Background
The 22.9 % incidence of metabolic syndrome has driven research focused on genetic, pharmacologic, stress, and diet induced changes in metabolic status [
1‐
4]. Interpretation and design of these studies requires an in depth understanding of the acute changes in metabolite flux induced by either fasting or meal consumption, yet the literature lacks studies that evaluate the timing of metabolic adaptations to acute changes in food consumption. Fasting durations range from 4 h to 48 h. The commonly reported “overnight fast” is indicative of the disregard for shifting physiological adaptations that accompany fasts of varying duration. Most importantly the food deprivation and resulting metabolic perturbations are often imposed without an explanation of physiological considerations.
The liver is the central tissue maintaining metabolic homeostasis as the animal shifts between fed and fasted states. Fasting induces hepatic glucose and ketone body production, glycogen depletion, and triacylglycerol accumulation [
5‐
7]. The triacylglycerol accumulation is a response to increased adipose tissue lipolysis [
7], while many of the metabolic adaptations are adaptive to prevent hepatic lipotoxicity [
6,
8,
9]. Feeding a carbohydrate containing meal stimulates hepatic glucose uptake and glycolysis, repletes glycogen stores, and induces fatty acid synthesis, while inhibiting β-oxidation and ketogenesis [
10‐
13]. Understanding the elasticity of hepatic metabolite flux and the central role of the liver in providing nutrients to peripheral tissues is essential to studies of metabolic perturbation.
We conducted two sets of studies to delineate the timing of hepatic metabolic adaptations that maintain homeostasis through the transitions from the fed to the fasted and back to the fed state across physiologically relevant durations in the mouse. The first set of studies set out to understand the duration of fast that induces glycogenolysis, gluconeogenesis, β-oxidation and ketogenesis. To assess this, we have measured serum metabolites, hepatic glycogen and triglyceride content, activity of rate limiting enzymes in gluconeogenesis and ketogenesis, and the mRNA expression of enzymes and nuclear hormone receptors that regulate flux through β-oxidation, ketogenesis, and gluconeogenesis. Subsequently, we examined the effects of re-feeding after a fast to understand the hepatic transition from the fasted to the fed state. Together these studies define the timing of changes in hepatic metabolism and aim to encourage informed application of these dietary manipulations to study metabolic diseases.
Methods
Animals
Twelve to fourteen week old male C57BL/6J mice were purchased from The Jackson Laboratory (Bar Harbor, ME). Individually housed mice were exposed to a 14-h light/10-h dark cycle, given ad libitum access to NIH-31 chow (Harlan Laboratories, Indianapolis, IN) and water, and adapted to the environment for 1 week prior to study initiation. Mice were housed on wood chip bedding (Harlan Laboratories; Cat #7090 Sani-Chips) to limit any potential consumption of nutrients from bedding during the fasting period. We initiated fasting at either 4, 8, 12, or 16 h prior to sacrifice 4 h after lights on for all mice. In the fast-refeed study, mice were given access to food at 4 h after lights on for 0, 1, or 2 h after 0, 8, or 16 h of fasting. These studies were approved by the Institutional Animal Use and Care Committee at the University of Arizona.
Sample collection and storage
We anesthetized mice with isoflurane using the bell-jar method and sacrificed mice by decapitation to collect trunk blood. The blood clotted at 4 °C overnight. To collect serum, we centrifuged the blood at 3,000 × g for 30 min. Serum was stored at −80 °C until analysis. Whole liver was collected, immediately frozen on dry ice, and stored at −80 °C. To obtain a homogenous liver sample, we powdered whole frozen liver using a liquid nitrogen cooled mortar and pestle.
Serum analyses
We used colorimetric assays to analyze serum β-OH butyrate (Cat. # 700190, Cayman Chemicals, Pittsburg, PA), glucose (Cat. # G7519, Pointe Scientific Inc., Canton MI), non-esterified fatty acids (NEFA; HR Series NEFA-HR, Wako Diagnostics, Richmond, VA), and triacylglycerols (Cat# T7531, Ponte Scientific Inc., Canton, MI).
Liver analyses
We extracted total liver mRNA with TRI Reagent® (Life Technologies, Grand Island, NY), performed reverse transcription using Verso cDNA synthesis kit (Thermo Scientific, Inc., Waltham, MA) and performed real-time PCR using SYBR 2X mastermix (Bio-Rad Laboratories, Hercules, CA) and the Biorad iQ™5 iCycler (Bio-Rad Laboratories, Hercules, CA). Prior to initiating the reverse transcription reaction, RNA was cleared of any phenol contamination using a water saturated butanol and ether method [
14]. Table
1 list the primers used to analyze expression of β-actin (ACTB), D-beta-hydroxybutyrate dehydrogenase type 1 (BDH1), D-beta-hydroxybutyrate dehydrogenase type 2 (BDH2), 3-hydroxy-3-methyl glutaryl CoenzymeA synthase II (HMGCS2), peroxisome proliferator activated receptor α (PPARα), uncoupling protein 2 (UCP2), glucose 6-phosphatase (G6Pase), phosphoenolypyruvate carboxykinase (PEPCK), and carnitine palmitoyltransferase (CPT1) mRNA. Raw amplification data was imported into LinReg PCR analysis software to establish efficiency of amplification [
15] and output data was converted to fold change in expression using the efficiency
-∆∆Ct method and with β-actin as the housekeeping gene [
16].
Table 1
Primer sequences for real-time PCR
β-actin | TCGGTGACATCAAAGAGAAG | GATGCCACAGGATTCCATA | 11461 |
β-OH Butyrate Dehydrogenase 1 | AGGCTGTGACTCTGGATTTGGG | CTGGATGGTTCTCAGTCGGTCA | 71911 |
β-OH Butyrate Dehydrogenase 2 | AGGAGCTGGAAAGACCGAGG | TCGCAATCCAGGATGGTTCCGT | 69772 |
3-hydroxy-3-methylglutaryl-CoA Synthase II | AGAGAGCGATGCAGGAAACTT | AAGGATGCCCACATCTTTTGG | 15360 |
Peroxisome Proliferator Activated Receptor α | AGAGCCCCATCTGTCCTCTC | ACTGGTAGTCTTGCAAAACCAAA | 19013 |
Uncoupling Protein 2 | ATGGTTGGTTTCAAGGCCACA | CGGTATCCAGAGGGAAAGTGAT | 22228 |
Glucose 6 Phosphatase | CGACTCGCTATCTCCAAGTGA | GTTGAACCAGTCTCCGACCA | 14377 |
Phosphoenolpyruvate Carboxykinase | CTGCATAACGGTCTGGACTTC | CAGCAACTGCCCGTACTCC | 18534 |
Carnitine Palmitoyltransferase I | CTCCGCCTGAGCCATGAAG | CACCAGTGATGATGCCATTCT | 12894 |
We used the Folch method to extract total lipid from powdered liver [
17]. Extracted lipids were assayed for triacylglyerol (Cat# T7531, Ponte Scientific Inc., Canton, MI) and expressed as mg triacylglycerol/g liver. Liver NEFA content was assessed by homogenizing 10–20 mg of powdered liver tissue in 10 volumes 0.1 M phosphate buffered saline. NEFA were extracted from the tissue homogenate by vortexing with 10 volumes of 100 % ethanol for 20 min. Subsequently, NEFA were measured in 50 μl of ethanol using a commercially available non-esterified fatty acid kit (HR Series NEFA-HR, Wako Diagnostics, Richmond, VA). Dilution in ethanol rather than phosphate buffered saline did not affect the absorbance resulting from the standards, but standards were diluted in 100 % ethanol so that samples and standards were in the same diluent. Liver glycogen content was measured using a previously described colorimetric assay [
18].
Using previously described enzyme activity assays that rely on the NADH to NAD
+ ratio, we measured the enzymatic drive of acetoacetate to β-OH butyrate (BDH1 activity) and gluconeogenic potential from tricarboxylic acid (TCA) cycle intermediates (PEPCK activity) in powdered liver tissue homogenates as previously described [
19,
20]. To measure liver adenosine triphosphate (ATP) content, we homogenized liver in somatic cell ATP releasing agent (Cat. FLASR, Sigma Chemical Co., St. Louis, MO) and measured ATP using an ATP Determination Kit (A22066, Molecular Probes, Eugene, OR) with luciferase read in real time on Clarity™ Luminescence Microplate Reader (BioTek Instruments, Winooski, VT). Liver cyclic adenosine monophosphate (cAMP), an integrative measure of hormone signaling within the hepatocyte, was measured in powdered liver tissue by enzyme-linked immunosorbent assay (ELISA; ADI-900-066, Enzo Life Sciences, Farmingdale, NY) and expressed per gram of tissue.
Statistical analysis
We analyzed the effect of fasting duration on all dependent variables using the mixed model in SAS Enterprise Guide 4.3 (SAS Institute Inc., Cary, NC). Probabilities of differences between means were determined using Tukey’s adjustment for multiple comparisons. The effects of re-feeding were analyzed with a two-way ANOVA including fasting duration (0 and 16 h) and re-feeding duration (0, 1, and 2 h) and their interaction as the main effects. The probabilities of differences between means were assessed within a fasting duration and within a re-feeding duration. Accordingly, a Bonferroni correction was employed. Independent variables were identified as classification variables in both models. All raw data was plotted in Graphpad PRISM® Version 5.04 for Windows (GraphPad Software, San Diego California USA,
www.graphpad.com).
Discussion
Here we show the timing of the fasting induced metabolic shift toward gluconeogenesis, glycogenolysis, β-oxidation, and ketogenesis in the mouse. Flux through each of these pathways was evaluated using biochemical indicators of pathway activity, enzyme activity assays, and measurement of mRNA expression. Gluconeogenic potential was assessed by the change in PEPCK mRNA and activity, glycogenolytic potential by the change in hepatic glycogen content, and the potential for hepatic glucose output by G6Pase mRNA [
5]. Elevated mRNA expression of gluconeogenic and ketogenic enzymes translates functionally into increased flux through these metabolic pathways. In fact, mice overexpressing PEPCK mRNA, have elevated hepatic glucose output both in vivo and in vitro [
5,
33], while overexpression of G6Pase mRNA results in hyperglycemia, hyperinsulinemia, and an over 50 % reduction in liver glycogen [
34,
35]. Overexpression of CPT1 or HMGCS2 increases hepatic β-oxidation and ketone synthesis [
36,
37], while HMGCS2 knockdown completely eliminates the fasting induced rise in serum β-OH butyrate [
38]. We measured mRNA for CPT1, a fatty acid translocase whose expression controls the rate of fatty acid flux into the mitochondria and the flux through β-oxidation [
39]. Serum β-OH butyrate concentration, BDH1 activity, and expression of mRNA encoding HMGCS2, BDH1, and BDH2 were measured as indicators of ketogenic potential from acetyl-CoA.
The 24 or 48 h food deprivation that is commonly employed in rodent studies appears to better model starvation than fasting [
40‐
44]. In fact, serum NEFA concentrations are similarly elevated by 8 h of fasting in the mouse and 24 h of fasting in the human [
45]. In rats, G6Pase mRNA is increased 3.5 times in 24 h fasted rats and not further increased by 48 h fasting [
32], while 48 h fast increases CPT1 mRNA 7.5 times relative to the fed animal [
46]. We observed similar 3.4 and 7.4 times increases in G6Pase and CPT1 mRNA, respectively, at 16 h fasting in the mouse (Figs.
2b and
3d). Thus, maximal changes in mRNA expression induced by fasting may be achieved within 16 h in mice. We conclude that 8 h fasting in the mouse represents a realistic time point for metabolic adaptations which occur early in a fast, as hepatic triglyceride accumulation and serum β-OH butyrate are minimally elevated and hepatic glycogen reserves are not yet depleted. In the mouse, 16 h of fasting, represents a complete induction of the fasting response, as hepatic triglyceride and serum β-OH butyrate concentrations are robustly elevated and hepatic glycogen stores are exhausted. Thus, refeeding after 8 h and after 16 h represents the transition from the fasted back to the fed state after a partial and complete activation of hepatic fasting metabolic adaptations, respectively.
Serum glucose homeostasis at the onset of fasting relied heavily on depletion of hepatic glycogen content, which was almost fully exhausted by 12 h of fasting (Fig.
2a). When the fasting duration exceeded 8 h, we observed increases in hepatic gluconeogenic potential (PEPCK activity and mRNA expression, Fig.
2c and
d) from TCA cycle intermediates. When re-fed, the increase in hepatic glycogen combined with a decrease in G6Pase and PEPCK mRNA indicate inhibition of hepatic glucose output. Elevated serum insulin immediately stimulates hepatic glycogen synthesis upon refeeding, and glycogen content is restored to the level of fed animals by 5 h of refeeding [
47,
48]. Interestingly, hepatic glycolysis remains low during the initial refeeding phase, while gluconeogenesis remains active until hepatic glycogen levels are restored [
47,
48]. In fact, early glycogen repletion is a consequence of maintained hepatic gluconeogenesis [
48‐
50]. Thus, although PEPCK and G6Pase mRNA expression decreases immediately upon termination of a fast, the switch from gluconeogenesis to glycolysis does not occur until several hours following refeeding. During early fasting hepatic glucose output is dominated by glycogenolysis and replenishment of glycogen stores through glycogenesis is prioritized upon refeeding. Therefore, hepatic glycogen appears to play a central role in maintenance of short-term glucose homeostasis during transitions between the fed and fasted state [
51].
Re-feeding immediately and robustly inhibits adipose tissue lipolysis and hepatic ketogenesis as observed by changes in serum NEFA and β-OH butyrate (Fig.
5b and
d). The declines in hepatic CPT1 and HMGCS2 mRNA expression parallel the decline in β-OH butyrate, yet the decline in hepatic ketogenesis is more robust and rapid than explained by changes in gene expression alone (Fig.
7d and
e). The inhibition of ketogenesis is not dependent on a depression in hepatic β-oxidation, as high levels of β-oxidation and low rates of de novo lipogenesis have been reported to continue several hours into refeeding [
47,
52,
53]. More likely, it results from insulin’s inhibition of HMGCS2 activity [
54].
The pancreatic hormones, insulin and glucagon, are critical mediators in coordinating the systemic response to changes in nutritional status, with the liver being a primary site of action. The insulin:glucagon ratio decreases with fasting and increases upon refeeding [
55,
56]. Regulation of metabolite flux, enzyme activity, and gene expression is exerted through changes in intracellular cAMP. Glucagon increases cAMP concentrations through G
αs signaling at its receptor while insulin decreases cAMP concentrations by enhancing phosphodiesterase activity [
57]. Elevated cAMP directly upregulates transcription of PEPCK, G6Pase, CPT1, and HMGCS2 through identified cAMP response elements (CRE) in the promoter region of these genes [
22,
58‐
60]. The observed increase in hepatic cAMP with fasting and decline upon refeeding reflect glucagon and insulin mediated control of gene transcription (Figs.
2e and
6e).
In addition to cAMP signaling through CRE, a number of hormonally regulated transcription factors control expression of gluconeogenic, β-oxidative, and ketogenic enzymes. When fasted, decreased insulin and increased glucagon result in dephosphorylation of forkhead box proteins (FoxO) and class IIa histone deacetylases (HDAC), respectively. This dephosphorylation allows nuclear translocation of these proteins and upregulation of G6Pase and PEPCK mRNA expression [
61]. FoxA2, inhibited by insulin dependent phosphorylation, stimulates transcription of β-oxidative, and ketogenic enzymes [
62]. This is merely a short list of insulin and glucagon regulated transcription factors meant to demonstrate the central role of glucoregulatory hormone signaling in orchestrating the hepatic mRNA transcript expression in transitions between the fed and fasted state [
63].
Peroxisome proliferator activated receptor α (PPARα), a nuclear hormone receptor that is activated by non-esterified fatty acids (NEFAs), is another transcription factor central to the metabolic shift initiated by fasting [
6]. Our results propose that liver NEFA concentrations are very sensitive to the initiation of a fast, as NEFA concentrations increased 5.4 times within 4 h of fasting. (Fig.
3b). NEFA activated PPARα binds to the promoter and encourages expression of target genes involved in flux through gluconeogenesis, β-oxidation, and ketogenesis [
6,
64‐
67]. Indicative of the integral role for PPARα in the gluconeogenic response to fasting, PPARα null mice display fasting hypoglycemia [
6,
64,
68]. PPARα null mice also lack the ability to properly transition to ketogenesis despite normal NEFA mobilization from adipose tissue [
6,
69]. PPARα induced expression of UCP2 is equally important to support the increased hepatic lipid oxidation and ketogenesis of fasting [
70]. In fact, fasting does not increase serum β-OH butyrate in the UCP2 knockout mouse [
71]. By uncoupling oxidative phosphorylation from the electron transport chain, UCP2 allows unbridled oxidation of NADH to NAD
+, increasing the pool of NAD
+ and allowing efficient oxidation of fatty acids to acetyl CoA through β-oxidation. Thus, PPARα signaling works to limit the potential hepatotoxic effects of lipid accumulation by enhancing lipid oxidation and ketone body synthesis. PPARα null mice display a muted gluconeogenic and ketogenic response to fasting induced lipid accumulation. Like fasting, obesity is characterized by hepatic lipid accumulation, hyperketonemia, enhanced hepatic glucose production, and decreased hepatic ATP content resulting from increased expression of UCP2 [
5‐
9,
31,
72‐
75]. These metabolic changes are downstream of increased PEPCK mRNA, protein, and activity, G6Pase mRNA and protein, and CPT1 and BDH1 mRNA and appear to be common to hepatic lipid accumulation [
32,
76‐
79]. Mice fed diets high in fructose, sucrose, or fat all develop hepatic lipid accumulation and aberrantly overexpress hepatic PEPCK, G6Pase, and UCP2 mRNA [
80‐
82]. Interestingly, PPARα null mice, which are unable to properly upregulate gluconeogenesis and ketogenesis in response to a fast, are protected from metabolic responses (hyperglycemia and hyperketonemia) common to obesity induced lipid accumulation. Thus it appears that PPARα is integral for the metabolic adaptations/maladaptations (increased ketogenesis and gluconeogenesis) in response to either fasting or obesity-induced lipid accumulation. Given that many of the metabolic pathways that are active during fasting are also active in obesity, careful consideration must be applied toward study design and data interpretation when food depriving diet- or genetically induced obese mice.