhibited daily rhythms on the standard, high-fat, 60% casein, 15% casein, and 60% soy-protein diets, with a peak/plateau at 11,19,11,[19][20][21][22][23]respectively. On the standard diet, elevation of insulin levels during the dark (wake-feeding) period is likely responsible for repression of Pck1, G6pc, and Tat in a genedifferential manner (Figure 1a). An atypical insulin peak at ZT11 in the late light (sleep-fasting) period on the high-fat and 15% casein diets, as well as possibly on the 15% soy-protein diet with the tendency of insulin rhythmicity (p= 0.059), may underlie phase-shifts and/or damping of the daily expression rhythms of the gluconeogenic enzyme genes (Figure 6a, d, and f). Prolonged fasting and the high-carbohydrate (no-protein) diet lowered insulin levels especially in the dark period, possibly at least in part via hypoglycemia (Figure 7a), lowering leptin levels (Figure 7e), and elevating corticosterone ( Figure 7c) and ghrelin levels (Figure 7d). The lowered insulin levels are in turn likely involved in the activation of the gluconeogenic enzyme genes on the prolonged fasting ( Figure 1b) and in the paradoxical G6pc activation on the highcarbohydrate (no-protein) diet (Figure 6b).
Rhythmicity of the glucagon/insulin ratio (Figure 7i) was detected only on the standard diet, with a plateau at ZT3-11. Insulin rhythms detected in other nutritional conditions ( Figure  7h) were apparently canceled by related fluctuations of glucagon levels. As reported previously (Gylfe and Gilon 2014), insulin and glucagon are secreted from mouse islets in an alternatively pulsatile manner, increasingly in parallel with glucose concentration in the range of 7 to 20 mM (126 to 360 mg/dl), which is the case for most conditions in this study ( Figure 7a). As described below (Figure 8), positive correlations between glucagon and insulin levels were observed on the standard diet as well as the 60% casein diet. GIP (Figure 7j) exhibited rhythmicity on the high-carbohydrate (no-protein) diet with a peak at ZT19, and fluctuation on the 15% soy-protein diet with two peak/plateaus at ZT11 and ZT23-3. The high-fat, high-carbohydrate (no-protein), 15% casein, and 15% soy-protein diets raised GIP levels at 4, 3, 4, and 2 time points, respectively, compared with the standard diet. The GIP up-regulation on the 15% casein diet was significant at 4 time points compared with the 60% casein diet. These GIP up-regulations are concordant with previous observations that dietary carbohydrate and fat stimulate GIP secretion more strongly than protein (Diakogiannaki et al. 2012).
GLP-1 (Figure 7k) exhibited fluctuation on prolonged fasting with peaks at ZT3 and 19, and rhythmicity on the 15% casein diet with a plateau at ZT7-11. Previous studies (Elliott et al. 1993;Ørskov et al. 1996) reported that plasma levels of GLP-1, as well as GIP, elevate in response to dietary ingestion, while their circadian natures are controversial (Gil-Lozano et al. 2014). The fluctuation of GLP-1 levels on the prolonged fasting suggests intrinsic oscillation, which remains to be confirmed by examining effects of conditions such as constant darkness.
Prolonged fasting, the highcarbohydrate (no-protein) diet, and the 15% casein diet each lowered GLP-1 levels at 2 time points, compared with the standard diet. Previously, elevation of GLP-1 levels by supplementation with dietary protein in humans has been demonstrated in a few studies (Lejeune et al. 2006;Raben et al. 2003), but not found in another (Karamanlis et al. 2007).

Correlations and networks of blood humoral factors
Pearson's correlation coefficients between every pair of humoral factors in each dietary condition (examples of scatter plots are shown in Supplementary Figure S7) are noted in the Cytoscape map (Figure 8). One of the most frequent positive correlations observed was between glucagon and GLP-1 (Supplementary Figure S7, A-C) on the standard diet (Figure 8a), the prolonged fasting (Figure 8b), the 60% casein diet (Figure 8e), and the 60% soy-protein diet (Figure 8g). This was surprising because GLP-1 typically inhibits glucagon secretion (Campbell and Drucker 2013). Further studies are needed to clarify the mechanisms and meanings of these positive correlations between glucagon and GLP-1. In contrast, positive correlations between insulin and GLP-1 (Supplementary Figure S7D) were detected only on the standard (Figure 8a) and 15% casein diets (Figure 8f), even though GLP-1 stimulates insulin secretion (Campbell and Drucker 2013).
Glucagon also exhibited positive correlations with insulin (Supplementary Figure S7, E and F) on the standard (Figure 8a) and 60% casein diets (Figure 8e), despite the fact that they play opposing roles in regulation of blood glucose levels. As described above in reference to Figure 7i, reciprocally pulsatile but totally parallel secretions of glucagon and insulin (Gylfe and Gilon 2014) seem to result in their positive correlations, which can be beneficial in dietary conditions such as low carbohydrate/protein ratios imposing the need for synthesis of both glucose and fat from amino acids.
We observed positive correlations between corticosterone and resistin on prolonged fasting ( Figure 8b) and the high-fat diet (Figure 8c) as well as negative correlations on the highcarbohydrate (no-protein) (Figure 8d), 60% casein (Figure 8e), and 15% casein diets (Figure 8f). A previous study (Shojima et al. 2002) reported that the synthetic glucocorticoid dexamethasone raised resistin mRNA and protein levels in adipose tissue, but more direct reciprocal effects between corticosterone and resistin on their secretions remain to be elucidated. On the high-fat diet, corticosterone ( Figure 7c) and resistin (Figure 7f) exhibited an up-and-down feature in their daily profiles, which was also the case for leptin ( Figure 7e) and GIP (Figure 7j), and positive correlations were detected among corticosterone, resistin, and leptin, and between resistin and GIP ( Figure 8c). Figure 8i shows frequencies for incorporation of each humoral factor into a closed triplet, which is formed by significant correlations (Figure 8a-h), and which is regarded as an elemental network cluster (Newman 2003). Glucagon gave the highest frequency with 15 in total, with major contributions from frequencies 4 and 6 on the 60% casein ( Figure 8e) and 60% soy-protein diets (Figure 8g), respectively. Glucagon shared clusters twice or more with ghrelin, insulin, or GIP on the 60% casein diet (Figure 8e), and with glucose, PAI-1, GIP, or GLP-1 on the 60% soyprotein diet (Figure 8g). Therefore, glucagon seems to occupy the central position in the humoral networks, especially on high-protein diets. While quite intriguing, biological meanings and formation mechanisms of theses glucagon-centered networks remain elusive. As for the factor(s) raising glucagon levels on the high-protein diets, amino acids themselves are also remarkable, since they directly stimulate glucagon secretion from the perfused dog pancreas (Ipp et al. 1977) and isolated rat α-cells (Pipeleers et al. 1985).

Discussion
A series of initial systematic characterizations (Figures 1-5) revealed that the daily expression rhythms of gluconeogenic enzyme genes Pck1, G6pc, and Tat ( Figure 1) were maintained in constant darkness (Figure 2), modulated by Clock mutation (Figure 3), and synchronized to varied photoperiods ( Figure 4) and time-restricted feeding ( Figure 5). Therefore, the daily expression rhythms of Pck1, G6pc, and Tat are likely under the control of the organism-intrinsic clock, which is entrained to the light/dark cycle, and which in turn entrains the expression of gluconeogenic enzyme genes via the feeding/fasting cycle, as illustrated in Figure 9. Concordantly, we previously demonstrated that varied photoperiods entrain the wake/sleep (activity/rest) cycle and then feeding/ fasting cycle (Matsumoto et al. 2010). Besides, remarkably, the results of time-restricted feedings ( Figure 5) are also concordant with the notion that the ingestion schedule, when separated from the light/dark cycle, can entrain the expression of gluconeogenic enzyme genes, independently from the central clock oscillator.
In addition to the feeding/fasting cycle, other pathways driven by the central clock oscillator are assumed to exist, since even under prolonged fasting (Figures 1b and 2b) daily ( Figure 1b) and circadian expression rhythms (Figure 2b) of Pck1, G6pc, and Tat were detected. A remarkable candidate for the pathway that mediates the effects of the central oscillator apart from the feeding/fasting cycle is that including corticosterone, which is under the control of the SCN-driven HPA axis (Figure 9). Like many other hormones (Veldhuis et al. 2008), plasma corticosterone is secreted in a pulsatile manner, exhibiting ultradian oscillation with a period length of about 1 h (Conway- Campbell et al. 2012;Dickmeis et al. 2013). A series of the ultradian oscillations in a b d c e Figure 9. Schematic illustration of proposed systemic oscillator-driven and nutrient-responsive hormonal regulation of daily expression rhythms for gluconeogenic enzyme genes in the liver. Hypothetical mechanisms for the regulations on the standard diet (a), prolonged fasting (b), the high-fat diet (c), the high-carbohydrate (no-protein) diet (d), and the 60% soy-protein diet as a representative of high-protein diets (e) are shown. Effects, changes in blood levels (elevation or lowering at least at 2 time points), and rhythmicities for humoral factors are shown as indicated in the lower-right diagram. See the text for detailed explanations.
turn exhibit a circadian pattern with a zenith at the sleep-to-wake transition phase and a nadir at the wake-to-sleep transition phase (Halberg et al. 1959;Migeon et al. 1956). This daily profile of corticosterone levels was also observed in various nutritional conditions (Figure 7c) of the present study. Therefore, daily rhythms of corticosterone levels are relatively stable, compared with other hormones (Figure 7c). This stable corticosterone rhythmicity was presumably brought about by oscillatory drive from the SCN to the adrenal cortex through the HPA axis and the autonomic nervous system (Dickmeis et al. 2013;Kalsbeek et al. 2012), and may underlie the expression rhythms of gluconeogenic enzyme genes even in the varied nutritional conditions.
On the standard diet (Figure 9a), corticosterone may trigger the activation of gluconeogenic enzyme genes in the light (sleep-fasting) period. Elevation of insulin levels with a plateau during the dark period ( Figure 7h) likely leads to eventual suppression of the corticosterone-triggered activation of gluconeogenic enzyme genes through differential regulation of these target genes by insulin (O'Brien et al. 2001), consistent with lowered Pck1 mRNA levels in the early dark period and lowered G6pc and Tat mRNA levels in the late dark period (Figure 1a). The ghrelin rhythmicity (Figure 7d) may play a role in the regulation of corticosterone ( Figure 7c) and insulin rhythmicities (Figure 7h), because of stimulation of adrenocorticotropic hormone secretion (Spencer et al. 2015) and inhibition of insulin secretion (Egido et al. 2002;Reimer et al. 2003), respectively, by ghrelin.
Prolonged fasting (Figure 9b) caused severe whole-day hypoglycemia ( Figure 7a) and resulted in general elevation of the mRNA levels of gluconeogenic enzyme genes with phase-shifts of their daily rhythms (Figure 1b). Concordantly, prolonged fasting caused overall elevation of corticosterone levels (Figure 7c), exhibiting attenuated, but still marginal rhythmicity (p = 0.074) with an apparent phase-shift. Insulin showed extreme downregulation in the dark period (Figure 7h), while glucagon exhibited steady levels (Figure 7g). Effects of prolonged fasting on corticosterone levels might be mediated, at least in part, by up-regulation of ghrelin ( Figure 7d) and down-regulation of adipokines such as leptin (Figure 7e) (Roubos et al. 2012). Another adipokine resistin was also down-regulated by the prolonged fasting (Figure 7f), and curiously showed positive correlation with upregulated corticosterone (Figure 8b), physiological significance being left to future study. Upregulation of ghrelin (Egido et al. 2002;Reimer et al. 2003), as well as rhythmicity-gaining downregulation of GLP-1 (Figure 7k) (Campbell and Drucker 2013), is also likely involved in lowering insulin levels.
The high-fat diet (Figure 9c) caused phase-shifts and/or damping of the daily expression rhythms of gluconeogenic enzyme genes, with relatively enhanced expression in the middle of the light (sleep-fasting) period (Figure 6a). These changes were likely underlain by damping of the corticosterone rhythm (Figure 7c) and by atypical insulin up-regulation in the late light (sleep-fasting) period ( Figure 7h). The corticosterone damping is in turn possibly underlain by damping of the ghrelin rhythm ( Figure 7d) (Spencer et al. 2015) and by general elevation of leptin (Figure 7e) (Roubos et al. 2012). Interestingly, corticosterone and leptin formed positive correlation network with resistin ( Figure 8c) that was up-regulated and gained rhythmicity on the high-fat diet (Figure 7f), physiological meaning and formation mechanism remaining to be investigated. As for insulin, candidates for factors causing its atypical upregulation ( Figure 7h (Figure 7k), but with daily profiles that were not necessarily similar to each other. In contrast, down-regulation of resistin (Figure 7f) was selective for prolonged fasting, while downregulation of glucagon ( Figure 7g) and rhythmicity-gaining up-regulation of GIP (Figure 7j) were selective for the high-carbohydrate (no-protein) diet. Therefore, the high-carbohydrate (noprotein) diet, presumably because of resultant hypoproteinosis acting as a stressor, mimics prolonged fasting in regulation of daily profiles of some humoral factors, but not others.
The high-protein diets, represented by the 60% soy-protein diet in Figure 9e, activated gluconeogenic enzyme genes prominently and atypically in the dark (wake-feeding) period (Figure 6c and e) when moderate hypoglycemia was caused (Figure 7a). This activation was likely led by a paradoxical increase in glucagon secretion during the wake-feeding period (Figure 7g). Concordantly, a previous study reported that feeding rats on a high-protein (82% casein, nocarbohydrate) diet during the dark period resulted in a three-fold elevation of glucagon levels 3 h after the feeding onset (Tiedgen and Seitz 1980). Compared to glucagon, other humoral factors exhibited only limited changes in their daily profiles on the high-protein diets (Figures 7 and 9e), but frequently formed correlation networks with glucagon (Figure 8e and g). Plausibly, ingested amino acids in the dark (wakefeeding) period on the high-protein diets directly (Ipp et al. 1977;Pipeleers et al. 1985) and selectively stimulate the secretion of glucagon, which in turn operates with other humoral factors agonistically or antagonistically.
Glucagon and insulin bind to hepatocellular cognate G-protein-coupled receptor and tyrosine kinase receptor, respectively, leading to activation and repression of gluconeogenic enzyme genes through PKA and Akt pathways. A large number of studies have revealed that regulatory regions of Pck1 (Croniger et al. 2002), G6pc (Hutton andO'Brien 2009), andTat (Boshart et al. 1993;Nitsch et al. 1993) genes interact with the glucocorticoid receptor as well as various transcription factors and cofactors including CREB, C/EBPs, FoxO1, CBP, CRTC2, and PGC1α, which are cellular signaling targets of glucagon and/or insulin, and whose regulatory roles for daily expression rhythms of gluconeogenic enzyme genes remain to be clarified. It is also important to investigate the direct and indirect interplays of these hormone-responsive factors with clock proteins, as most typically demonstrated for CRYs (Jang et al. 2016;Lamia et al. 2011;Toledo et al. 2018;Zhang et al. 2010).
In conclusion, we propose that daily expression rhythms of gluconeogenic enzyme genes are under the control of the integrated endocrine system, which synchronizes with the central clock oscillator, and which also responds to the ingested nutrients. This study provides insights for understanding the mechanisms and developing the interventions for regulating daily profiles of blood glucose levels, especially in the context of daily dietary fluctuation and day-to-day nutritional variation.