The major architectural feature of fasting liver is a pronounced decline in cell size (down to ~75% of its fed diameter after 3 days of fasting) rather than a loss of cell number. In addition, the liver's metabolic zonation in upstream, periportal and downstream, pericentral regions remains intact. These findings indicate that the liver can quickly resume its homeostatic functions once feeding resumes. Our microarray data show that the adaptive response of the liver to fasting at the level of gene expression is most pronounced during the early phase, with the upregulation of ammonia detoxification persisting up to 72 hours of fasting. Since the technically similar study of Bauer et al.  reported enhanced expression of lipid-catabolizing and urea-cycle enzymes after 24 and 48 h of fasting, the collective data show that the response to fasting in the liver starts already at 12 hours of fasting and becomes maximal between 24 and 48 hours. The hepatic response to total food deprivation, therefore, does not proceed through the global "sugar-fat-protein" sequence that is described for the adaptation on the whole-body level [1–5].
The expression of genes involved in lipid metabolism and ketone-body synthesis, many under PPARα coordination [29, 26], was strongly regulated towards fatty-acid oxidation and ketone-body formation. Interestingly, this adaptive response was seen between 12 and 48 hours ( and present study) of fasting only, and then faded out. In the rat, this response was recently reported to occur between 3 and 5 days of fasting . Fatty-acid oxidation was accommodated by the identical time frame of the upregulation of the expression of TCA-cycle enzymes and the proteins of the electron-transport chain in response to fasting. The associated oxidative stress and mitochondrial radical formation was apparently sufficiently strong to induce the unfolded-protein (ER stress) response in liver. Thus far, to our knowledge, activation of the unfolded protein response has not been associated with fatty-acid oxidation in the fasting liver, but it is induced by a high-fat diet . Similarly, mitochondria in fasting muscle protect themselves against the oxidative stress that results from fat oxidation  by accumulating the uncoupling proteins UCP2 and UCP3 [33, 34].
The role of the liver in gluconeogenesis during fasting is well documented [35–37]. However, the expression of enzymes associated with gluconeogenesis was upregulated only during the first day of fasting and was mainly confined to the malate-aspartate shuttle and Pepck1. In fact, apart from Pepck1, the expression of none of the committed steps in gluconeogenesis was regulated. It is likely that the enhanced expression of TCA-cycle and malate-aspartate shuttle enzymes, and the enhanced expression of Pepck1 enhance the flux towards either glucose-6-phosphate or lactate. It is, therefore, remarkable that the expression of glucose-6-phosphatase, a periportal enzyme, and pyruvate kinase, a mainly pericentral enzyme, are not regulated, while the expression of lactate dehydrogenase is only upregulated at 12 and 24 hours. Similarly, the expression and activity of glucose-6-phosphatase in rat liver are upregulated mildly during the first 48 hours of fasting only . Our unpublished data show a similar response in the kidneys of fasting mice, in which Pepck1 is 2–3 fold upregulated at all time points, whereas glucose-6-phosphatase is not regulated. The pronounced accumulation of glycogen in pericentral hepatocytes starting after 24 hours of fasting, which was also observed in 72- and 96-hour fasted rats , indicates that pericentral hepatocytes, which do not express glucose-6-phosphatase [40, 41], channel glucose-6-phosphate towards glycogen. Since all relevant enzymes are also expressed in periportal hepatocytes, which do not accumulate glycogen, we assume that these hepatocytes contain enough glucose-6-phosphatase to produce glucose.
The liver produces ~60% of the newly produced glucose in starvation, while the kidneys account for ~40% . A recent, but controversial series of experiments suggest that, in addition to the liver [35–37] and kidney [35, 42], the small intestine also has the capacity to produce glucose upon prolonged fasting [43, 6]. It contributes indirectly, by providing lactate and alanine to the liver in short-term fasting [17, 44], and directly by the production of glucose  (perhaps up to 27% of whole-body glucose production in extended fasting in the rat ). The concept is controversial, since other studies were unable to detect glucose formation from glutamine in the isolated small intestine of 72 hours fasted rats . Furthermore, the expression of the key gluconeogenic enzyme phosphoenolpyruvate kinase (Pepck1) in the mouse small intestine was reported to amount to only 0.05–1% of that in the liver after 12 h hours of fasting , also arguing against intestinal gluconeogenesis. We, therefore, compared Pepck mRNA levels in these two organs by qPCR in the fed and 3-days fasting condition (Table 2). While Pepck1 expression in the gut at 12 hours of fasting only amounted to ~5% of that in liver, its expression increased to 18 and 53% of that in the liver after 24 and 72 hours of fasting, respectively. This finding demonstrates that the issue of intestinal gluconeogenesis during prolonged fasting deserves additional study.
The plasma concentrations of both glucose and lactate remained unchanged during the first 24 hours of fasting, declined temporarily by 35–40% at 48 hours, and returned to control values between 48 and 72 hours. The maintenance of normal concentrations of glucose and lactate during the first 24 hours of fasting is most likely the result of gluconeogenesis in the liver and kidney. Since the expression of enzymes that are shared by the glycolytic and gluconeogenic pathways, declines after 24 hours of fasting, the observed decline at 48 hours may represent a declining contribution of the liver to gluconeogenesis. As we argued earlier, the accumulation of glycogen in the pericentral hepatocytes between 24 and 72 hours of fasting indicates that gluconeogenic intermediates flow towards glucose-6-phosphate and accumulate as glycogen due to the low expression of glucose-6-phosphatase in these hepatocytes. Most likely, a similar or higher flow of gluconeogenic intermediates is present in the periportal hepatocytes, but is exported as glucose due to the high concentration of glucose-6-phosphatese in these cells. Furthermore, the putative production of glucose in the 72-hour fasted intestine can also contribute to the circulating glucose level.
The urea-cycle enzymes distinguish themselves from most other genes in the liver in that they were upregulated in expression throughout the period of fasting that was studied. Furthermore, cytosolic glutamate-oxaloacetate transaminase, which mediates the availability of aspartate to the urea-cycle enzyme argininosuccinate synthetase, was also strongly upregulated at 72 hours. Similarly, Oat and Prodh, which supply glutamate to glutamine synthetase for glutamine synthesis in pericentral hepatocytes, were strongly upregulated at all time points studied, but glutamine synthetase itself was not regulated (and even downregulated in another study ). Since few amino-acid catabolizing enzymes were upregulated (the exception being the metabolism of sulphur-containing amino acids), most amino-groups were probably carried to the liver as alanine or glutamine, although neither glutamate-pyruvate transaminase nor liver glutaminase was upregulated. The coordinate control of ammonia detoxification and the source of ammonia during prolonged fasting therefore deserve attention.
An important question is to what extent we can extrapolate the observations in a small mammal like the mouse to larger animals like humans. The ability to tolerate the absence of food does indeed decline with body size: in the mouse the maximum duration of fasting is 4 days , in rat 12–15 days , in children 4 weeks and in adult humans 8–9 weeks [49, 13]. Qualitatively, however, the response to fasting is probably comparable between these mammals, as long as the time scale is adjusted to the size of the animal. Rather than questioning the comparability of small and large animals, our data question whether the implicit extrapolation of the "sugars-fats-proteins" succession of energy substrates during fasting that is based on whole-body energy expenditure [1, 50] to individual organs is valid. Microarray studies in rodents that have prospected the adaptive response to fasting of the small intestine , liver ( and present study), muscle [23, 24, 51], and a more limited study in kidney focusing on circadian differences in gene expression , reveal a different scenario. Muscle and kidney respond to fasting with a progressive change over time in mRNA concentrations of enzymes involved in protein, carbohydrate and fat metabolism. The response in liver peaked at 24–48 hours of fasting in mouse, while most adaptive changes had abated by 72 hours. The intestine, finally, showed an early, but temporary peak of adaptive changes in amino-acid, carbohydrate and fat metabolism at 12 hours of fasting, while a late response, existing almost exclusively of amino-acid catabolizing and gluconeogenic enzymes, gradually developed towards 72 hours of fasting. These differences in pattern and amplitude of gene expression change in different organs can be used to look for circulating biomarkers that reflect the functions of organs during adaptive responses.