Type 2 diabetes (T2D) and non-alcoholic fatty liver disease (NAFLD) are chronic metabolic diseases that associate with overnutrition. T2D manifests as a defect in blood glucose homeostasis with altered insulin secretion and liver glucose metabolism and insulin resistance( Reference Rizza 1 ). NAFLD comprises a spectrum of changes ranging from raised liver fat (steatosis) to inflammation and fibrosis( Reference Williams, Aspinall and Bellis 2 ), and associates with insulin resistance and T2D( Reference Ortiz-Lopez, Lomonaco and Orsak 3 ). High-carbohydrate diets and particularly dietary sugars are a risk factor for both T2D and NAFLD( Reference Romaguera, Norat and Wark 4 – Reference Moore, Gunn and Fielding 6 ). The transcription factor carbohydrate-response element-binding protein (ChREBP, encoded by MLXIPL), is a major mediator of the hepatic changes induced by high-carbohydrate diets and is conventionally regarded as a key mechanism for promoting conversion of dietary carbohydrate to fat for energy storage( Reference Yamashita, Takenoshita and Sakurai 7 – Reference Oosterveer and Schoonjans 11 ). This review evaluates the hypothesis that a key function of ChREBP is to preserve liver ATP homeostasis in conditions of compromised intracellular metabolite homeostasis as occurs with overnutrition on high-carbohydrate diets. In health, liver cell ATP is robustly homeostatic despite large diurnal changes in glucose metabolism. The hepatic concentrations of intermediates of glucose metabolism comprising phosphate esters (PE) are dynamic but over a narrow concentration range, that has been described as the stability paradox( Reference Hochachka 12 ). Evidence is reviewed that key changes in hepatic gene expression that associate with high-carbohydrate diets or metabolic disease are an adaptive response to maintain intracellular PE and ATP homeostasis.
Effects of type 2 diabetes and non-alcoholic fatty liver disease on hepatic glucose metabolism and de novo lipogenesis
The liver maintains blood glucose homeostasis by net production of glucose in the post-absorptive state and net uptake of glucose in the postprandial state in response to the rise in blood glucose in the portal vein which delivers the products of carbohydrate digestion from the gut. In the post-absorptive state the liver produces glucose by glycogen degradation and gluconeogenesis. The glucose 6-phosphate (G6P) generated in the cytoplasm from these pathways is transported into the endoplasmic reticulum on the G6P-transporter (encoded by SLC37A4) and hydrolysed within the lumen by glucose 6-phosphatase (G6pc) to glucose and inorganic phosphate (Pi)( Reference van Schaftingen and Gerin 13 ). After a meal when the glucose concentration in the portal vein rises, hepatic glucose metabolism undergoes a large metabolic transition from net production to net uptake. The liver extracts about one-third of the glucose load and of this about 60 % is converted to glycogen; the rest is metabolised by glycolysis and the pyruvate is either transported into mitochondria or exported as lactate( Reference Cherrington 14 ). This rapid metabolic transition is mediated by endocrine and neural signals, and by metabolite signals linked to the increase in glucose concentration in the portal vein( Reference Cherrington 14 , Reference Agius 15 ).
T2D is characterised by defects in secretion of insulin and glucagon and by elevated hepatic glucose production, lack of suppression of hepatic glucose production by insulin and lower hepatic glucose uptake( Reference Rizza 1 ). Poorly controlled T2D is associated with low flux through glucokinase( Reference Rizza 1 ) and low glucokinase activity( Reference Caro, Triester and Patel 16 ) and mRNA expression( Reference Haeusler, Camastra and Astiarraga 17 ). T2D is also associated with raised plasma TAG levels and a greater fractional contribution of hepatic de novo lipogenesis (DNL), representing the synthesis of fatty acids from acetyl units( Reference Hellerstein, Schwarz and Neese 18 ), to TAG secretion( Reference Vedala, Wang and Neese 19 ).
In NAFLD, glycerolipid accumulates in hepatocytes because of an imbalance between fatty acid uptake (from NEFA and dietary lipids) plus synthesis (via DNL) relative to fatty acid disposal by mitochondrial β-oxidation plus secretion as TAG in VLDL( Reference Donnelly, Smith and Schwarzenberg 20 – Reference Lambert, Ramos-Roman and Browning 22 ). In the healthy fasted state, the major sources of substrate for hepatic TAG secretion are NEFA derived from adipose lipolysis with a smaller contribution of NEFA from dietary chylomicrons that escape uptake by extra hepatic tissues and a negligible contribution of DNL( Reference Hellerstein, Schwarz and Neese 18 ). NAFLD is associated with increased fatty acid supply by adipose tissue lipolysis( Reference Lambert, Ramos-Roman and Browning 22 ) and an increased fractional contribution of DNL to TAG secretion( Reference Donnelly, Smith and Schwarzenberg 20 – Reference Lambert, Ramos-Roman and Browning 22 ) from <5 % in controls to 15–20 % in patients( Reference Donnelly, Smith and Schwarzenberg 20 , Reference Lambert, Ramos-Roman and Browning 22 ). Both the rate of VLDL–TAG secretion derived from DNL and the fractional contribution of DNL to VLDL secretion are elevated in NAFLD and correlate with liver fat content( Reference Lambert, Ramos-Roman and Browning 22 ). Other hepatic metabolic abnormalities in NAFLD include impaired suppression of hepatic glucose production by insulin( Reference Seppälä-Lindroos, Vehkavaara and Häkkinen 23 ); elevated gluconeogenesis and increased flux through tricarboxylic acid cycle oxidation and anaplerosis, without increased turnover of ketone bodies( Reference Sunny, Parks and Browning 24 ).
A key paradox is that despite a greater fractional contribution of DNL to TAG secretion in NAFLD, the net flux through DNL is very small (<5 %) relative to carbohydrate turnover( Reference Hellerstein 25 ). Accordingly DNL is not a quantitative route for disposal of carbohydrate. Similar considerations apply in relation to ethanol which increases fractional DNL to >30%; however, the major disposal product is acetate released into the circulation rather than DNL which represents <5 % of ethanol metabolism( Reference Hellerstein 25 ). However, DNL has a regulatory role through changes in cellular levels of malonyl-CoA (the product of acetyl-carboxylase) which inhibits mitochondrial fatty acid oxidation( Reference Savage, Choi and Samuel 26 ). This may explain the correlation between raised fractional DNL and raised TAG secretion rate( Reference Diraison, Moulin and Beylot 21 ). DNL may represent a spill over of acetyl-units (as citrate efflux from mitochondria) as suggested by the association between intrahepatic lipid and flux through tricarboxylic acid cycle oxidation plus anaplerosis (and increased gluconeogenic flux), in the absence of elevated ketone body production( Reference Sunny, Parks and Browning 24 ).
Effects of dietary carbohydrate on de novo lipogenesis and hepatic glucose production
The fractional contribution of hepatic DNL to hepatic TAG secretion is dependent on dietary carbohydrate intake( Reference Parks 27 ). Healthy subjects consuming diets with lower (25–50 %) or higher (25–50 %) carbohydrate content for 5 d showed a near-linear relation between fractional DNL and dietary carbohydrate( Reference Schwarz, Neese and Turner 28 ). Raised fractional DNL with high carbohydrate was also observed in isoenergetic conditions( Reference Hudgins, Hellerstein and Seidman 29 ) and more so with dietary sugars than with starch( Reference Hudgins, Seidman and Diakun 30 ). Fractional DNL correlates with raised VLDL–TAG secretion rates and raised plasma TAG levels( Reference Hudgins, Hellerstein and Seidman 31 ). Fructose has a greater effect than glucose on DNL, plasma lipid levels and hepatic steatosis( Reference Parks, Skokan and Timlin 32 – Reference Sevastianova, Santos and Kotronen 34 ). Fructose increases DNL more strongly and more rapidly (within hours) than glucose when administered acutely( Reference Hellerstein, Schwarz and Neese 18 , Reference Hellerstein 25 ). This rapid effect of fructose is attributed to its high flux of metabolism, whereas the effect of chronic consumption of glucose-containing high-carbohydrate diets is thought to be predominantly transcriptional. Despite the large fractional increase in DNL (from 1 to about 30 %) caused by either acute fructose infusion or chronic carbohydrate surplus( Reference Schwarz, Neese and Turner 28 , Reference Lammert, Grunnet and Faber 35 ) the net daily flux through DNL in comparison with dietary carbohydrate load is small (⩽5 %) in both man and rodents( Reference Hellerstein 25 , Reference Lammert, Grunnet and Faber 35 ) indicating that even in healthy subjects on high-carbohydrate diets the role of DNL is regulatory for fuel selection for oxidation( Reference Savage, Choi and Samuel 26 , Reference Schwarz, Neese and Turner 28 ) and not as a quantitative route of disposal of dietary carbohydrate( Reference Hellerstein 25 ).
Another effect of high-carbohydrate diets in man is an increase in hepatic glucose production. During chronic carbohydrate under-feeding or over-feeding (for 5 d) hepatic glucose production increased in parallel with DNL with increasing carbohydrate load despite the elevated plasma insulin( Reference Schwarz, Neese and Turner 28 ). Elevation in basal glucose production also occurred with a high-carbohydrate diet in isoenergetic conditions( Reference Bisschop, de Metz and Ackermans 36 ). The parallel elevation in glucose production and DNL with high-carbohydrate diets are similar to the metabolic characteristics of NAFLD( Reference Sunny, Parks and Browning 24 ).
Effects of dietary carbohydrate on hepatic gene expression: role of carbohydrate-response element-binding protein
In animal models high-carbohydrate diets induce liver enzymes of DNL( Reference Uyeda and Repa 8 , Reference Towle 9 ). Two transcription factors involved are sterol regulatory element binding protein 1c that is induced by insulin and targets predominantly enzymes of DNL and ChREBP that is induced by high glucose and targets enzymes of DNL but also several additional genes, including pyruvate kinase ( Reference Yamashita, Takenoshita and Sakurai 7 – Reference Towle 9 ). Because ChREBP is induced by dietary carbohydrate, the generally held explanation for ChREBP function is that it represents a mechanism for efficient conversion of dietary carbohydrate to fat as a ‘thrifty gene’( Reference Uyeda and Repa 8 – Reference Oosterveer and Schoonjans 11 ). The decrease in DNL( Reference Iizuka, Bruick and Liang 37 – Reference Dentin, Benhamed and Hainault 39 ) and in steatosis( Reference Iizuka, Miller and Uyeda 38 , Reference Dentin, Benhamed and Hainault 39 ) in ChREBP knock-down models( Reference Iizuka, Bruick and Liang 37 ) is consistent with such a hypothesis but it does not exclude the alternative explanation( Reference Hellerstein 25 ) that the role of DNL is regulatory for fuel selection because it is not a quantitative route of carbohydrate disposal( Reference Hellerstein 25 , Reference Schwarz, Neese and Turner 28 ).
ChREBP binds to the promoter of its target genes in complex with Max-like protein X (Mlx), a member of the Myc/Max superfamily of transcription factors( Reference Towle 9 , Reference Peterson and Ayer 40 ). Mlx has other partners, including MondoA, the paralogue of ChREBP( Reference Peterson and Ayer 40 ). Gene microarrays in hepatocytes using a dominant-negative Mlx, identified several candidate target genes of Mlx, including not only enzymes of DNL and L-pyruvate kinase (PKLR), but also two enzymes of gluconeogenesis (G6pc and fructose bisphosphatase-1) which catalyse generation of Pi from PE and the inhibitory protein of glucokinase encoded by the Gckr gene( Reference Ma, Robinson and Towle 41 ). These target genes were also confirmed by chromatin immunoprecipitation-PCR( Reference Arden, Petrie and Tudhope 42 , Reference Arden, Tudhope and Petrie 43 ) or chromatin immunoprecipitation-sequencing( Reference Poungvarin, Chang and Imamura 44 ). Glucokinase which is the major determinant of hepatic glucose disposal( Reference Agius 15 ), is not a target gene for ChREBP and it is repressed rather than induced by high glucose( Reference Arden, Petrie and Tudhope 42 ). The induction by high glucose via ChREBP-Mlx of the two major counter-regulators of glucokinase, the glucokinase inhibitor protein (Gckr) and G6pc which dephosphorylates the product of the glucokinase reaction, argues against a role for ChREBP-Mlx in net conversion of glucose to fat.
To evaluate the hypothesis that the function of ChREBP is to maintain homeostasis of PE and ATP in conditions of high-carbohydrate diets or compromised intracellular metabolite homeostasis, the following issues need to be considered: (1) To what extent are the hepatic concentrations of PE of glucose metabolism elevated postprandially? What is the function of these changes? (2) What are the consequences of disrupted regulation on cell Pi and ATP homeostasis? (3) What mechanisms exist to restrain the elevation in PE within physiological limits? (4) What is the effect of ChREBP deletion on hepatic PE and ATP homeostasis? (5) What are the signals for ChREBP activation?
Hepatic concentrations of phosphate esters in the fasted-to-fed transition
In the rodent model, the hepatic concentration of G6P, the first intermediate of glucose metabolism (Fig. 1) and also downstream PE intermediates of glycolysis and the pentose phosphate pathway are elevated about 2-fold in the fed state on a starch-containing diet compared with the fasted state( Reference Casazza and Veech 45 ). The increase in G6P is sustained during the absorptive state( Reference Fernández-Novell, Roca and Bellido 46 ). Larger elevations in PE, particularly of the pentose pathway intermediates, occur on a sucrose-containing diet( Reference Casazza and Veech 45 ). Changes in G6P are associated with parallel changes in fructose 6-phosphate (F6P) which is in equilibrium with G6P via phosphoglucoisomerase, and with elevation in fructose 2,6-bisphosphate (F2,6P2) which is a ‘regulatory’ rather than ‘intermediate’ metabolite that is a potent activator of phosphofructokinase-1 and thereby glycolysis( Reference Hue and Rider 47 ). F2,6P2 is generated from F6P by the bifunctional enzyme phosphofructokinase-2/fructose biphosphatase-2, which also degrades F2,6P2 to F6P. The postprandial elevation in G6P has a major role in stimulating glycogen storage (Fig. 1) because G6P is an allosteric effector of glycogen phosphorylase and glycogen synthase( Reference Agius 15 ). The elevation in F2,6P2 mediates stimulation of glycolysis by activation of phosphofructokinase-1, with consequent elevation in its product fructose 1,6-bisphosphate, which is an allosteric activator of pyruvate kinase which converts phosphoenolpyruvate, the last PE intermediate of glycolysis to pyruvate. The greater elevation in PE caused by a sucrose-diet compared with a starch-diet( Reference Casazza and Veech 45 ) can be explained by the different pathways of glucose (Fig. 1) and fructose (Fig. 2) metabolism in liver as discussed later.
Fig. 1. Glucose stimulates glucose disposal by allosteric mechanisms but it also causes feed-back regulation of gene expression. Elevated glucose concentration in the portal vein causes rapid dissociation of glucokinase (GK) from the Gckr regulatory protein (RP) resulting in a moderate increase in the cellular concentrations of glucose 6-P (G6P) and fructose 2,6-bisphosphate (F2,6P2) which cause allosteric activation of enzymes promoting glycogen storage and glycolysis. Chronic elevation in G6P and F2,6P2 causes activation of transcription factors which down regulate the GK gene (Gck) and upregulate the G6pc and Pklr genes. This gene regulation serves to maintain homeostasis of the cell phosphometabolite pool and results in restraint of glucose uptake and elevated glucose production.
Fig. 2. Fructose is transported and phosphorylated by a distinct mechanism from glucose. Low concentrations of fructose cause moderate elevation in fructose 1-phosphate (F1P), which dissociates glucokinase (GK) from its regulatory protein (RP) leading to stimulation of glucose metabolism (1). Millimolar concentrations of fructose cause marked elevation in F1P (2) because the activity of ketohexokinase (KHK) is higher than that of downstream enzymes. During reconversion of ADP to ATP by oxidative phosphorylation (OP; 3), the cell content of inorganic phosphate (Pi) becomes depleted (4) resulting in depletion of ATP (5) and degradation of AMP (6) to uric acid (7).
Effects of glucose on hepatocyte phosphate esters:the restraining role of glucokinase inhibitor protein and glucose 6-phosphatase
In hepatocytes, the G6P content is a sigmoidal function of glucose concentration (5–35 mm) with half-maximal (S0·5) stimulation at about 25 mm glucose as for glucose phosphorylation( Reference Agius 15 , Reference van Schaftingen, Veiga-da-Cunha and Niculescu 48 ). This represents a lower affinity than for glucokinase kinetics (S0·5 8 mm) and is explained by the inhibition of glucokinase by the Gckr protein( Reference van Schaftingen, Veiga-da-Cunha and Niculescu 48 ). Gckr sequesters glucokinase in the nucleus of hepatocytes in the fasted state, and enables its translocation to the cytoplasm in response to elevated glucose in the portal vein postprandially (Fig. 1)( Reference Agius 15 ). Allosteric effectors that enhance dissociation of glucokinase from Gckr (e.g. fructose 1-phosphate (F1P) and Pi) promote translocation of glucokinase to the cytoplasm and elevation in G6P( Reference Davies, Detheux and Van Schaftingen 49 – Reference Agius 51 ). Accordingly the G6P elevation like glucose phosphorylation( Reference de la Iglesia, Mukhtar and Seoane 50 , Reference Agius 51 ) is a function of the glucokinase: Gckr protein ratio and of allosteric ligands of Gckr (e.g. F1P and Pi) which decrease its affinity for glucokinase( Reference Davies, Detheux and Van Schaftingen 49 – Reference Agius 51 ). The activities of enzymes that degrade G6P (e.g. G6pc) also influence the elevation of G6P in response to glucose, as shown by overexpression of G6pc which lowers the G6P elevation by high glucose( Reference Seoane, Trinh and O'Doherty 52 ) and by inhibitors of G6P hydrolysis, such as chlorogenic derivatives that target the G6P transporter, which markedly enhance the elevation in G6P( Reference Arden, Petrie and Tudhope 42 , Reference van Dijk, van der Sluijs and Wiegman 53 ). The induction by high glucose via Mlx-ChREBP of G6pc and Gckr( Reference Ma, Robinson and Towle 41 ) therefore has a restraining effect on the elevation in G6P( Reference Arden, Petrie and Tudhope 42 , Reference Arden, Tudhope and Petrie 43 ).
Effects of fructose on hepatic phosphate ester, inorganic phosphate, ATP and uric acid production
The larger elevation in hepatic PE on a diet containing sucrose or fructose( Reference Casazza and Veech 45 ) can be explained by the high hepatic capacity for phosphorylation of fructose to F1P (by ketohexokinase (KHK)) and the absence of a mechanism for hydrolysis of F1P, the first intermediate (Fig. 2). Fructose transport into hepatocytes is characterised by a high K m (50–100 mm) and V max (Reference Sestoft and Fleron54) and may involve various transporters including Glut5, Glut2 and Glut8( Reference Karim, Adams and Lalor 55 ). The KHK gene encodes two isoforms: a ubiquitous isoform with very low-affinity (KHKa) and the liver isoform (KHKc) which is expressed at high activity in human and rat liver and has a high-affinity (K m approximately 0·5 mm) for fructose( Reference Hayward and Bonthron 56 – Reference Masson, Henriksen and Stengaard 58 ). Accordingly, the liver is the major site of fructose clearance( Reference Van den Berghe 57 ). After phosphorylation of fructose to F1P, the latter is metabolised by aldolase-2 to dihydroxyacetone-P and glyceraldehyde, which is phosphorylated by triokinase to glyceraldehyde 3-P. The triose phosphates generated are metabolised by glycolysis and gluconeogenesis. Two key consequences of the high KHKc capacity are: (i) a higher rate of metabolic flux from fructose to pyruvate compared with glucose to pyruvate; and (ii) a much larger elevation of the cellular PE concentrations than occurs with high glucose. This is explained by the high capacity of KHKc relative to the capacity for clearance of triose phosphates by glycolysis and gluconeogenesis( Reference Van den Berghe 57 ) and accordingly hepatic F1P can rise to very high levels by micromolar or millimolar fructose reaching the liver in the portal vein( Reference Davies, Detheux and Van Schaftingen 49 , Reference Masson, Henriksen and Stengaard 58 ). A further mechanism for the elevation in PE with fructose is that the elevation in F1P dissociates glucokinase from Gckr (Fig. 2) and elevates G6P derived from glucose by translocating glucokinase to the cytoplasm( Reference van Schaftingen, Veiga-da-Cunha and Niculescu 48 – Reference Agius 51 ). This mechanism operates also at small fructose loads (20 mg/kg) with modest effects on F1P( Reference Van Schaftingen and Davies 59 ) and is explained by the high affinity of Gckr for F1P( Reference van Schaftingen, Veiga-da-Cunha and Niculescu 48 , Reference Davies, Detheux and Van Schaftingen 49 ).
In man the fructose concentrations in the portal vein can reach 2·5 mm ( Reference Holdsworth and Dawson 60 ). This is expected to raise hepatic F1P to millimolar concentrations( Reference Davies, Detheux and Van Schaftingen 49 ). Studies testing the effects of fructose on hepatic Pi and ATP levels have generally used intravenous fructose loads ranging from 0·5 to 1·5 g/kg body weight given as a bolus load or infusion( Reference Masson, Henriksen and Stengaard 58 ). MRI for real-time monitoring of hepatic PE, Pi and ATP showed that fructose infusion causes acute elevation of PE and depletion of Pi (about 50 %) and ATP (by 24–29 %)( Reference Masson, Henriksen and Stengaard 58 ), with similar effects in human and rodent liver( Reference Masson, Henriksen and Stengaard 58 ). During F1P accumulation the ADP generated is used together with Pi for oxidative phosphorylation leading to depletion of hepatic Pi and thereby attenuation of oxidative phosphorylation and depletion of ATP (Fig. 2) The depletion of Pi and ATP results in degradation of AMP to inosine monophosphate via AMP-deaminase and the further metabolism of inosine monophosphate to uric acid( Reference Mayes 61 , Reference Petrie, Patman and Sinha 62 ).
In man and in higher primates, uric acid produced from degradation of AMP and purines is not further metabolised to allantoin because of lack of a functional uric acid oxidase gene( Reference Oda, Satta and Takenaka 63 ). Accordingly plasma uric acid levels are higher in man than in most other mammals that convert uric acid to allantoin. Uric acid production by isolated hepatocytes is a very sensitive index of compromised cell ATP homeostasis in response to low concentrations of fructose( Reference Petrie, Patman and Sinha 62 ). However, changes in blood uric acid concentration in man after fructose consumption are not a sensitive index of elevated hepatic uric acid production, because the basal rate of production of uric acid by human liver (3–5 nmol/min per g liver)( Reference Grunst, Dietze and Wicklmayr 64 ) is low relative to the plasma pool size (1·5 mmol). Therefore, a 2- to 3-fold increase in uric acid production, as occurs at 2 mm fructose( Reference Petrie, Patman and Sinha 62 ), may not be detectable in the peripheral circulation. However, evidence for a several-fold increase in uric acid production in man after fructose challenge (0·5 g/kg per h) was evident by catheterisation of the hepatic vein( Reference Grunst, Dietze and Wicklmayr 64 ) or by 14C-adenine labelling and measurement of 14C-excretory products in the urine( Reference Johnson, Tekkanat and Schmaltz 65 ).
Hepatic ATP homeostasis: robust in response to glucose but not fructose
Robustness of liver function is contingent on maintenance of ATP homeostasis because ATP is required for anabolic pathways such as protein synthesis, urea synthesis, gluconeogenesis and other ATP-dependent processes( Reference Brown 66 ). Pathways of hepatic glucose and fructose metabolism begin with ATP-requiring steps, generating G6P or F1P, respectively. ATP homeostasis is maintained if the ATP-production rate by oxidative phosphorylation matches ATP consumption. Oxidative phosphorylation is dependent on the trans-mitochondrial proton gradient generated by the electron transport chain and on the concentrations of the substrates ADP and Pi. It is therefore dependent on cell Pi homeostasis. The hepatocellular total phosphate content (PE + Pi) is about 50 µmol/g of which about 10 % is Pi, representing cytoplasmic and mitochondrial pools which are substrates for glycogenolysis/glycolysis and oxidative phosphorylation, respectively( Reference Sestoft and Kristensen 67 ). Pi transport capacity from the cytoplasm into mitochondria exceeds the rate of ATP synthesis( Reference Brown 66 ). However, Pi transport across the cell membrane (influx and efflux) is slow relative to oxidative phosphorylation capacity( Reference Sestoft and Kristensen 67 ). In steady state homeostasis of ATP and Pi is preserved because ATP consumption is balanced by ATP production and Pi incorporation is balanced by Pi regeneration.
Hepatic glucose disposal is a good example of metabolic robustness because high glucose concentrations (15–25 mm) as occur in the portal vein or high rates of glycogenolysis do not compromise ATP homeostasis, because G6P and downstream metabolites are modestly elevated( Reference Casazza and Veech 45 ). However, fructose metabolism is an example of fragility because 2 mm fructose causes excessive accumulation of F1P( Reference Davies, Detheux and Van Schaftingen 49 ) with consequent depletion of Pi and ATP and degradation of AMP to uric acid( Reference Petrie, Patman and Sinha 62 ). Two key components of the robustness of glucose metabolism are the glucose 6-phosphohydrolase system within the endoplasmic reticulum (comprising the G6P transporter and G6pc) that dephosphorylates G6P with a high V max and K m ( Reference van Schaftingen and Gerin 13 ), and the glucokinase inhibitor, Gckr. Patients with glycogen storage disease type 1, who have gene defects in G6pc or the G6P transporter, respond to glucagon challenge by acute depletion of Pi and ATP and elevated uric acid production( Reference Greene, Wilson and Hefferan 68 ), because Pi incorporation into PE during activation of glycogen phosphorylase is not balanced by Pi regeneration by G6pc. The induction of G6pc and the G6P transporter (SLC37A4) by high glucose can be rationalised as an adaptive mechanism to maintain Pi and ATP homeostasis( Reference Ma, Robinson and Towle 41 , Reference Arden, Petrie and Tudhope 42 ). The induction of Gckr (and the decrease in glucokinase:Gckr ratio) by glucose can also be rationalised as an adaptive response for ATP homeostasis, as shown by hepatocyte studies where an experimentally induced elevation in the glucokinase:Gckr ratio leads to compromised ATP homeostasis and increased uric acid production in response to high glucose challenge( Reference Arden, Petrie and Tudhope 42 ).
Carbohydrate-response element-binding protein and liver ATP homeostasis: evidence from carbohydrate-response element-binding protein-deletion models and human gene variants
If the function of ChREBP is to maintain hepatic ATP homeostasis by controlling gene expression of enzymes or regulatory proteins that influence PE concentrations (e.g. G6pc, fructose biphosphatase-1, Gckr, Pklr), then ChREBP deletion models would be expected to be associated with raised PE levels, compromised ATP homeostasis and elevated uric acid production. Consistent with this hypothesis, ChREBP deletion models have 2- to 3-fold higher levels of hepatic G6P, dihydroxyacetone-P, 3-phosphoglycerate and phoshoenolpyruvate on either chow or starch diets( Reference Iizuka, Bruick and Liang 37 ) with somewhat greater differences (about 5-fold) in the hyperphagic ob/ob mouse( Reference Iizuka, Miller and Uyeda 38 , Reference Dentin, Benhamed and Hainault 39 ). A study by Burgess and colleagues( Reference Burgess, Iizuka and Jeoung 69 ) that determined the hepatic phosphorylation potential, representing the ratio of free concentrations of [ATP]/[ADP]×[Pi] in ChREBP−/− mice showed a 50 % lower phosphorylation potential associated with a 24 % decrease in total ATP and 4-fold elevation of free AMP. The ChREBP−/− mice were viable on chow or starch diets but not on sucrose or fructose diets( Reference Iizuka, Bruick and Liang 37 , Reference Burgess, Iizuka and Jeoung 69 ). Two possible explanations for the mortality on fructose are either increased expression of fructose transporters and/or KHKc leading to greater F1P elevation and more severe ATP depletion; or a defect in ATP recovery. The first possibility seems unlikely because Khk mRNA levels were lower in the ChREBP−/− model( Reference Iizuka, Bruick and Liang 37 ). In support of the second possibility, gene microarrays in cells with ChREBP knock-down identified dihydrofolate reductase which is involved in purine biosynthesis in the down-regulated genes( Reference Tong, Zhao and Mancuso 70 ), and genes involved in purine metabolism were also identified as ChREBP targets by chromatin immunoprecipitation-sequencing( Reference Poungvarin, Chang and Imamura 44 ). A role for ChREBP in ATP homeostasis is also indicated by raised plasma uric acid levels in rats with partial ChREBP knock-down on a high-fructose diet( Reference Erion, Popov and Hsiao 71 ).
Plasma levels of uric acid in man are elevated in association with insulin resistance( Reference Vuorinen-Markkola and Yki-Järvinen 72 ) and NAFLD( Reference Li, Xu and Yu 73 ) and increased hepatic production of uric acid is inferred in these conditions( Reference Vuorinen-Markkola and Yki-Järvinen 72 ). Genome-wide studies in man have identified twenty-eight loci in association with raised uric acid levels( Reference Köttgen, Albrecht and Teumer 74 ). They include variants for MLXIPL and two of the ChREBP target genes, PKLR and GCKR. The association of these genes with raised uric acid, a marker of ATP depletion( Reference Petrie, Patman and Sinha 62 ) is consistent with a role for ChREBP, Pklr and GCKR in ATP homeostasis.
Mechanism for glucose-induced activation of carbohydrate-response element-binding protein: glucose 6-phosphate or multiple phosphate esters?
The mechanism by which high glucose activates ChREBP has not been resolved. ChREBP is transcribed as either a full-length ChREBP-α (864 residues), which is cytoplasmic at low glucose and activated by high glucose which causes translocation to the nucleus, recruitment to target genes and transcriptional activation, or as a shorter ChREBP-β isoform (lacking the first 117-residues at the N-terminus), that is constitutively active at low glucose( Reference Herman, Peroni and Villoria 75 ). Several aspects of the post-transcriptional regulation of ChREBP-α have been explored, including covalent modification by phosphorylation( Reference Uyeda and Repa 8 , Reference Sakiyama, Wynn and Lee 76 – Reference Tsatsos, Davies and O'Callaghan 78 ), O-GlcNAc-modification( Reference Guinez, Filhoulaud and Rayah-Benhamed 79 , Reference Sakiyama, Fujiwara and Noguchi 80 ) and acetylation( Reference Bricambert, Miranda and Benhamed 81 ); mutational studies for mapping of the N-terminal glucose sensory domain and 14-3-3 binding sites( Reference Li, Chang and Imamura 82 – Reference Ge, Huang and Wynn 84 ), and identification of putative metabolites that mediate the response to high glucose. The proposed metabolites include: G6P( Reference Li, Chen and Harmancey 85 , Reference Dentin, Tomas-Cobos and Foufelle 86 ), xylulose 5-P( Reference Uyeda and Repa 8 ), the regulatory metabolite F2,6P2 ( Reference Arden, Tudhope and Petrie 43 , Reference Petrie, Al-Oanzi and Arden 87 ), ketone bodies( Reference Nakagawa, Ge and Pawlosky 88 ) and the product of the hexosamine pathway UDP-N-acetylglucosamine( Reference Guinez, Filhoulaud and Rayah-Benhamed 79 , Reference Sakiyama, Fujiwara and Noguchi 80 ) which is the substrate for O-GlcNAc-modification of proteins. The induction of ChREBP target genes by high glucose is abolished by glucokinase inhibitors and markedly enhanced by inhibitors of G6P hydrolysis( Reference Arden, Petrie and Tudhope 42 , Reference Parks, Skokan and Timlin 32 , Reference van Dijk, van der Sluijs and Wiegman 53 ). These inhibitors abolish and enhance, respectively, the elevation in PE pool in hepatocytes comprising G6P and downstream metabolites( Reference Arden, Petrie and Tudhope 42 , Reference Parks, Skokan and Timlin 32 , Reference van Dijk, van der Sluijs and Wiegman 53 ). This rules out an effect of glucose and points to mechanism(s) consequent to PE elevation. Current evidence suggests that a combination of mechanisms is most likely involved in ChREBP-α activation by glucose. For example mutation of residues that are substrates for phosphorylation by cyclic AMP dependent protein kinase and AMP-dependent protein kinase (S196, S626, T666; S566) and for dephosphorylation by glucose( Reference Uyeda and Repa 8 ), did not generate a glucose-insensitive ChREBP indicating involvement of additional mechanisms( Reference Tsatsos and Towle 77 ). Covalent modification of ChREBP by O-GlcNAc which is dependent on UDP-N-acetylglucosamine generated by the hexosamine pathway caused ChREBP protein stabilisation and transcriptional activation( Reference Tsatsos and Towle 77 , Reference Tsatsos, Davies and O'Callaghan 78 ). However, increased flux through the hexosamine pathway at low glucose does not activate ChREBP target genes( Reference Arden, Tudhope and Petrie 43 ). A role for G6P as the signalling metabolite for ChREBP activation( Reference Li, Chen and Harmancey 85 , Reference Dentin, Tomas-Cobos and Foufelle 86 ) was proposed based on two sets of evidence: (i) a glucose-like effect of 2-deoxyglucose which is phosphorylated to its 6-PE but not further metabolised; (ii) attenuation of the high glucose effect by overexpression of enzymes that metabolise G6P (e.g. G6pc, glucose 6-phosphate dehydrogenase, phosphofructokinase-1, phosphofructokinase-2). However, 2-deoxyglucose does not activate ChREBP target genes in hepatocytes despite accumulation of its 6-PE( Reference Arden, Tudhope and Petrie 43 ). Attenuation of the glucose response by overexpression of enzymes that degrade G6P does not rule out hexosamine pathway involvement, because G6P is in equilibrium with F6P, the substrate for the hexosamine pathway. A role for F2,6P2 was proposed based on inhibition of recruitment of ChREBP-Mlx when this metabolite was selectively depleted at high glucose with a kinase-deficient bisphosphatase-active variant of phosphofructokinase-2/fructose bisphosphatase-2. However, additive activation of ChREBP target genes by glucose and xylitol which do not have additive effects on F2,6P2 indicates involvement of additional metabolites or mechanisms( Reference Petrie, Al-Oanzi and Arden 87 ).
On the premise that G6P activates MondoA( Reference Peterson and Ayer 40 ) and ChREBP( Reference Li, Chen and Harmancey 85 , Reference Dentin, Tomas-Cobos and Foufelle 86 ), McFerrin and Atchley( Reference McFerrin and Atchley 89 ) conducted a bioinformatic analysis for the putative ‘G6P binding motif’ within the N-terminal conserved ‘glucose sensing module’ based on G6P-binding sites of glucokinase and other enzymes, and identified a ‘low-specificity’ motif (Sx[ST]xx[ST]) similar to that found in glucosephosphate isomerase and glutamine:fructose 6-phosphate amidotransferase( Reference McFerrin and Atchley 89 ). Based on the ‘low-specificity’ of this motif and that F6P rather than G6P is the common substrate of these two enzymes, this outcome favours F6P or multiple PE.
Conclusion
ChREBP is induced and activated by high-carbohydrate diets( Reference Yamashita, Takenoshita and Sakurai 7 , Reference Uyeda and Repa 8 ) and more so by a fructose-containing diet( Reference Koo, Miyashita and Cho 90 ). The mechanism by which ChREBP-α is activated remains unresolved. Current evidence suggests a complex mechanism involving phosphorylation and O-GlcNAc modification of multiple residues and multiple PE including F2,6P2. The target genes of ChREBP include enzymes of DNL and several proteins which have a lowering effect on PE intermediates of glucose metabolism such as G6P (GCKR, G6PC) and phosphoenolpyruvate (Pklr). These target genes support the hypothesis that a key function of ChREBP is to maintain homeostasis of PE and thereby ATP in conditions of compromised cell PE homeostasis. This hypothesis is further supported by: (i) compromised ATP phosphorylation potential and raised hepatic PE in ChREBP knock-down mouse models; (ii) raised plasma uric acid levels in association with common human variants in the ChREBP gene (MLXIPL) and two of its target genes (GCKR and PKLR)( Reference Köttgen, Albrecht and Teumer 74 ). The hepatic changes that associate with T2D and NAFLD share several similarities to the changes induced by high-carbohydrate diets. It is proposed that these changes in metabolic disease may likewise be, at least in part, an adaptive response to preserve ATP homeostasis. Recognition of this mechanism is crucial for the design of therapeutic strategies for metabolic disease with sustained efficacy( Reference Agius 91 ).
Acknowledgements
The author thanks Tabassum Moonira for reading the manuscript, and all past and previous colleagues in the laboratory who contributed to the work reviewed. Research in the author's laboratory is funded by Diabetes UK (BDA 11/0004231).
Financial support
None.
Conflicts of Interest
None.
Authorship
The author was responsible for all aspects of preparation of this paper.
