Effects of food withdrawal on liver structure
During the first 12 hours of fasting, mice lost ~12% of their body weight (that is, 24% if expressed on a per-day basis). Thereafter, weight loss remained steady at a rate of ~7% per day, so that mice had lost ~30% of their initial weight after 72 hours of fasting (Figure ). Note that we expressed daily differences in the rate of weight loss on a per-day basis to define a common denominator for the 12 h- and 24 h-fasted animals. Liver wet weight declined more than body weight (Figure ), especially during the first 12 hours of fasting, and amounted to ~44, ~5, ~11 and ~10% per day after 12, 24, 48 and 72 hours of fasting, respectively. After 72 hours, the liver had, therefore, lost almost 50% of its initial weight. The basic architecture of the liver lobules (Figure , HA) and the zonation of gene expression as studied by the expression of glutamine synthetase and carbamoylphosphate synthetase (Figure , GS and CPS) remained unaffected. Staining for the appearance of active-caspase 3 revealed no changes in the number of apoptotic cells upon fasting, not even after 72 hours (data not shown). In agreement, the apoptotic genes that were represented on the microarrays showed no significant change in expression in fasted compared to fed mice. Since there was no reduction in the number of liver cells during fasting, we took two approaches to estimate the decrease in average cell size. The summation of 50 hepatocyte diameters in three 72 hours fasted and three control animals, amounted to 25% reduction in cell diameter. Based on the liver wet weight, the average cell diameter decreased 20% in the fasted animals.
Figure 1 Macro- and microscopic analysis of the fasting liver. A) Change in whole-body and liver weight during fasting as percentage of fed weight (n ≥ 8). Asterisks label significant changes (P < 0.01). The blue and pink lines represent the daily (more ...)
Effects of fasting on metabolism
Ammonia levels had increased 2.0-, 3.7- and 5.2-fold after 24, 48 and 72 hours of fasting, respectively (P < 0.005; Figure ). Glucose and lactate concentrations remained stable until 48 h of fasting, but decreased 34 and 43%, respectively (P < 0.05 and 0.005, respectively, Figure ) in the next 24 hours. The plasma concentration of many amino acids changed at some time point of fasting, but only the changes in the concentration of taurine showed a trend with time (Figure and Supplementary Table , Additional file 1
). Accumulation of taurine helps protect cells from hypertonicity [25
], as may occur during shrinkage of fasting hepatocytes.
Figure 2 Changes in plasma metabolite concentrations during fasting. A) Glucose and lactate (mM, primary Y-axis), and ammonia concentrations (μM, secondary Y-axis) after 0, 12, 24, 48 and 72 hours of fasting. B) Adaptive changes in concentrations of a (more ...)
Top 10 canonical pathways influenced by fasting
Global gene-expression profile in the liver
To gain a comprehensive overview of the physiological response of the liver to fasting, whole-genome measurements were made. Compared to the fed group, 201, 504 and 119 transcripts, including expressed sequence tags and RIKEN sequences, met our boundary condition for significance (≥ 1.4-fold change with P < 0.01) after 12, 24, and 72 hours of fasting, respectively (Figure ; for a complete list of more than 1.4-fold up- or downregulated genes, see Additional file 2
). The dendrogram generated by supervised hierarchical clustering (Figure ) shows a clear separation between fed and fasted conditions. Among the arrays coming from fasted animals, those from 72 hours stand out, while the branches of the two earlier time points are intertwined, indicating that expression profiles are rather similar after 12 and 24 hours of fasting. This is also reflected in the Venn diagrams where the overlap between 12 and 24 hours is larger than the overlap with 72 hours.
Figure 3 Number of genes in the liver that are affected by fasting. A) Number of differentially expressed genes (≥ 1.4-fold change in expression; P < 0.01) at each time point studied. The left-sided Venn diagram shows the number of up-regulated (more ...)
Global analysis reveals a strong early and an abated late response to fasting
We used GenMAPP and, in particular, MetaCore™ software to deduce the biological processes that change with an increasing duration of fasting from the liver transcriptome data. In MetaCore, the degree of association of the uploaded datasets with predefined metabolic pathways is defined by P-values, with lower P-values being more relevant. The expression of 465 genes that met our thresholds (56%) could be linked to the MetaCore™ suite. Their distribution across time points is shown in Figure . The graphs show the numbers of unique, similar and common genes for all three, and for two initial time points separately, showing that the response to fasting at 24 hours was similar to, but more pronounced than that at 12 hours.
Figure 4 Adaptive changes in metabolic processes in the liver during fasting as analyzed by MetaCore™ software. A) The gene content imported to MetaCore™ is aligned between the time points. The parameters for comparison are ≥ 1.4 fold change (more ...)
We performed gene-set enrichment analysis in three different functional ontologies using MetaCore™: cellular processes, biological processes and canonical pathways. Based on the Gene Ontology categorization of cellular processes, fasting predominantly affected the metabolic processes, in particular the carboxylic-acid metabolizing processes, lipid and glucose metabolism. The enrichment analysis for biological processes showed, more specifically, that genes involved in amino-acid, lipid, carbohydrate and energy metabolism responded most significantly to fasting (Figure ). The graph presents P-values as parameter of the likelihood that coordinate changes in the pathways shown were indeed present at the different time points of fasting. As statistical parameter, the P-value encompasses no variation. The changes in all processes except amino-acid metabolism showed a response that peaked at 24 hours after food withdrawal and declined thereafter. The response during the late phase of fasting was dominated by amino-acid metabolism, although lipid and carbohydrate metabolism remained significantly regulated. The Figure further reveals that the changes in energy metabolism were significant at 24 hours of fasting only. The common denominator of the overall fasting response was, therefore, metabolism of amino acids, carbohydrates, and lipids.
Since the global analysis does not reveal a direction in the changes and lacks functional detail, we scrutinized the pathways with most pronounced regulation for functional implications. A list of the 10 top-scoring canonical pathways, shown in Table , points to gluconeogenesis, urea synthesis, and PPARα-regulated fatty-acid oxidation as the major characteristics in the response of the liver to fasting.
Amino-acid catabolism and urea synthesis
Of all the pathways studied in the liver, the adaptive changes in amino-acid metabolism persisted throughout the fasting period (Figure ). Of the enzymes in this group, those of the urea cycle were upregulated at all three time points (Figure ). Among the genes consistently affected were argininosuccinate synthetase 1 (Ass1, Assy; 3.7-, 2.5- and 4.5-fold upregulated) and argininosuccinate lyase (Asl, Arly; 5.0-, 5.8-, and 12-fold upregulated at 12, 24 and 72 hours, respectively. The first and rate-determining enzyme of urea cycle, carbamoylphosphate synthetase (Cps), was not represented on the microarrays, but its expression level, as estimated by qPCR, was increased 3.5-fold at all 3 time points (manually added to Figure ). Urea synthesis occurs in periportal hepatocytes, whereas ammonia detoxification via glutamine synthesis occurs pericentrally. Genes for the pericentral enzymes ornithine-aminotransferase (Oat) and proline dehydrogenase (Prodh), which provide glutamate for glutamine synthesis, were upregulated 2.5-, 3.0- and 3.0-fold and 2.1-, 2.5- and 2.0- fold at 12, 24 and 72 hours, respectively. Glutamine synthetase (Glns) itself was, however, not regulated.
Figure 5 Amino-acid catabolism in fasting liver. The map was created in the GenMAPP suite to show a comprehensive overview of amino-acid metabolism in response to fasting. Warm colors (from yellow to red) represent down-regulation, while cold colors (light blue (more ...)
Remarkably, the expression of amino-acid catabolizing enzymes themselves was barely affected by fasting. Only the degradation of branched-chain keto-acids (products of branched-chain amino-acid transamination elsewhere) was upregulated, as shown by the upregulation of acetyl-coenzyme A dehydrogenase, medium chain (Acaddm), enoyl-coenzyme A, hydratase/3-hydroxyacyl-coenzyme (Ehhadh), hydroxyacyl-coenzyme A dehydrogenase, short chain (Hadhsc), acetyl-coenzyme A acyltransferase 1 (Acaa1), and 3-hydroxy-3-methylglutaryl-coenzyme A lyase (Hmgcl) – all within first the 24 hours (Figure ). This finding suggests that the adaptations in amino-acid catabolism during fasting mainly occur outside the liver. Since neither glutamate-pyruvate transaminase nor ammonia-inducible liver glutaminase were upregulated, the capacity of the liver to deaminate the amino-carriers alanine and glutamine must have been sufficient.
TCA cycle and electron-transport chain
The strong induction of the urea cycle suggests a strong stimulation of amino-acid oxidation or gluconeogenesis. In agreement with this hypothesis, both the expression of enzymes of the tricarboxylic-acid (TCA) cycle and oxidative phosphorylation were induced in fasted liver, again mainly at 24 hours. Aconitase 2 (Aco2), isocitrate dehydrogenase 3β (NAD+) (Idh3b), oxoglutarate dehydrogenase (Ogdh), dihydrolipoamide S-succinyltransferase (Dlst), fumarate hydratase 1 (Fh1) and malate dehydrogenase 1 (Mdh1) were all upregulated at 24 hours of fasting (1.9-, 1.5-, 3.1-, 2.0-, 1.4- and 1.6-fold, respectively; Figures and ), indicating an increased capacity of the cycle. Dlst and Fh1 were 2.0 and 1.6 times induced at 12 hours of fasting, while Aco2 expression was also 1.8-fold increased at 72 hours.
Fasting upregulates genes of the malate-aspartate shuttle and the gluconeogenic enzyme Pepck1 in mouse liver. The color code of the GenMAPP view is the same as in Figure 5.
In agreement with an increased capacity of the TCA cycle, the expression of the genes of the electron-transport chain was strongly stimulated (Figure ; a legend for the MetaCore canonical pathways is provided in Additional file 3
). Four genes belonging to NADH-ubiquinone oxidoreductase complex: NADH dehydrogenase [ubiquinone] 1α subcomplex subunit 10 (Ndufa10
), NADH dehydrogenase [ubiquinone] 1α subcomplex subunit 13 (Ndufa13
), NADH dehydrogenase [ubiquinone] flavoprotein 1 (Ndufv1
) and NADH dehydrogenase [ubiquinone] flavoprotein 2 (Ndufv2
), were all approximately 1.6-fold upregulated. Expression of the genes of the ATP synthase complex, ATP synthase subunit α (Atp5a1
), ATP synthase δ chain (Atp5d
) and ATP synthase lipid-binding protein (Atp5g1
), was 1.6–1.9-fold induced at 24 hours. Ubiquinol-cytochrome-c reductase complex core protein 1 (Uqcrc1
) was 2.2-fold upregulated after 24 hours, whereas the energy-dissipating uncoupling protein 2 (Ucp2
) was 1.8-fold downregulated at this time point (Figure ). Taken together, these data indicate that the capacity for ATP synthesis in the liver is strongly upregulated during the first day of food deprivation.
Figure 7 Electron-transport chain. Experimental data are visualized on a MetaCore map as blue (for downregulation) and red (upregulation) histograms ('thermometers'). The height of the histogram corresponds to the relative expression value for a particular gene, (more ...)
PPARα regulation of lipid metabolism in fasting. The figure description is the same as in Figure 7.
Phosphoenolpyruvate carboxykinase 1 (Pepck1), a key enzyme in the gluconeogenic route, was upregulated 2.0-, 2.5- and 2.7-fold on the microarrays and 3.2-, 3.2, and 2.9-fold in the qPCR measurements at 12, 24 and 72 hours of fasting, respectively (Figures , and , and Table ). Cytosolic glutamate oxaloacetate transaminase 1 (Got1) was also strongly upregulated at all three time points (5-, 6-, and 21-fold). In addition, malate dehydrogenase (Mdh) and mitochondrial glutamate oxaloacetate transaminase (Got2) were induced (1.6- and 1.8-fold respectively). Together, these data suggest an increased capacity of the malate-aspartate shuttle across the mitochondrial membrane, which would accommodate an enhanced carbon flux from the mitochondria.
Comparison of intestinal and liver Pepck1 expression in fasting by qRTPCR, expressed in relative units after normalization by 18 S expression (n = 6)
All other steps that were affected by fasting were shared by the glycolytic and gluconeogenic pathways and were regulated during the first day of fasting only (Figure ). Phosphoglycerate mutase 1 (Pgam1) was 1.4-fold upregulated at 12 hours of fasting, while glucosephosphate isomerase 1 (Gpi1), aldolase 1A isoform (Aldoa), triosephosphate isomerase 1 (Tpi1), glyceraldehyde-3-phosphate dehydrogenase (Gapdh) and enolase 1α (Eno1) were 1.5-, 1.7-, 1.5-, 1.7-, 2.1- fold upregulated at 24 hours of fasting, respectively. These data indicate that the enhanced capacity of the gluconeogenic pathway would largely depend on enhanced TCA and malate-aspartate cycling, and that this adaptive response in gene expression might be restricted to a single day in the mouse.
Liver glycogen accumulation upon prolonged fasting
The near total return to "normalcy" of gene expression at 72 hours (only the genes for urea cycle enzymes, glutamate-synthesizing enzymes, and Pepck1 remained induced) was striking. Because glucose-6-phosphatase expression was not upregulated, we explored the possibility that glucose precursors were channelled into glycogen. As expected, (amylase-sensitive) periodic acid-Schiff (PAS) staining showed the complete disappearance of glycogen from the liver after 12 hours of fasting (Figure ), but some staining had returned at 24 hours and intense staining was seen in 72-hours fasted liver. Whereas glycogen was localized around the portal veins in fed liver, it was deposited exclusively around the central veins after 72 hours of fasting, with sharp borders towards the empty cells.
Figure 9 Glycogen storage in mouse liver during fasting. In fed liver, periportal hepatocytes contain most glycogen (left panel). Twelve hours of fasting totally depletes the glycogen stores, but at 24 hours, glycogen starts to re-accumulate and has accumulated (more ...)
Fatty-acid catabolism and ketone-body synthesis
The enhanced expression of fatty-acid catabolizing enzymes was also limited to the initial phase of fasting. The expression of the transcription factor Pparα, a major regulator of fatty-acid oxidation, was 2.1-fold upregulated at 24 hours of fasting (Figure ). Furthermore, the mitochondrial carnitine/acylcarnitine fatty-acid translocase (Cac or Slc25a20) was 1.6-fold upregulated at 12 hours, while carnitine palmitoyltransferase 2 (Cpt2) on the inner mitochondrial membrane was 1.8-fold upregulated at both 12 and 24 hours of fasting. However, the expression of Cpt1, which is present on the outer mitochondrial membrane and is sensitive to malonyl-CoA inhibition, remained unchanged. The 4 acyl-coenzyme A dehydrogenases (Acad -v, -l, -m and -sh), involved in oxidation of very long-, long-, medium- and short-chain fatty acids, were all upregulated in the first 24 hours of fasting (1.5–2.6 fold). The β-subunit of the trifunctional protein (Hadhb) was 2- and 2.1-fold upregulated at 12 and 24 hours, respectively, while another subunit, hydroxyacyl-coenzyme A dehydrogenase, short chain (Hadhsc) showed increased expression after 24 hours of fasting only, indicating altogether a strong stimulation of fatty-acid oxidation at the gene-expression level during the first day of fasting.
The expression of HMGCoA synthase 2 (Hmgcs2
) was also strongly stimulated during the first day of fasting (3.4- and 2.9-fold at 12 and 24 hours, respectively; Figure ), indicating an increased capacity of the synthesis of ketone bodies from acetyl-CoA. This process is further facilitated by increased expression of genes involved in branched-chain keto-acid degradation (Acadm, Hadhsc
; see section on amino-acid catabolism) at 12 and 24 hours. Interestingly, neither the cytoplasmic HMGCoA synthase (Hmgcs1
) nor HMGCoA reductase (Hmgcr
), the key enzyme in de novo cholesterol synthesis pathway [26
], have changed the expression levels in fasted liver.
Among the genes involved in fatty-acid synthesis, enoyl coenzyme A hydratase domain containing 3 (Echdc3) was 1.6 and 1.7-fold downregulated at 12 and 24 hours, while stearoyl-coenzyme A desaturase 1 (Scd1) showed a 2.6-fold decrease in expression at 72 hours of fasting. These data underscore the importance of enhanced lipid catabolism in the liver, which, in the mouse, apparently occurs during the first day of fasting only.
Oxidative stress and unfolded protein response
The enhanced expression of TCA cycle and oxidative-phosphorylation enzymes often causes oxidative stress. Indeed, cytosolic superoxide dismutase (Sod1
) was 2.2-fold upregulated after 24 hours, and the early growth response protein 1 (Egr1
), its transcriptional regulator [27
], 2.9-fold. Furthermore, catalase (Cat
) and stress-regulated mitogen-activated protein kinase 14 (Mapk14
) were both 1.4-fold upregulated at this time point. In addition, metallothionein 1 gene, known to be involved in protection against oxidative stress and metal toxicity [28
], was intensely upregulated (8.6-, 5.5- and 13.5-fold, at 12, 24 and 72 hours, respectively).
Interestingly, the 3 top-scoring processes obtained from a biological-process enrichment analysis all belonged to the unfolded-protein response (endoplasmic reticulum (ER) stress). To present the relevant data in a single figure, we created a network using the shortest-path algorithm (Figure ). The resulting network provides links based on the known interaction data between the nodes from the query data set, and also between the nodes that regulate the given genes or metabolites. It shows 8 heat-shock and 6 other proteins, all upregulated 1.5–2.5 fold, indicating upregulation of this stress-response pathway in fasted liver. Downstream of the ER stress pathway, proteasome degradation was also upregulated, but again only after 24 hours of fasting (Figure ). A list of these and some additional genes regulated in the ER stress and proteasome degradation, with their change level, is shown in Table .
Figure 10 The unfolded-protein response in fasting. The network was generated and linked with available experimental data in the MetaCore™ suite. Nodes with red or blue circles in top right corner of the network objects, represent up- or down-regulation, (more ...)
Proteasome degradation in fasting. The colour code of the GenMAPP view is the same as in Figure 5.
Genes involved in the protein-folding response and oxidative stress that are regulated by fasting