Influence of plasma macronutrient levels on hepatic metabolism: role of regulatory networks in homeostasis and disease states

Abstract

The human liver acts as a homeostatic controller for maintaining the normal levels of plasma metabolite concentrations by uptake, utilization, storage and synthesis of essential metabolites. These hepatic functions are orchestrated through a multilevel regulation composed of metabolic, signaling and transcriptional networks. Plasma macronutrients namely, glucose, amino acids and fatty acids are known to influence these regulatory mechanisms to facilitate homeostasis. We composed a regulatory circuit that elicits the design principle behind the metabolic regulation in liver. We have developed a detailed dynamic model for hepatic metabolism incorporating the regulatory mechanisms at signaling and transcriptional level. The model was analyzed to capture the behavior of hepatic metabolic fluxes under various combinations of plasma macronutrient levels. The model was used to rationalize and explain the experimental observations of metabolic dysfunctions through regulatory mechanisms. We addressed the key questions such as, how high carbohydrate diet increases cholesterol and why a high protein diet would reduce it; how high fat and high protein diet increases gluconeogenesis leading to hyperglycemia; how TCA (tricarboxylic acid) cycle is impaired through diet induced insulin resistance; how high fat can impair plasma ammonia balance; how high plasma glucose can lead to dyslipidemia and fatty liver disease etc. The analysis indicates that higher levels (above 2.5–3 fold) of macronutrient in plasma results in impairment of metabolic functions due to perturbations in the regulatory circuit. While higher glucose levels saturate the rate of plasma glucose uptake, higher amino acids activate glucagon and inhibit IRS (insulin receptor substrate) through S6K (S6 kinase), whereas higher fatty acid levels inhibit IRS through DAG–PKC (diacylglycerol and protein kinase C) and TRB3 activation. Moreover the ATP–ADP ratio is reduced under such conditions and β-oxidation is up-regulated through activation of PPARα (peroxisome proliferator-activated receptor alpha) leading to reduced anabolic capacity and increased cataplerosis in TCA cycle. The above factors together decrease insulin sensitivity and enhances glucagon effect through underlying signaling and transcriptional network leading to insulin resistance in liver. Such a metabolic state is known to result in diabetes and non-alcoholic fatty liver disease.

---

Introduction

The plasma homeostasis of most of the vital metabolites is maintained by the intervention of hepatic metabolism.^1,2 The versatility of the central metabolic pathway in the liver enables it to interconvert the metabolites and maintain the hepatic energy supply when required.^3,4 Due to the non-linear nature of the effect of glucose, amino acids and fatty acids on insulin and glucagon secretions and subsequent signaling pathway, it is difficult to predict the metabolic changes that can be induced through different macronutrient compositions in diet. It is long known that dietary and behavioral patterns of individuals are responsible for lifestyle diseases and the underlying changes in the metabolic status.⁵ Several experimental investigations over the last two decades have reported the effect of variation in dietary composition on hepatic metabolism in mice, rats, hamsters and humans.^6–14 However, such studies do not provide a mechanistic explanation for the phenotypic observations such as, how a high fat and high protein diet increases gluconeogenesis and hyperglycemia; how a high carbohydrate diet increases cholesterol levels; how a high protein–low fat diet can reduce cholesterol synthesis; how high fat diet induce defects in TCA flux leading to an insulin resistance state; how a high fat diet increases plasma ammonia levels; how high plasma fat and protein levels can affect hepatic glucose release leading to hyperglycemia; how high glucose levels can affect hepatic fatty acid uptake and lead to dyslipidemia and NAFLD (non-alcoholic fatty lever disease) etc.

Our aim of this study was to develop a mechanistic model to answer these questions in a regulatory perspective. To analyze these effects we developed a mathematical model incorporating the regulatory circuit in the hepatic metabolism. Unlike the other models in literature, this is a first effort in literature to integrate the regulatory circuit comprising of signaling and transcriptional network with metabolic network. This would enable to rationalize the phenotypic responses and associated disease states through a regulatory perspective.

The developed model was used to obtain steady state fluxes for various metabolic reactions in response to variation in plasma metabolite levels. The analysis reveals that extremely high levels of fatty acids and amino acids can reduce insulin sensitivity compromising the anabolic capacity of insulin and consequently leading to a metabolic state that represents insulin resistance. Certain combinations in the levels of macronutrients would result in metabolic fluxes that represent a diabetic state wherein the hepatic glucose release, gluconeogenesis and lipolysis are active even under high insulin levels.¹⁵ Conversely, under low glucose conditions (higher physical activity and exercise) where a catabolic state is anticipated, with increasing circulating levels of fatty acids and triglycerides reduces the catabolic capacity. Whereas, higher amino acids would help in increasing the overall catabolic rate and facilitate higher rate of glucose release. Thus, the study highlights the metabolic states attained due to various levels of macronutrients in plasma and subsequent complexity in the regulation that leads to disease states.

Regulatory circuit

Apart from being used as metabolic substrates, plasma macronutrients (glucose, amino acids and lipids) act as global regulators of metabolic pathways.^16,17 The regulatory actions are mediated by hormones that are triggered by sensing these metabolites through pancreas. The plasma levels of these macronutrients are known to influence the secretion of hormones namely, insulin and glucagon.^18–20 The hormones (insulin and glucagon) further activate specific signaling pathways that eventually influence the activity of the downstream enzymes that catalyze metabolism. These macronutrients also regulate the signaling components and transcriptional factors that regulate gene expression mediated by insulin and glucagon in a highly nonlinear manner.^21–23 These effects of macronutrients on hormonal secretion, signaling pathways and transcriptional factors result in a metabolic regulatory pattern that varies with different levels of macronutrients present in the plasma. The interactions and crosstalk between the signaling, transcription and metabolic pathways (as reported in several bits and parts in literature) were used to develop a comprehensive regulatory circuit. While glucose enhances insulin secretion and reduces glucagon secretion, amino acids have a tendency to enhance the secretion of both insulin and glucagon at different thresholds.^24–26 Fatty acids and triglycerides also influence insulin secretion.²⁷ Although fatty acids increase insulin secretion, it inhibits insulin signaling beyond a certain threshold.^28,29 The regulatory mechanisms of these macronutrients at multiple levels results in a highly non-linear metabolic flux landscape based upon the different quantities of these macronutrients in the diet. This leads to a complex interplay of metabolites, signaling proteins and gene expression that decides the cellular metabolic state. The schematic of the interactions between macronutrients, hormones and signaling is shown in Fig. 1. The detailed molecular network of the regulatory pathways is depicted in S1 Fig. 1*.†

Fig. 1 The schematic of the interactions between plasma macronutrients (glucose, amino acids and fatty acids) on pancreatic hormones (insulin and glucagon) and metabolic regulatory signaling pathways.

Results

Model development

In order to provide an explanation to the experimental observations, a detailed model was developed incorporating hepatic metabolism (see Fig. 2) and its regulation through hormonal signaling and transcriptional network as shown in Fig. 1. In our study, we mainly concentrate on the effects of plasma glucose, amino acids and fatty acid concentration on the hepatic metabolism and explain several observed phenotypic responses to different dietary conditions from a regulatory perspective. Moreover, this is the first time in literature, that we have integrated the signaling and transcriptional effects with metabolism and have analyzed the effect of plasma macronutrient concentration on the tissue metabolism. However, it should be noted that there are several models that specifically model signaling or the metabolic pathways independently.^30–41 We have integrated these several published models as components of our comprehensive model along with a module for whole body plasma metabolite homeostasis. We applied a system level approach that is composed of four modules such as blood, metabolism, signaling and transcription. The modeling framework involved representation of biological pathways by mathematical functions given by mass action, Michaelis–Menten and Hills kinetic functions. A mass balance was performed on the network to obtain the ordinary differential equations to capture the dynamics of each state variable in the system. The overall model is composed of 272 rate equations, 170 state variables and 801 parameters. For details on development of each module and integration see Methodology section. The model was developed and simulated using Matlab 2014b (http://mathworks.com).

Fig. 2 The schematic of the metabolic network of the hepatic metabolism used in the model.

Model calibration

The model was calibrated from the source/component models referred from the literature. The parameters for hepatic metabolism were extracted from Konig et al. 2012 and Xu et al. 2011,^38,42 for insulin signaling from Sedaghat et al. 2002,⁴³ for glucagon signaling from Xu et al. 2011 and Mutalik et al. 2005,^42,44 for mTOR signaling from Vinod and Venkatesh 2009,⁴⁵ and insulin secretion kinetics from Dalla Man et al. 2007.⁴⁶ We tried to retain the reported parameter values from the source models allowing minimal deviation in them. The parameters for the model integration and optimization were estimated by the authors using optimization algorithms in Matlab. We used modular partial calibration methodology wherein each subsystem was optimized separately to a desired response and further integrated and re-calibrated to yield the similar optimal solutions.

Model validation

The model was validated by obtaining the dynamic concentration profiles of various metabolites and signaling molecules for resting state during 24 h fasting condition and comparing it with the literature data and the simulation results of the source models (see S1 file Fig. M1 and M2(I), M2(II) and M2(III)†). The model was validated to capture the reported qualitative behavior of the concentration profiles while the quantitative information was retained by matching the fold changes or the observed rates over the time frames. For most of the concentration profiles we used human data reported in literature; however we resorted to the scaled data from other animal models such as mouse and rats in the instances of lacking human data.

The dynamic profiles indicate that the storage compounds such as glycogen and triglycerides are degraded to maintain the other metabolites at a homeostatic level. The values matched the known homeostatic levels reported in literature (see S1 file Fig. M2(I) and M2(II)†). It can be noted that the regulators of the storage compounds, for example, mTOR (mammalian target of rapamycin) for amino acids, GSK3 (glycogen synthase kinase 3) for glycogen and SREBP (sterol response element binding protein) and PPARα and PPARβ (peroxisome proliferator-activated receptor alpha and beta) for triglycerides, also show unsteady behavior causing the storage compounds to be degraded. The key signaling molecules in the insulin signaling pathway attain a basal steady state since the pathway is not operational under fasting conditions (see S1 file Fig. M2(III)A, B & C†). However, during fasting, the glycogen signaling pathway is operational indicated by the activation of signaling molecules cAMP (cyclic adenosine monophosphate) and PKA (protein kinase A) (see S1 file Fig. M2(III)D & E†). The model for transcriptional network yielded the reported qualitative trends under resting and postprandial conditions (see S1 file Fig. M2(III)G, H and I†). The model was thus able to compare the physiological resting state thereby obtaining the steady state fluxes of various metabolic reactions.

Steady state metabolic flux distribution

The model was further used to determine/predict the effect of plasma macronutrient concentration on the steady state fluxes in the various metabolic pathways. Steady sate fluxes were obtained for different levels of plasma macronutrient (glucose, fatty acids and amino acid levels) for up to 4 fold changes of each of the macronutrient in plasma. These metabolite combinations were used as a proxy for the dietary macronutrient input to the system (i.e. combinations of low, medium, high and very high levels of carbohydrates (∼glucose), proteins (∼amino acids) and fats (∼fatty acids) in diet). The effects of these macronutrients were recorded through MFD (metabolic flux distribution) in the hepatic metabolic pathways. The representative MFD for the constant plasma macronutrient levels with high carbohydrate and protein with normal fat levels is shown in Fig. 3. The figure shows that under such a scenario, the fatty acid and cholesterol synthesis increase despite the normal dietary fat consumption. The lipogenesis flux is strongly activated under such a condition. These results are also in line with the experimental observations for the high carbohydrate diet in hamsters.⁴⁷ However, for a scenario wherein the plasma fatty acids levels are higher with normal carbohydrates and amino acids, the MFD show a increase in the gluconeogenesis flux with decreased fatty acid synthesis (see Fig. 4). The increase in gluconeogenesis under high fat diet has been experimentally confirmed in rats.⁹ This analysis highlights the non-linear dependence of metabolic fluxes on the plasma macronutrient levels. Such a metabolic flux distribution was used to obtain the fold change in the individual flux value relative to that observed under physiological resting state. We discuss the effect of macronutrient level on the fluxes through the fold change values in various metabolic reactions.

Fig. 3 The metabolic flux distribution for the scenario where the plasma glucose is set to 10 mmol l⁻¹, plasma amino acid are set to 0.5 mmol l⁻¹ and fatty acids set to 0.68 mmol l⁻¹. This represents the diet with high carbohydrates and protein with normal fat content. The yellow arrow shows the diversion of the metabolic flux towards the lipogenesis. The color code represent the fold change-green to colorless for negative fold change to zero (≤0), colorless to blue for zero to one (0–1) and blue to red for one to greater than one (1 to >1).

Fig. 4 The metabolic flux distribution for the scenario where the plasma glucose is set to 5 mmol l⁻¹, plasma amino acid is set to 0.25 mmol l⁻¹ and fatty acids set to 2.4 mmol l⁻¹. This represents the diet with normal carbohydrates and protein with high fat content. The yellow arrow represents the diversion of the metabolic flux towards gluconeogenesis. The color code represent the fold change-green to colorless for negative fold change to zero (≤0), colorless to blue for zero to one (0–1) and blue to red for one to greater than one (1 to >1).

We further used the model to study the intracellular metabolic flux variations with respect to the different combination of plasma macronutrient levels. The diet combinations involved variations in plasma amino acids and plasma fatty acid concentrations for different plasma glucose levels, namely, low (3 mM), normal (5 mM), high (10 mM) and very high (15 mM). Henceforth we denote the plasma concentrations of macronutrients on the scale of low to very high levels. Table 1 represents the concentrations and fold change values for each of the macronutrient correspondingly to the scale of low to very high level.

Table 1 The different levels of macronutrient and the corresponding physiological concentrations used during simulation

Macronutrient/level	Normal/ref. mmol	Low mmol	Medium mmol	High mmol	Very high mmol
Glucose	5	<5	5–8	8–12.5	>12
		<1 fold	1–1.6 fold	1.6–2.5 fold	>2.5 fold
Fatty acids	0.68	<0.68	0.68–1.36	1.36–2.04	>2.04
		<1 fold	1–2 fold	2–3 fold	>3 fold
Amino acids	0.25	<0.25	0.25–0.5	0.5–0.75	>0.75
		<1 fold	1–2 fold	2–3 fold	>3 fold

---

Gluconeogenesis and glycolysis

First, we present the effect of diet on glucose synthesis/assimilation by characterizing the gluconeogenesis/glycolysis fluxes. To address the glucose metabolism, the net steady state flux response around the substrate cycles in the pathway were recorded.⁴⁸ The flux differences at the irreversible steps in the glycolytic pathway were recorded to characterize the net flux towards gluconeogenesis and glycolysis. In this case, the flux value around the reversible reaction F6p to F16bp was recorded by normalizing the difference in flux through fructose 1,6-biphosphatase (gluconeogenesis) and phosphofructokinase (glycolysis) (see Fig. 5). In the figure, the value of 1 on the color bar represents the normalized value of the difference in the gluconeogenesis and glycolysis flux (normalized by the basal difference in the flux). As expected, gluconeogenesis is preferred for both low and normal glucose condition irrespective of the amino acid and fatty acid levels. High gluconeogenesis flux can be observed even under normal glucose condition but with a very high fatty acid levels. Glycolysis is dominant on increasing the plasma glucose levels at high amino acids–low fatty acids and low amino acids–high fatty acid levels.

Fig. 5 The graph represents the normalized difference in the gluconeogenesis (F16bpase flux) and glycolysis (PFK flux) for the F61bp to F6p and F6p to F16bp flux, respectively, for varying folds of plasma amino and fatty acids for four different glucose levels. A positive value on the color bar represents the prevalence of gluconeogenesis and a negative value represents the net flux to be as glycolysis. The subplots A, B, C & D represents the flux variations for plasma glucose concentration of 3 mM, 5 mM, 10 mM and 15 mM, respectively. Gluconeogenesis (F16bpase flux) increases with decreasing glucose levels and increasing amino-fat levels. However, with increasing glucose levels, gluconeogenesis further decreases with increasing fat levels under moderate amino acid levels. Glycolysis (PFK flux) increases with increasing glucose levels and decreasing fat levels. However at higher glucose levels the trend becomes nonlinear with fat and amino acids. At very high glucose levels, glycolysis is higher at either very low to moderate amino-fatty acid levels. It is mostly inhibited at very high amino-fat levels.

A similar trend is reflected in the glucose to G6p (glucose 6 phosphate) and G6p to glucose flux as observed in Fig. 6. It shows the normalized difference of the flux through glucose 6 phosphatase (gluconeogenesis) and glucokinase (glycolysis). The GK (glucokinase) flux increases with increasing glucose levels, whereas, G6pase flux increases with decreasing plasma glucose concentrations. However at higher glucose levels, GK flux is overcome by G6pase at higher amino and fatty acid levels.

Fig. 6 The normalized difference in the gluconeogenesis (glucose 6 phosphatase flux-G6pase) and glycolysis (glucokinase flux-GK) for the G6p to glucose and glucose to G6p flux, respectively, for varying levels of plasma amino and fatty acids for four different glucose levels. A positive value on the color bar represents the prevalence of gluconeogenesis and a negative value represents the net flux to be as glycolysis. The subplots A, B, C & D represents the flux variations for plasma glucose concentration of 3 mM, 5 mM, 10 mM and 15 mM, respectively. The GK flux increases with increasing glucose levels, whereas, G6pase flux increases with decreasing glucose concentrations. However at higher glucose levels, GK flux is overcome by G6pase at higher amino-fat levels. The GK flux is highest at high glucose-amino and low fat levels.

Under very high fatty acid and amino acid levels, gluconeogenesis is dominant over glycolysis indicating higher glucose release into the plasma. This is also reflected by plotting the glucose transport flux into the plasma, wherein high glucose release is observed under low plasma glucose condition and under high amino acid and high fatty acid levels (see Fig. 7). This is mainly due to the inhibition of AKT (protein kinase B), a glycolytic regulator, and activation of PKA that triggers gluconeogenesis under such conditions (see S1 file Fig. N1 and N2†). The inhibition of AKT at higher fatty acid levels even under high glucose levels is due to the over activation of PKC (protein kinase C) which has a negative feedback on AKT phosphorylation (S1 file Fig. N3†). It can be noted that under moderately high glucose levels, low amino acid and high fatty acid levels results in an efficient glucose uptake (i.e. no release). This may be significant under transient condition during food intake, where the plasma glucose levels rise to moderately high levels.

Fig. 7 The glucose transport flux for varying levels of plasma amino and fatty acids for four different glucose levels. A positive value on the color bar represents the glucose release into the plasma from the liver, whereas a negative value represents the uptake of the glucose by hepatic tissues. The subplots A, B, C & D represents the flux variations for plasma glucose concentration of 3 mM, 5 mM, 10 mM and 15 mM, respectively. Glucose release increases with decreasing glucose concentration in blood and vice versa. However, under higher glucose levels, glucose uptake increases at high fat and moderate amino acid levels. Under higher glucose levels, glucose release increases with higher amino and fatty acid levels.

The glycolytic flux through PFK (phosphofructokinase) in the pathway (downward flux towards pyruvate) is plotted in Fig. 8. In this case, under normal glucose, amino acids and fatty acid levels, the glycolysis is partially active with gluconeogenesis also operational indicating that gluconeogenesis is being accounted for by glycogen breakdown, while some of the flux contributes to the energy requirements of the liver. It can be distinctively noted that the PFK flux increases with increasing amino acid levels under low to moderately high glucose levels and moderate fatty acid levels. This increase in glycolytic flux indicates the saturation of glycogen levels and the flux is directed towards pyruvate. However, under very high glucose levels, the glycolytic flux is highest either at low fatty and amino acid or moderate amino acid and high fatty acid levels (see Fig. 8D).

Fig. 8 The phosphofructokinase flux (PFK), for varying levels of plasma amino and fatty acids for four different glucose levels. The subplots A, B, C & D represents the flux variations for plasma glucose concentration of 3 mM, 5 mM, 10 mM and 15 mM, respectively. PFK flux increases with increasing glucose levels and decreasing fat levels. However, at higher glucose levels the trend becomes nonlinear with fat and amino acids. At very high glucose levels, it is higher at either very low or very high amino acid levels. It is mostly inhibited at very high amino and fatty acid levels.

Glycogen metabolism

The metabolic process that affects glucose synthesis and storage is glycogen metabolism. The glycogen metabolism is characterized by plotting the normalized difference between the flux through glycogen phosphorylase (glycogen breakdown) and glycogen synthase (glycogen synthase), which catalyzes G6p to glycogen and glycogen to G6p reactions, respectively (see Fig. 9). The glycogen catabolic flux (i.e. glycogen breakdown) shows a similar trend as gluconeogenesis. Under low glucose levels, the glycogen breakdown is reduced under high fat levels which may result in lower glucose supply to the plasma.

Fig. 9 The difference in the glycogenolysis (glycogen phosphorylase) and glycogen synthesis (glycogen synthesis) for the G6p to glycogen and glycogen to G6p flux, respectively varying levels of plasma amino and fatty acids for four different glucose levels. A positive value on the color bar represents the prevalence of glycogen breakdown and a negative value represents the net flux towards glycogen synthesis. The subplots A, B, C & D represents the flux variations for plasma glucose concentration of 3 mM, 5 mM, 10 mM and 15 mM, respectively. Glycogen synthesis increases with increasing glucose concentration and decreases with increasing fat and amino acids for normal glucose levels. However at very high glucose levels, it increases with increasing amino acid levels provided that fat levels are moderate. It is reduced at high amino and fatty acid levels and at low amino acid and high glucose levels. Glycogen breakdown increases with decreasing glucose levels and increasing amino and fatty acid levels.

The glycogen synthesis is highest under moderately high glucose with either low fatty acid–high amino acid or high fatty acid–low amino acid levels. This can be due to higher activation of AMPK which inhibits glycogen synthesis under such condition (see ESI file II-Fig. S4†). Under very high glucose levels glycogen synthesis is not as efficient as that under moderate glucose levels. The signaling molecule regulating the glycogen metabolism (i.e. phosphorylated GSK3 (inactive) helps in glycogen synthesis) is also shown in Fig. 10. It can be seen that irrespective of the glucose levels, GSK3p is inhibited strongly at very high amino and fatty acid levels in the plasma, whereas it is highest under low amino and fatty acid condition for moderately high glucose levels. Under very high glucose levels, GSK3p is highest (almost 10 folds) for all the levels of fatty acid and amino acids, except high fatty and high amino acid levels due to inhibition of insulin signaling at very high amino and fatty acids.

Fig. 10 The levels of phosphorylated glycogen synthase kinase (GSK3), varying levels of plasma amino and fatty acids for four different glucose levels. The subplots A, B, C & D represents the flux variations for plasma glucose concentration of 3 mM, 5 mM, 10 mM and 15 mM, respectively. The phosphorylation of glycogen synthase kinase increases with increasing glucose levels and decreases with increasing amino and fatty acid levels. It is mostly inhibited at high amino-fatty acid levels.

Pyruvate and lactate metabolism

The trends observed in the glycolysis flux are also directly reflected in the pyruvate uptake from plasma, wherein the uptake of pyruvate is low under higher glycolytic conditions (S1 file Fig. M3†). Moreover, pyruvate is released under higher glycolysis conditions, wherein the PFK to F16bpase flux difference is highest as noted in Fig. 5 (darker regions in the plots). While pyruvate uptake is high under low glucose levels, the pyruvate release increases at very high amino acid glucose levels under low fat levels due to its higher synthesis rate. This indicates that the accumulation of carbon from high glucose is mainly channeled towards fatty acid synthesis. In liver, pyruvate is also synthesized from alanine and lactate under normal physiological conditions, whereas, under excessive pyruvate production, the fate of pyruvate can be towards lactate through lactate dehydrogenase. The figure shows the normalized difference of the pyruvate to lactate reversible flux, wherein the positive flux represents the net flux towards pyruvate formation and the negative value represents the net flux towards lactate formation (see Fig. 11). The lactate to pyruvate flux is reduced for low glucose levels only under moderate to high amino acid and low fatty acid levels, thereby reducing the efficiency of gluconeogenesis as reflected in Fig. 5. This is also reflected in the trends of NADH/NAD ratio (nicotinamide adenine dinucleotide) under these conditions. Whereas pyruvate synthesis increases with increasing fatty acid levels under lower and normal glucose levels.

Fig. 11 The reversible lactate dehydrogenase flux for varying levels of plasma amino and fatty acids for four different glucose levels. A positive value on the color bar represents the lactate to pyruvate flux, whereas a negative value represents the pyruvate to lactate flux. The subplots A, B, C & D represents the flux variations for plasma glucose concentration of 3 mM, 5 mM, 10 mM and 15 mM, respectively. Pyruvate synthesis increases with increasing fatty acid levels for lower and normal glucose levels. However under higher glucose levels, lactate synthesis is favored with increasing fatty acid levels for medium amino acid levels.

However, under higher glucose levels lactate synthesis is favored with increasing fatty acid concentrations under moderate amino acid levels. This increase in lactate flux is due to higher pyruvate synthesis from its glycolytic precursors and the favorable NADH/NAD ratio under the conditions of high glucose and fatty acid levels (see Fig. 12). Similar trend is further reflected in the lactate transport flux, wherein higher lactate uptake is associated with increased flux towards pyruvate synthesis and vice versa (see S1 file Fig. M4†). Moreover, the flux through pyruvate carboxylase which catalyzes pyruvate to oxaloacetate (see Fig. 13) follows the similar trend as of depicted by the gluconeogenesis flux and lactate dehydrogenase flux (Fig. 5 and 11). This flux utilizes partial TCA cycle, wherein the pyruvate is reutilized for the gluconeogenesis. This flux also facilitates the utilization of alanine and lactate as the substrates for gluconeogenesis through pyruvate. Under high glucose, amino and fatty acid levels, both lactate dehydrogenase and pyruvate carboxylase fluxes are higher. This indicates that the competition between these two fluxes decides the major fate of pyruvate (i.e. glucose or lactate) under these conditions (see Fig. 11D & S1 file Fig. M5(D)†).

Fig. 12 The NADH/NAD ratio for varying levels of plasma amino and fatty acids for four different glucose levels. The subplots A, B, C & D represents the flux variations for plasma glucose concentration of 3 mM, 5 mM, 10 mM and 15 mM, respectively. The NADH/NAD ratio increases with increasing amino acid levels and decreases with increasing fatty acid levels for normal glucose levels. With increasing glucose levels the ratio decreases slightly and is nonlinear with amino acid and fatty acid levels. However, under high glucose levels, higher levels of amino and fatty acids can restore normal ratio.

Fig. 13 The pyruvate dehydrogenase flux for varying levels of plasma amino and fatty acids for four different glucose levels. The subplots A, B, C & D represents the flux variations for plasma glucose concentration of 3 mM, 5 mM, 10 mM and 15 mM, respectively. The pyruvate dehydrogenase flux increases with increasing amino acid level and decreases with higher fatty acid levels for all glucose levels. However the flux increases monotonously with increasing glucose concentration and is inhibited at very high fatty and amino acid levels.

TCA cycle and ATP–ADP ratio

Flux towards TCA cycle is an indicator of the ATP (adenosine triphosphate) and fatty acid synthesis. Here, the TCA flux is indicated by pyruvate dehydrogenase flux that catalyzes pyruvate to acetyl-CoA. The TCA flux is low for low glucose level when gluconeogenesis is prevalent (see Fig. 13). However, under normal glucose levels, only under high fatty acid levels the TCA flux is reduced (see Fig. 13B). On further increasing glucose, TCA cycle is most preferred at low fat high amino acid levels, and strongly inhibited under high amino acid high fatty acid levels. The flux from pyruvate to ACoA (acetyl coenzyme A) is an amphibolic flux, being activated by insulin and glucagon signaling, thereby influences both anabolic and catabolic process. The flux is high through glucagon signaling activation under high amino and fatty acid levels (indicating catabolism), while it is high through insulin signaling activation under high glucose high fatty acid levels (indicating anabolism). Further, to characterize the energy status of the liver with respect to dietary macronutrient composition, we plotted the ratio of the ATP breakdown to ATP production rates, which is denoted by adenylate kinase and oxidative phosphorylation flux, respectively (see S1 file Fig. M6†). Subsequently, we also plotted the resulting ATP/ADP ratio (see Fig. 14).

Fig. 14 The ATP/ADP ratio for varying levels of plasma amino and fatty acids for four different glucose levels. The ratio above one represents the net surplus of ATP over ADP. The subplots A, B, C & D represents the flux variations for plasma glucose concentration of 3 mM, 5 mM, 10 mM and 15 mM, respectively. The ATP/ADP ratio is maintained at normal under normal glucose and moderate amino-fatty acid levels. The ratio decreases with decreasing glucose levels below normal. However, for higher glucose levels, high level of fatty acids is required to maintain the normal ratio. The ratio decreases with higher amino acid levels for high glucose levels. The ratio is drastically reduced at very high glucose and fatty acid levels.

The energy utilization is higher under low glucose levels, indicating that in liver the energy is being utilized for the gluconeogenesis. This is evident from the ESI file S1 Fig. M6(A)† and Fig. 14A, that under low glucose levels, while ATP breakdown is higher, the ATP/ADP ratio is lower indicating increase in ADP (adenosine diphosphate) levels. However, on increasing glucose levels, under moderate amino acid levels, ATP synthesis dominates under most levels of fatty acids (the ratio is less than one in ESI file II-Fig. M6(C & D)†). It is interesting to observe that under high fat high amino acid levels, the ATP synthesis is low indicating a reduced drive towards anabolic reactions. This is also reflected in ATP/ADP ratio, under low glucose concentrations where gluconeogenesis dominates (see S1 file Fig. M6A†). Under normal glucose concentrations and high amino acid levels the ATP/ADP ratio is lower indicating catabolic effect (Fig. 14B). However under low amino acid levels, the ATP/ADP ratio is near normal under all conditions. On increasing glucose (moderately high), the ATP/ADP ratio drops in most cases, except under moderate amino acid and high fatty acid levels. On further increasing the glucose (very high), it can be seen that ADP dominates with deficient ATP indicating abnormal anabolic conditions (see Fig. 14D). This is due to lower oxidative phosphorylation caused by higher insulin which inhibits PKA and calcium i.e. regulators of oxidative phosphorylation. The lower oxidative phosphorylation under high glucose levels is also accompanied by the higher ATP consumption due to anabolic condition resulting into the steep fall in the ATP/ADP ratio.

Pentose phosphate pathway

Next we consider the flux towards pentose phosphate pathway indicating the degree of anabolic reactions (biosynthesis) and the measure for the supply of redox equivalent (NADPH). This flux is represented by the rate of the flux through G6p dehydrogenase (abstracted for conversion of G6p to R5p (ribulose 5 phosphate)) (see S1 file Fig. M7†). As expected, the pentose phosphate pathway is off at low glucose levels irrespective of the amino and fatty acid levels. However it is reduced at higher fat levels for normal glucose levels. Under normal glucose levels, the pentose phosphate pathway is operational under basal levels. The maximum pentose phosphate pathway flux is observed under moderately high glucose and amino acid levels. Further at very high glucose, pentose pathway is operational at basal levels for moderate amino acid and high fatty acid levels. The pentose flux is inhibited at high amino and fatty acid levels. At very higher glucose levels, pentose flux is functional either at moderate amino acid levels or low to moderate amino acids and high fat levels (see S1 file Fig. M7(D)†). Moreover, in S1 file Fig. M7(D),† the operational region of pentose phosphate flux maps the conditions where the F16bpase and PFK fluxes are highly reduced, which implies the diversion of the flux towards pentose phosphate pathway under such dietary conditions. Such response enhances lipogenesis by providing more reducing equivalents under high fatty acids and higher glucose levels.

Fat metabolism

We next consider the flux towards lipogenesis (i.e. fatty acid and triglyceride synthesis). We firstly quantify the fatty acid synthesis by characterizing the flux through acetyl-CoA to malonyl-CoA catalyzed by acetyl-CoA carboxylase (ACC) (see S1 file Fig. M7†). Under low glucose levels, due to higher gluconeogenesis, lipogenesis is minimal for all levels of amino acid and fatty acids. Under normal glucose level, there is an enhanced lipogenesis, albeit at normal level under normal fatty and amino acid levels. On increasing glucose concentration further, the maximum lipogenesis is observed for moderate amino acid-low fatty acid and moderate fatty acid-low amino acid levels. It can be noted that, for very high plasma glucose levels, the lipogenic flux is operational in the region where the pentose phosphate pathway is also active (see S1 file Fig. M7(D)† and Fig. 15D). However, at higher glucose levels, both under low amino acid/low fatty acid and high amino acid/high fatty acid, the lipogenesis is completely inhibited which is due to lower levels of ATP levels countering the lipogenesis (an anabolic process). This is also associated with the state of the lipogenic regulators,⁴⁹ wherein CHREBP (carbohydrate response element binding protein) an activator of lipogenesis increases with glucose and inhibited by higher fatty acids and amino acids due to activation of AMPK (AMP activated protein kinase) (an inhibitor of CHREBP) under such conditions (see S1 file Fig. N5†). TRB3 (tribbles homolog 3) is an inhibitor of lipogenesis. It is activated at high fatty acid under normal glucose level, which inhibits AKT activity that is required for lipogenesis (see Fig. N6†).

Fig. 15 The flux through lipogenesis [fatty acid synthesis] represented by acetyl CoA carboxylase (ACC) flux that catalyzes ACoA to malonyl CoA for varying levels of plasma amino and fatty acids for four different glucose levels. The subplots A, B, C & D represents the flux variations for plasma glucose concentration of 3 mM, 5 mM, 10 mM and 15 mM, respectively. At normal glucose levels, fatty acid synthesis is higher at low fatty acid and moderately higher amino acid levels. It increases with increasing glucose and moderately high levels of amino acids. It is inhibited at high fatty acid and high amino acid zone.

The triglyceride metabolic flux is characterized by the flux ratio for triglyceride synthesis to triglyceride breakdown (see Fig. 16). The triglyceride synthesis is low, as expected, under low glucose levels irrespective of the dietary amino and fatty acid levels. Under normal glucose levels, its synthesis is high under moderate fatty/amino acid levels. The triglyceride synthesis space increases on further increasing glucose, with high synthesis rates noted for low to moderate amino acid and high fatty acid levels. The triglyceride synthesis is activated by PPARγ (peroxisome proliferator-activated receptor gamma), which in turn is activated by insulin and fatty acids (see S1 file Fig. N7†). This helps in the anabolic accumulation of triglycerides in the liver under these conditions. However, for very high glucose level, the system limits ATP for anabolic reactions to happen thereby reducing triglyceride synthesis. Although PPARγ is activated at higher fatty acid levels, AKT is inhibited due to activation of FOXO (fork head box protein) which is operational under high fatty acid levels (see S1 file Fig. N8†). Moreover, the activation of PPARα under very high fatty acid levels induces triglyceride and fatty acid breakdown thereby reducing lipogenesis (see S1 file Fig. S9†). The triglyceride release into the blood also mimics a similar behavior as that of its synthesis (see S1 file Fig. M8†).

Fig. 16 The flux ratio of triglyceride synthesis to triglyceride breakdown for varying levels of plasma amino and fatty acids for four different glucose levels. The value below one represents the net flux is towards triglyceride breakdown and vice versa. The subplots A, B, C & D represents the flux variations for plasma glucose concentration of 3 mM, 5 mM, 10 mM and 15 mM, respectively. Triglyceride synthesis decreases with lower and very high glucose levels and increasing fatty acid levels under normal glucose levels. It is higher at moderate glucose, amino acid and high fatty acid levels. However at very high glucose levels, higher amino acid level and moderate fatty acid level increases TG synthesis.

Cholesterol metabolism

The cholesterol biosynthesis flux is characterized by the flux through the HMG-CoA (3-hydroxy-3-methyl-glutaryl-CoA) reductase that catalyzed the conversion of HMG-CoA to mevalonate (a rate limiting step in cholesterol biosynthesis pathway) (see Fig. 17). Under low glucose levels the cholesterol biosynthesis is the lowest and increases marginally under normal glucose, fatty acid and amino acid levels. The activation of glucagon under these conditions results in activation of PKA which inhibits cholesterol synthesis. On further increasing glucose levels the cholesterol biosynthesis increases further under marginally higher levels of amino acids and fatty acids. Under very high plasma glucose levels, moderately high amino acids and the high fatty acid level results in maximum flux towards cholesterol biosynthesis. This is due to the higher SREBP levels activated by insulin and fatty acids under this condition. However, higher amino acid levels reduce the flux towards the biosynthesis of cholesterol. SREBP a regulator of HMGR is reduced due to inhibition of AKT at higher amino and fat acid levels, whereas fat activates SREBP along with insulin, hence higher cholesterol synthesis (see S1 file Fig. S10†).

Fig. 17 The cholesterol biosynthesis flux, that is represented by the flux trough HMG-CoA reductase flux which catalyzed HMG-CoA to mevalonate for varying levels of plasma amino and fatty acids for four different glucose levels. The subplots A, B, C & D represents the flux variations for plasma glucose concentration of 3 mM, 5 mM, 10 mM and 15 mM, respectively. Cholesterol synthesis increases with increasing amino acids to moderate levels while glucose is maintained at normal levels. It increases with increasing glucose and fatty acid levels at higher glucose concentration; however it decreases at higher amino acid levels. It is reduced at lower glucose levels and inhibited at high amino and fatty acid levels.

Amino acid and protein metabolism

The analysis shows that the amino acid uptake increases with increasing amino acid and fatty acid levels, while it decreases with increasing glucose levels (see S1 file Fig. M9†). This suggests that the gluconeogenesis from amino acids is mainly operational under low glucose level. Further, it can be noted that amino acid uptake is lowest under high glucose, low amino acids and high fatty acid levels (see Fig. M9(C & D)†). Higher glucose levels essentially reduce gluconeogenesis which makes the amino acid uptake flux redundant. Since, amino acids are mainly a source of carbon for glucose and protein synthesis in liver, such a flux is observed. It is also interesting to note that the conditions that show higher amino acid uptake overlap with that of higher gluconeogenesis, indicating that amino acids are one of the major substrates for gluconeogenesis. The protein metabolism was characterized by plotting the normalized flux difference between protein breakdown and synthesis flux (see Fig. 18). The protein synthesis in liver is mainly under high amino acid and low fat levels.

Fig. 18 The normalized flux difference between protein breakdown and protein synthesis for varying levels of plasma amino and fatty acids for four different glucose levels. The negative value on the color bar represents the net protein synthesis flux. The subplots A, B, C & D represents the flux variations for plasma glucose concentration of 3 mM, 5 mM, 10 mM and 15 mM, respectively. Protein synthesis increases with increasing amino acids and increasing glucose levels and low fatty acid levels. However it is reduced by increasing fatty acid levels.

Further, protein synthesis increases with higher glucose levels. However it is reduced with increasing fatty acid levels, thereby increasing its breakdown under low glucose, low amino acid and high fatty acid levels. Protein synthesis decreases with increasing fatty acids due to inhibition of AKT and subsequent activation of PKA that activates protein breakdown. Wherein the protein synthesis is regulated by insulin and amino acid mediated activation of mTOR and S6Kp which also increases with increasing amino acid and glucose levels (see S1 file Fig. N11 and N12†). It should be noted that, under low and normal plasma glucose levels, the protein synthesis is in parallel to amino acids being channeled towards gluconeogenesis, whereas, under high glucose levels, protein synthesis is in contrast to region of gluconeogenesis.

The balance of the nitrogen in the system is regulated through urea cycle.^50,51 The flux through urea cycle is characterized by carbamoyl phosphate synthase flux that catalyzed ammonia to carbamoyl phosphate (see Fig. 19). The flux through urea cycle increases with increase in amino acid levels and decrease in glucose and fatty acids levels in plasma. It can be noted that under high protein and high glucose levels, moderately higher levels of fatty acid are required to maintain the flux through urea cycle. Subsequently, it can be seen that ammonia release is maximum under high fatty acid/high amino levels wherein the urea cycle flux is inhibited (see S1 file Fig. M10†). This is due to the inhibition of the urea cycle flux due to the activation of PPARα by fatty acids and deactivation of PKA due to increased glucose levels. Thus, the ammonia in the system is also dependent on the dietary composition of fatty acids and amino acids.

Fig. 19 The urea cycle flux represented by the normalized rate of carbamoyl phosphate synthase that catalyzed ammonia to carbamoyl phosphate for varying levels of plasma amino and fatty acids for four different glucose levels. The subplots A, B, C & D represents the flux variations for plasma glucose concentration of 3 mM, 5 mM, 10 mM and 15 mM, respectively. Urea cycle flux increases with increasing amino acids and decreases with increasing glucose and fatty acid levels. However, under high amino acid and high glucose levels, moderately higher levels of fatty acids restore the normal urea cycle flux. It is highest at low glucose, low fat and high amino acid levels.

Discussion

In order to quantify the effect of plasma macronutrients on metabolic fluxes in liver, a detailed model including signaling and transcriptional regulations was developed. The model predictions revealed several signatures of metabolic performance under different levels of fat, amino acids and glucose in the plasma. Using these regulatory signatures we could qualitatively rationalize several experimental observations in the metabolic phenotypes associated disease states reported in literature. The model reveals that glucose, fatty acids and amino acids have differential effects on the secretion and activity of the metabolic hormones (insulin and glucagon) thereby resulting in a highly nonlinear metabolic control. The analysis indicated that a steady state metabolic flux is collectively determined by the regulatory effects of signaling components, transcriptional factors and the metabolic controllers (ATP/ADP and NADH/NAD ratios).

Alternative to the results reported above, we summarize the overall effect of diet (plasma macronutrient levels) on the key metabolic pathways in a tabular form (see ESI file, Excel file, S2_Table†). The table reports a relative flux ratio to the flux under physiological resting. The trends in the results of our model were motivated to explain the qualitative metabolic responses observed in experiments reported in literature. However, the quantitative validation of the model predictions with each of the experimental observations is out of the scope of present manuscript.

How high levels of fatty acids and proteins can increase gluconeogenesis and decrease glycogen synthesis leading to hyperglycemia?

The gluconeogenesis is known to be fairly constant in healthy individuals under varying dietary perturbations.⁵² However, for steady state perturbations, the analysis demonstrated that gluconeogenesis was activated at lower plasma glucose levels and was also induced even at constant glucose levels with increasing fatty acid composition. These results were in agreement with the observations reported on humans.^1,3 This was due to the inhibition of insulin signaling pathway that reduced the glycolytic flux and glycogen synthesis, resulting in a higher net gluconeogenic flux. Moreover, under high plasma glucose levels with increasing amino acid levels above 2.5 to 3 fold have shown to inhibit insulin action leading to de novo glucose synthesis from amino acids. Similar effects were observed in the investigation on rats fed on high protein diet.⁵³ Chevalier et al. have observed such effects in obese individuals, wherein increased rate of protein catabolism contributed to greater rate of gluconeogenesis and subsequent increase in glucose release.¹⁶ The sensitivity of gluconeogenesis increased with amino acids under higher glucose levels and similar results were also reported for a protein rich-low carbohydrate diet in humans.⁵⁴ The inhibition of insulin signaling pathway was associated with enhanced effect of glucagon signaling pathway being responsible for glycogen breakdown and gluconeogenesis, which represented an insulin resistant state.

Glycogen synthesis is a key mechanism in storing the excess glucose from the blood into liver. Defects in glycogen metabolism have been shown to be one of the main reasons for hyperglycemia.^55,56 The glycogen synthesis flux was quite sensitive to plasma levels of amino acids and fatty acids. Glycogen synthesis followed the plasma glucose and insulin levels, whereas its potential was reduced at very high amino acid levels thereby disabling sufficient glucose uptake. Such an effect of high protein diets on hepatic glycogen metabolism in mice and rat have been documented in literature.^10,57 Taylor et al. have demonstrated that postprandial glycogen storage flux follows the insulin to glucagon ratio in blood⁵⁸ which is in agreement with our analysis. Under low glucose level (i.e. under starvation or higher physical activity), where glycogen breakdown is anticipated, increasing amino acids can further increase glycogen breakdown, whereas higher levels of fatty acids reduced glycogen breakdown. This reduction in glycogen breakdown flux under high fat diet was also confirmed in rats.^12,59 This suggested that for an obese individual, whose circulating fatty acid levels are high, it would be difficult to obtain a faster rate of glycogen breakdown and subsequent glucose release as compared to a normal individual under lower plasma glucose condition.

How high fat diets induce defects in TCA flux leading to an insulin resistance state?

The TCA cycle in liver acts as an amphibolic pathway, which serves both anabolic and catabolic purpose in hepatocyte through its ability of anaplerosis and cataplerosis, respectively.⁶⁰ Under surplus energy (ATP) condition the flux was diverted towards lipogenesis or amino acid synthesis (anabolic) and under lower ATP states, the pyruvate, fatty acids and the amino acids are collectively utilized for the synthesis of ATP (catabolic) and gluconeogenic precursors, via TCA cycle. Therefore, the net abundance of these metabolites and the energy status of the cell decided whether the TCA cycle operate under catabolic or anabolic mode. The analysis indicated that the pyruvate dehydrogenase flux increased linearly with increasing glucose and decreased with increasing fatty acid levels. However, under low glucose levels this flux increases with increasing amino acid levels to cope up with the ATP requirement of the cell in a catabolic manner. With increasing plasma glucose levels, excess glucose was diverted to lipogenesis via pyruvate dehydrogenase that deployed partial TCA cycle in an anabolic manner. However, under very high fatty acid levels, β-oxidation was activated due to another homeostatic constraint, i.e. to maintain fatty acid levels. The pyruvate carboxylase flux increased under very high glucose and high fatty acid levels, thereby diverting the TCA flux towards gluconeogenesis. Under higher fatty acid levels PPARα was activated by PGC1 (PPAR gamma coactivator 1) mediated mechanism which further enhanced fatty acid breakdown. Therefore, higher levels of acetyl CoA generated through β-oxidation inhibited pyruvate dehydrogenase thereby reducing the glycolytic flux towards TCA cycle. TCA cycle was thus activated catabolically to utilize excess acetyl CoA in the form of energy or de novo glucose synthesis. At cellular level, this mechanism acts to economize the energy production through either of the substrates (glucose or fat) under surplus conditions. The two observations of increased lipolysis and gluconeogenesis were also confirmed by a study on humans reported by ref. 11 and 14. Therefore, glucose homeostasis is destabilized by excess fatty acids due to the inherent metabolic control in TCA cycle which would eventually lead to a diabetic state, under high fat dietary intake.

How lipogenesis and triglyceride synthesis are affected due to high carbohydrate and fat diet leading to a diabetic state?

In lipogenesis, fatty acid synthesis was favored with increasing glucose (up to 2 folds) levels and moderate amino acid levels, however, it decreased with increasing fatty acid levels and very high amino acid levels due to the inhibition of insulin signaling and activation of PKA (i.e. catabolic activity). This suggested that lipogenesis was favored under low fatty acid and high glucose levels which also assured the maintenance of fatty acid homeostasis in the cell. The variation in lipogenic flux was in line with recent experimental studies performed on rats that were fed on high carbohydrate and high fat diet.^8,61 At very high glucose and fatty acid levels, the lipogenic flux reduced due to the fall in ATP levels and induced β oxidation through the activation of PPARα. One of the major fates of high levels of circulating plasma glucose was to be stored as triglycerides via lipogenesis which also required higher consumption of ATP in the cell. However, at higher glucose levels, oxidative phosphorylation was compromised due to high insulin levels which inhibited the activators (PKA and calcium) of oxidative phosphorylation. This puts forth a constraint on the disposal of glucose through lipogenesis at very high glucose levels. Moreover, it was also limited by the correspondingly lower flux through the pentose phosphate pathway which supplied NADPH (nicotinamide adenine dinucleotide phosphate) as reducing power required for lipid synthesis. This phenomenon provided an insight into the patho-physiology of diabetic conditions wherein higher plasma glucose might put a positive feedback on its circulating levels due to reduction in the lipogenesis.

Similarly in triglyceride metabolism, triglyceride synthesis increased with increasing glucose and fatty acid levels, however at very high glucose and amino acid levels the TG synthesis is reduced. The reduction in TG synthesis with increasing amino acid levels was in line with the study that demonstrated the reversal of hepatic steatosis with high protein diet in mice.^4,62,63,64 On the other hand, triglyceride breakdown increased with decrease in glucose and fatty acid levels below the normal level. This is also confirmed by ref. 61 and 65 in their study on rats. Triglyceride breakdown was further induced at very high glucose levels due to lack of ATP in the system. In terms of diabetic pathogenesis, this suggested that, at very high glucose levels (>14 mmol l⁻¹), fatty acid levels might increase due to TG breakdown, which would further increase the negative feedback of the fatty acid on the insulin action that aggravates the diabetic state by decreasing the rate of glucose uptake.

How a high carbohydrate diet increases cholesterol levels? How a high protein–low fat diet can reduce cholesterol synthesis and help in reducing hypercholesterolemia?

Liver is the major site for biosynthesis of cholesterol. Cholesterol synthesis increased with increasing glucose and fatty acid levels and reduced at very high amino acid levels. High carbohydrates activates SREBP (a regulator of HMGCoA reductase) through the increase in insulin levels thereby diverting the flux towards cholesterol synthesis. However it increased with low fat and moderate amino acid levels under high glucose levels. These results are in agreement with the dietary studies on humans and rats.⁶⁶ This suggested that certain amount of amino acid (1.25 to 2.5 fold of normal) was essential for cholesterol synthesis along with fatty acid and glucose. Therefore, the analysis demonstrated that maintaining the plasma amino acids either below 1.25 folds or above 3 fold (unusually high) levels can help in reducing cholesterol even under higher glucose and fatty acid levels. These effect of low carbohydrate, high fat and high protein diet on cholesterol homeostasis in mice was also documented.⁶⁷ Higher levels of amino acids inhibits the AKT mediated activation of SREBP thereby reducing the flux through cholesterol synthesis. The observation suggested that, higher levels of plasma amino acids under a diabetic state can help in reducing the HMG-CoA reductase flux there by reducing hypercholesterolemia.

How high glucose and fat reduces protein synthesis? How a high fat diet increases plasma ammonia levels?

In case of protein metabolism, protein synthesis increased with increasing amino acids and glucose levels and decreasing fatty acid levels. Protein breakdown increased with increasing fatty acid levels and decreasing glucose and amino acid levels. These effects were also demonstrated in rats fed on high fat diet.^63,68 This is due to the inhibition of insulin signaling and subsequent activation of glucagon signaling by higher fatty acid levels. The metabolic flux observed under high fat levels explain the limitation of protein synthesis or decrease in muscle density under diabetic state. Although higher glucose levels help in protein synthesis, when followed by higher fatty acid levels the protein synthesis was hampered. The urea cycle facilitated the homeostasis of the ammonia that is generated during amino acid breakdown. The urea cycle flux increased with higher amino acid and lower glucose levels under moderate fatty acid levels.⁶⁹ The higher amount of amino acid influx to the liver induced a gluconeogenic state in liver; wherein most of the amino acids were used for de novo synthesis of glucose. Therefore, the nitrogen part of the carbon backbone of the amino acids was liberated as ammonia which was disposed through urea cycle.⁷⁰ With increasing glucose levels the potential of urea cycle decreased due to reduction in gluconeogenic flux by insulin and utilization of amino acids for protein synthesis. Moreover, with increasing fatty acid levels, the levels of ammonia rose with increasing amino acids due to reduction in the urea cycle flux. A recent study demonstrate the suppression of urea cycle enzymes by a high fat diet in hamsters.⁷ High fat levels induce the activation of PPARα that inhibits urea cycle flux whereas high carbohydrate levels inhibits PKA which activates urea cycle thereby leading to reduced urea synthesis resulting into increase in ammonia levels. Due to its neurotoxicity the ammonia levels were strictly under homeostatic control, therefore even 2 to 3 fold increments in plasma ammonia levels are detrimental. Hence, the analysis indicated the importance of not allowing the circulating levels of plasma fatty acid and amino acid levels to go very high simultaneously for ammonia homeostasis.

How high protein and fat levels can affect hepatic glucose release leading to hypoglycemic or hyperglycemic states?

One of the important transport flux is the hepatic glucose release which is reported to be distorted in case of diabetic condition.⁷¹ Insulin is known to regulate hepatic glucose production in direct and indirect mechanisms.⁷² The analysis demonstrated that at lower plasma glucose and with increasing plasma amino acid levels the hepatic glucose release rate increased as reported by,⁷³ whereas at high amino acid and fatty acid levels the release rate was restricted to a normal level (instead of increasing). Under conditions of starvation or higher physical activity, the lower plasma glucose levels led to an increase in the plasma glucagon levels. Glucagon triggers gluconeogenesis and glycogenolysis with the activation of cAMP, PKA and calcium signaling in liver. However at very high levels of amino acids and fatty acid levels insulin secretion was triggered which further inhibited the action of PKA through AKT. Under such a condition, although the plasma glucagon level was high there was no subsequent rise in the hepatic glucose release. This shows that higher circulating levels of plasma amino and fatty acids can reduce hepatic glucose release irrespective of the plasma glucagon levels.

Under resting state and normal glucose levels, increasing fatty acids to 3–4 folds increased glucose release by 20–25% due to the inhibition of AKT by fatty acids. These effects of high fat diet on fasting glucose were demonstrated in healthy men.⁷⁴ Under the postprandial state, with increasing plasma glucose levels, the glucose uptake increased; however, the uptake rate decreased with increasing amino acid and fatty acid levels, even leading to glucose release. This reduction in insulin's action under high fat and relatively low carbohydrate diet is demonstrated in a study conducted on humans.⁷⁵ Under such condition, the higher levels of amino acids triggered glucagon secretion and subsequent activation of PKA and S6K which inhibited insulin signaling along with further inhibition by fatty acid. Henkel et al. have reported a similar increment in plasma glucagon levels under postprandial state in the subjects with glucose intolerance and type 2 diabetes.⁷⁶ Moreover, it led to a lower ATP/ADP ratio which limited the conversion of glucose to G6p leading to higher cellular glucose and the reversal of glucose uptake flux. Therefore, even under high levels of circulating plasma insulin, the cellular state was shifted to a catabolic mode with activation of gluconeogenesis instead of glycolysis and resulted in glucose release instead of its uptake. Such a condition depicted a diabetic state or insulin resistance irrespective of the insulin levels just due to the metabolic shift that the macronutrients induced in the cells.¹⁵ In a diabetic state, wherein plasma glucose levels are already higher, higher intake of amino acids and fatty acids can further aggravate glucose levels.

How high glucose levels can affect hepatic fatty acid uptake leading to dyslipidemia and non-alcoholic fatty liver disease (NAFLD)?

Higher levels of plasma fatty acids and triglycerides are also indicators of a disease state in obesity and dyslipidemia.^77–79 The hepatic fatty acid uptake increased with 2–2.5 fold of plasma fatty acid levels and was further reduced at higher fatty acid levels under resting glucose condition; however, it increased with 2–2.5 fold increase in plasma glucose levels. The fatty acid uptake was mainly dependent on the cellular ATP/ADP ratio and insulin levels. The fatty acid uptake was drastically reduced at very high glucose levels except for very high levels of plasma amino acids and fatty acids. This was due to the lower levels of ATP under very high glucose levels which limited the conversion of fatty acids to triglycerides. In such a condition, even though the plasma insulin levels were higher the hepatic fatty acid uptake was reduced which can lead to higher levels of plasma fatty acids due to distortion in the capacity of this flux to maintain homeostasis.^3,80

The triglyceride release followed the fatty acid uptake flux in the range of lower to moderate levels of plasma glucose levels; however, it was inhibited at higher levels of amino acids due to inhibition of insulin signaling. The release was completely suppressed at very high glucose levels due to lack of cellular ATP levels and insulin resistance induced by very high amino acid and fatty acid levels.⁸¹ Although fatty acid uptake increased under very high levels of all the three macronutrients, the triglyceride synthesis was suppressed. This condition can result in higher levels of cellular fatty acid and further inhibition of insulin signaling by a DAG–PKC mediated mechanism thereby leading to insulin resistance,⁸² and non-alcoholic fatty liver disorder.^80,82,83 The above observation provided insights into how a diabetic state (hyperglycemia) can lead to higher plasma fatty acid levels and the resulting metabolic states can put a positive feedback on insulin resistance, and thus stabilizing the diabetic state.

Conclusion

In summary, the metabolic status of a tissue depends upon the ratios of the metabolic controllers such as ATP/ADP and NADP/NADPH, and the phosphorylation states of the regulatory signaling proteins. The metabolic state of a tissue then influences the transport fluxes from the tissue which in turn govern the plasma metabolite levels. The transport fluxes are the resultant effects of plasma macronutrient levels and the subsequent hepatic metabolic state. The phosphorylation states of the signaling molecules also strongly influence the levels of ATP/ADP ratio. This is further translated to overall metabolic pathways that use ATP–ADP as co-substrates and affects the synthesis and transport process of key metabolites. In this study, we demonstrated the perturbations in these regulatory mechanisms due to plasma macronutrients and several resulting metabolic states representing healthy and disease states.

Thus, the developed model provided insights on the functioning of cellular metabolism that arise due to several combinations of the plasma levels of the major macronutrients that are part of our daily diet. These plasma profiles are highly dynamic in nature due to time varying dietary interventions and cells have to constantly regulate its metabolism to achieve homeostasis. Any perturbations due to either external factors such as diet and exercise or internal factors such as hormonal ratios and signaling or transcriptional events can influence the metabolic phenotype. Therefore, our analysis reveals the signatures of plasma metabolite profiles that can defile the homeostasis due to de regulatory effects caused by specific levels of macronutrient and their combinations. The analysis can be further extrapolated to understand the dietary requirements so as to assist the homeostasis by appropriate dietary composition. Nevertheless, this study helps in visualizing the metabolic profiles under abnormal plasma levels of key metabolites which might occur due to various disease states.

Methodology – mathematical model for liver metabolism

The model consists of central metabolic pathway including glycolysis, gluconeogenesis, glycogen metabolism, TCA cycle, fatty acid synthesis and oxidation, protein synthesis and breakdown, urea cycle, pentose phosphate pathway, cholesterol biosynthesis and hexosamine pathway (see Fig. 2).^{34,35,38,42,55,84} The model was further integrated with sub-models for several signaling and transcription networks. Moreover, we have extended the model to incorporate the whole body plasma metabolite homeostasis to analyze its effect on liver. The developed model integrates several reported sub-models in conjunction with models developed for signaling and transcriptional regulation adopting a systems level approach.⁸⁵ The overall model for the liver metabolic module consisted of 272 rate equations, 170 ODEs and 801 parameters. The integrated model is composed of four modules viz.,¹ blood (metabolites and hormones),² metabolism,³ signaling and⁴ transcription. The detailed model and parameters are explained in ESI file S3.†

The blood module represents the dynamics of plasma metabolite concentrations at whole body level. It includes the kinetics of hormonal secretions (i.e. insulin and glucagon) in the blood from pancreas in response to plasma macronutrient levels.^86–88 The blood module accounts for the facilitated transport from blood to tissue of seven metabolites viz., glucose, lactate, pyruvate, amino acids, fatty acids, glycerol, triglycerides, and the passive transport of oxygen and carbon dioxide.⁸⁴

In the metabolism module, the metabolic pathways (as mentioned above) required for the processing carbohydrates, lipids and proteins in liver were modeled along with their regulations at metabolic, signaling and transcriptional levels. The hormonal (insulin and glucagon) and nutrient (glucose, amino acids and fatty acid) signaling pathways were adopted from literature^43,44,89 and integrated together for metabolic regulation. The signaling network composed of the feedbacks and crosstalk between insulin signaling mediated through AKT and mTOR signaling and glucagon signaling mediated through calcium and cAMP signaling. Furthermore, the transcriptional network was modeled to incorporate the long-term/genetic effects of plasma macronutrients on the synthesis and activation of metabolic enzymes and the signaling proteins. The transcriptional network consisted of the ten transcriptional factors such as SREBP, ChREBP, CREB (cAMP response element-binding protein), CEBPα (CCAAT enhancer binding protein alpha), PGC1, TRB3, FOXO, PPAR (γ, α, β) and AMPK along with the inputs from the signaling and metabolic networks.

The regulation of a metabolic enzyme by a signaling/transcriptional component was modeled by assuming the parallel activation of other enzymes in a linear pathway.⁹⁰ This assumption ensures that the activation or inhibitions of all the enzymes in a linear pathway are similar to yield a balanced flux through the pathway. The regulatory effects of the signaling endpoints were incorporated in the metabolic reactions, wherein, these regulations were assumed to influence the maximal rate of an enzymatic reaction. The anabolic regulatory effects on the metabolic pathways were mediated by the insulin signaling components and the catabolic effects were mediated by the glucagon signaling components. The modules are interconnected through several common components such as metabolites and active hormonal concentrations that synchronize together to establish a metabolic state as a result of an input function. The parameters of the models were obtained by flux balance analysis, regression and by the least square fit technique used for in silico fitting of an expected output response for a sub network. The optimal estimates of the parameters were those that gave best least square fit by minimization of the sum of errors for an objective function to the data obtained from literature either through experimental data or through validated model simulations. We tried to retain the reported parameter values from the source models allowing minimal deviation in them. We only tended to estimate the parameters for integrating the sub modules. Each sub module was independently calibrated to a known/reported experimental profiles and then integrated together to minimize the sum of the errors after integration. This allows us to constrain our calibration space and minimize the risk of overfitting. In this sense we reduce the degree of freedom by relying more on the reported parameters and models and the experimental data to fit the modular parameters (interactions and crosstalk between modules).

Blood module

The blood module depicts the surrounding medium of the liver tissue. It consists of the metabolites that have been considered as transport metabolites to the tissues and the hormones that are responsible for the metabolic regulations in the tissue. Two pools of the blood streams were considered viz., arterial blood and capillary blood (i.e. equivalent to venous blood) supplies to the tissue. It was assumed that the arterio-venous difference in the metabolite concentration is equal to the tissue metabolite uptake. Therefore the events of plasma metabolite flow was considered such as, the arterial blood is supplied to the capillary bed around the tissue and the plasma metabolites diffuse either passively or by facilitated manner to the interstitial fluid surrounding the tissue membrane from where the metabolites are taken up by the tissue. The interstitial fluid and capillary plasma metabolite concentrations are assumed to be in equilibrium. The resultant blood after the exchange and transport of the metabolites is termed as the venous blood. The physiological blood flow rate and the volume of the blood were considered to be constant. As per the experimental evidence, the blood flow rate regulation by the plasma hormonal concentrations (insulin) was also accounted.

We have also considered the plasma concentrations of two major metabolic regulatory hormones namely, insulin and glucagon. The secretion of hormones is known to be regulated by nutrients in the plasma.^19,91 The plasma concentrations of insulin were modeled as a function of plasma glucose, amino acids and fatty acids by fitting an appropriate Hill function to the experimental data from literature.^24–27,92 The Hill fit for the plasma insulin levels with respect to plasma glucose was obtained from experimental data reported by Konig et al. 2012.

The experimental data for the effect of amino acids on plasma insulin was extracted from the dynamical data reported by Calbet and MacLean, 2002 and Loon et al. 2000, for different amino acid inputs^86,88 The data for the effect of fatty acids/lipids on plasma insulin was extracted from the dynamical data reported by Gravena et al. 2002 and Manco et al. 2004 (ref. 93 and 94). Since there was scarcity of the dose response curves for amino acid and fatty acid effects on plasma insulin levels, the dynamical data was used to obtain steady state points and was used to obtain the Hill fits based on the fold changes in plasma insulin levels for different amino acids (see S1 file Fig. M1(A–C)†).

The plasma glucagon concentration was modeled as function of plasma glucose and amino acid concentrations.^95,96 In our study, we varied the arterial plasma concentrations of glucose, amino acids and fatty acids and measured the steady state response of the metabolic fluxes and the metabolite concentrations.^86–88,97

The rate of insulin secretion was modeled as

where V_Glu, V_AA and V_FFA are the maximal insulin concentrations with respect to glucose, amino acids and fatty acids, respectively. C_aGlu, C_aAA and C_aFFA are the concentrations of glucose, amino acids and fatty acids in the arterial blood. n_g, n_a, n_f and K_Glu, K_AA, K_FFA are the Hill coefficients and the half saturation constants for glucose, amino acids and fatty acids, respectively. This rate was further incorporated into the kinetic model for the liver and plasma insulin levels developed by Dalla Man et al. (2007). The plasma glucagon concentration was modeled as function of glucose and amino acid levels.^86,97,98where V_GluGlcn is the maximum glucagon infusion rate, q₁ and p₁ are the weight factor and the rate, respectively. V_AAGlcn is the maximum infusion rate of glucagon due to amino acids and n and K_AA are the corresponding Hill coefficient and half saturation constant. To obtain the plasma concentrations of glucagon, these secretion rates were incorporated into the kinetic model developed by Liu et al. (2009).

The effect of plasma insulin concentration on the blood flow was derived by fitting a Hill equation to the profiles from the literature. The effect of plasma insulin on hepatic blood flow was modeled from the dynamical data reported by Frayn 2003,⁹⁹ wherein the 2.5 fold change in blood flow was reported for a 5 fold change in the plasma insulin levels (see S1 file Fig. M1(D)†).

where, V_max is the maximum rate, Ins is the plasma insulin concentration, n is the Hill coefficient and K_Ins is the Michaelis–Menten constant. The passive and facilitated metabolite transport across the tissue and blood compartment was modeled as per eqn (5) and (6) respectively.where, C_bj and C_cytj are the jth metabolite concentrations in the blood and the cytosol, respectively. ε_j and T_j are the effective permeability issue surface are product and the maximal transport rate of the metabolite across the tissue for passive and facilitated transport, respectively. K_bj and K_cytj are respective saturation constants for blood and cytosolic metabolites for blood tissue transport. The metabolite concentrations in the blood were modeled using the framework as given below.where, C_bj is the jth metabolite concentration in the capillary blood, bld_flw is the blood flow rate to the liver, Ins_bldEff is the effect of the insulin on blood flow, C_aj is the jth metabolite concentration in the arterial blood, Tis_tj is the rate of metabolite transport across the tissue and blood, V_bld is the volume of the capillary blood.

Metabolism module

This module consists of a detailed model of hepatic metabolism that comprises of the central metabolic pathway including glycolysis and gluconeogenesis, glycogen synthesis and breakdown, TCA cycle, oxidative phosphorylation, fatty acid synthesis and oxidation, protein synthesis and breakdown, urea cycle, pentose phosphate pathway, cholesterol biosynthesis and hexose amine pathway. The model for glycolysis, glycogen metabolism and gluconeogenesis was adopted from Konig et al. 2012. The detailed model was developed for lipid and amino acid and lipid metabolism which was further integrated with the existing model for carbohydrate metabolism. The general form of metabolic reactions was written in Michaels–Menten formalism.where M_i is the concentration of the ith metabolite, V_c is the volume of the compartment (cytosol or mitochondria), V_prodj and V_consk is the rate of production and consumption of the ith metabolite, respectively. Tis_t is the transport rate of the metabolite across blood cytosol or cytosol mitochondrial compartment. The production and consumption rates were modeled using the Michaelis–Menten functions as given belowwhere V_maxj is the maximum rate of the jth reaction, is the product of the regulation by the metabolite, signaling and the transcription. M_s,j is the sth metabolite in the jth reaction and Km_s,j is the corresponding saturation constant. A and I are the activators and the inhibitors pertaining to the activatory (Reg_actr,j) or inhibitory (Reg_deactr,j) regulation of the flux, respectively. is the regulation exerted by the signaling and transcriptional networks, wherein sig_acta is the positive regulation by the ath signaling molecule and trans_actb id the positive regulation by the bth transcription factor. Sig_deact and trans_deact are the negative regulations exerted by the signaling and transcription events on the jth flux, respectively.

Modeling metabolic regulation

The regulation of the signaling component on the metabolic enzymes were modeled by assuming parallel activation mechanism wherein, if a signaling/transcription component is known to regulate a enzyme in a certain manner (activation or inhibition), then the subsequent linear pathway was assumed to be correspondingly activated by that signaling/transcription component to ensure the flux balance. Apart from this, the regulations by several signaling/transcription components on a single enzyme was assumed to be by the OR gate for activation effects and by AND gate for inhibitory effects as given in eqn (17). The formalism used for modeling these regulations are as given below. An example of glycolysis regulatory function is illustrated below.

Reg_(Glu(G6p)) = (0.25)(1 + AKT_(Ptvglysis) + SREBP_(Ptvglysis) + AMPK_(Effglysis) + CHREBP_(Ptvglysis))FOXO_Ntv; (17)

The regulations of the metabolic reactions were modeled to modulate the metabolic enzymes. Several signaling and transcription factors are known to regulate metabolism (see Table 2). The influence of various signaling and transcriptions such as activation and deactivation of these enzymes were derived from the dose response data from the literature. The unknown rates were deduced by the fitting the output curve to the desired response and followed by appropriate parameterization. The unknown rates for the metabolic regulation by the signaling pathways were obtained by fitting the pathway rate parameters to the time-course data of plasma metabolite levels (i.e. glucose, amino acid and fatty acids). The rational was to obtain the fold change in the metabolic rates required to obtain the reported experimental profiles for plasma metabolite. These fold changes were translated to the appropriate Hill fits for the effect of signaling endpoints on the metabolic enzymes. From these Hill fits the three parameters V_max, Km and n were deduced, wherein the ‘V_max’ is the maximum fold change required, ‘Km’ the half saturation constant and ‘n’ as the Hill coefficients assumed to be sensitive (n = 2–4).

Table 2 Regulation of hepatic metabolism by metabolites

Reaction enzyme	Positive regulation	Negative regulation
Glucokinase		F6p
Phosphofructokinase	AMP	Citrate
Glycogen phosphorylase	AMP	Glucose
Ga3p dehydrogenase		Glucose
Pyruvate kinase	F16p	Amino acids
Pyruvate dehydrogenase		NADH, ACoA (Pi)
Citrate synthase	AMP
Isocitrate dehydrogenase		SCoA (Pi)
AKG dehydrogenase	AMP	SCoA (Pi)
Citrate shuttle^103		PalCoA (Pi)
Cit_ACoA_OAA (ATP citrate lyase)		PalCoA (Pi)
ACoA_Mal-CoA (acetyl CoA carboxylase)		PalCoA (Pi)
FFA_PalCoA (acyl CoA synthase) (Saggerson, 2008)		Mal-CoA (Pi)
PalCoA_ACoA (β oxidation)		ACoA (Pi)
Carnitine shuttle (carnitine acyltransferase)		Mal-CoA
Gmt_AKG (glutamate dehydrogenase)		FFA
ACoA_Gmt_NAG (N-acetyl glutamate synthase)	Arginine
NH4_Crbphos (carbomyl phosphate synthase)	NAG
Citrulin_Arg (argininosuccinate lyse)	AMP
(Glucosamine 6 phosphate N-acetyl transferase)	FFA	Glnac (Pi)
(N-Acetyl glucosamine pyrophosphorylase)	Glucose
HMG-CoA_Mevl (HMG-CoA reductase)		Mevl (Pi)

---

We have included the pentose phosphate pathway, urea cycle, cholesterol biosynthesis¹⁰⁰ and hexosamine pathways^101,102 along with the central metabolic pathway. While pentose phosphate pathway is the major source of NADPH, urea cycle takes care of the deamination or removal of the ammonia (NH₄) generated while gluconeogenesis and amino acid catabolism, through urea.^51,69 Hexosamine pathway is the indicator of the metabolic status of the cell under nutrient stress. This pathway is composed of the inputs from the derivatives of glucose, amino acids and the fatty acid metabolism. At higher levels of these metabolites, the glucosamine formation are triggered which further is responsible for the glycosylation of the metabolic enzymes. N-Acetyl glucosamine an end product of the hexosamine pathway is the indicator of the metabolic stress in the cell.

Signaling module

This is for the first time in literature, that we have integrated the hormonal signaling (insulin and glucagon) pathway along with the calcium, cAMP and mTOR signaling pathways. These models were adopted different literature sources and integrated together with the appropriate modeling formalisms. The model for insulin signaling was adopted from the Sedaghat et al. (2002) and the glucagon signaling was adopted from Mutalik et al.⁴⁴ and Xu et al.⁴² Insulin and glucagon hormones and the signaling pathways are mutually antagonistic pathways wherein the downstream of insulin signaling inhibits the activation of cAMP i.e. the glucagon signaling component. Similarly the calcium activated DAG increases the phosphorylation of inactivated PKC which further inhibits the insulin signaling through IRS. While AKT and GSK3 acts as major anabolic regulatory signaling component of insulin signaling pathways, cAMP and PKA are the major metabolic regulatory components of the glucagon signaling pathway. Further, AKT and amino acids signal to activate mTOR^104,105 and its downstream S6K that has an inhibition of IRS.^{21–23,89,106} Table 3 lists the feedback regulations in the signaling integrated pathways. The general formalism of modeling the signaling pathways is as given belowwhere, K_synth and K_deg are the basal synthesis and degradation rate of ith signaling protein S, K_phsj and K_dphsk are the phosphorylation and the dephosphorylation rates of the signaling molecule, respectively. R_pregj and R_dpregj are the regulatory interactions of the phosphorylation and dephosphorylation of S, respectively.

Table 3 Regulation of hepatic metabolism by signaling components

Signaling components	Positive regulation	Negative regulation	References
IRS		PTP, PKC, S6K	21
AKT	mTORC2	Glnac, TRB3	107
PKC	DAG, Glnac, FFA		28 and 29
GSK3	PP1, Phk	Cal, PKA, FFA	108
mTOR	Amino acids		109
S6K	Amino acids	AMPK	110
TSC	AMPK	AKT	111 and 112
cAMP	Gprt	PDE3	113 and 114
PKA	cAMP		115
PDE3	AKT	PKA	116

---

The regulatory effects of the signaling endpoints were incorporated in the metabolic reactions, wherein, these regulations were assumed to influence the maximal rate of an enzymatic reaction. The anabolic regulatory effects on the metabolic pathways were mediated by the insulin signaling components and the catabolic effects were mediated by the glucagon signaling components. The appropriate regulatory functions were modeled to integrate the signaling pathways to the metabolic pathways as described in the previous section.

Transcriptional module

The metabolism in liver is known to be regulated by several transcription factors¹¹⁷ such as SREBP^118–120, ChREBP,¹²¹ PPAR (γ,α,β),^122,123 CREB, CEBP,¹²⁴ PGC1, TRB3, FOXO¹²⁵ and AMPK.¹²⁶ Table 4 lists the components that inter-regulate transcriptional factors. Although it is known that the glucose uptake by liver is mediated by GLUT2 which is known to be insulin independent, the expression of GLUT2 is regulated by the insulin dependent transcriptional factor SREBP1c and glucose. SREBP1c is activated in PI3K dependent manner and is responsible for the expression of glucokinase enzyme, a rate limiting step in the glycolysis. Moreover, the expression of glycolytic and lipogenic genes are regulated by the action of SREBP1c, in the liver. Higher glucose levels also triggers the activation of a ChREBP transcription factor i.e. responsible for glucose mediated up regulation of lipogenesis through LPK, ACC and FAS gene transcription. Insulin signaling along with fatty acids activates a transcription factor PPARγ that is responsible for fatty acid transport and triglyceride synthesis in the liver. The catabolic transcriptions are mediated by the glucagon signaling, wherein cAMP activated PKA phosphorylates the transcription factor CREB which induces the transcription of the genes responsible for the enzymes of the gluconeogenesis pathway such as PEPCK, G6Pase and pyruvate carboxylase. CREB further activates the gluconeogenic cofactor PGC1 which increases the expression of the gluconeogenic genes. Another transcription factor activated under low glucose level and triggered by cAMP is CEBPa that regulates the transcription of the genes responsible for the ammonia metabolism i.e. urea cycle under higher protein diets or excessive amino acid breakdown during exercise. PPARα is the transcriptional activator of the fatty acid oxidation which triggers the expression β oxidation enzymes in the liver. FOXO is a metabolic regulatory transcription factor that down regulates glycolysis and influences on the gluconeogenic gene expression under fasting condition. TRB3 is another transcription factor i.e. activated by PPARα in response to the fatty acids and glucagon signaling which further inhibits AKT activation thereby down regulating the effect of insulin signaling. Furthermore, a major regulator of energy homeostasis is AMP activated protein kinase which is activated under energy stress or starvation, due to the changes in the AMP/ATP ratios in the cell. It is a potent transcriptional regulator that down regulates the anabolic pathways such as glycogen synthesis, fatty acid synthesis and protein synthesis.where, T_synth and T_deg are the basal synthesis and degradation rate of ith transcription factor T, T_actj and T_dactk are the activation rates of the expression and degradation of the transcriptional factor, respectively. T_aregp andT_dpregp are the product of regulatory interactions of that activate and deactivate the transcriptional factor T, respectively. A_p and I_p are the activator and inhibitor concentrations, respectively.

Table 4 Regulation of transcriptional factors by signaling components and macronutrients

Transcription factors	Positive regulation	Negative regulation	References
SREBP	S6K, AKT, PKC	cAMP, FOXO, AMPK	119 and 127
ChREBP	Glucose	PKA, AMPK	121 and 128
PPARγ	AKT, FFA	AMPK	122 and 129
PPARα	PKA, FFA, PGC		130
CREB	PKA	AKT	131 and 132
CEBPa	cAMP	PKC	124 and 133
TRB3	PI3K, PKC, PPAR PGC1		134 and 135
PGC1	FOXO, CREB	AKT,	136
FOXO	Glnac, AMPK	AKT, PPARγ	125 and 137
AMPK	AMP	AKT, PKA, ATP	126

---

References

1. V. Chandramouli, K. Ekberg, W. C. Schumann, S. C. Kalhan, J. Wahren and B. R. Landau, Quantifying gluconeogenesis during fasting, Am. J. Physiol.: Endocrinol. Metab., 1997, 273(6 36–6), E1209–E1215.
2. J. Radziuk and S. Pye, Hepatic glucose uptake, gluconeogenesis and the regulation of glycogen synthesis, Diabetes/Metab. Res. Rev., 2001, 17(4), 250–272.
3. X. Chen, N. Iqbal and G. Boden, The effects of free fatty acids on gluconeogenesis and glycogenolysis in normal subjects, J. Clin. Invest., 1999, 103(3), 365–372.
4. S. J. Pilkis and D. K. Granner, Molecular physiology of the regulation of hepatic gluconeogenesis and glycolysis, Annu. Rev. Physiol., 1992, 54, 885–909.
5. H. Kitano, K. Oda, T. Kimura, Y. Matsuoka, M. Csete and J. Doyle, et al., Metabolic syndrome and robustness tradeoffs, Diabetes, 2004, 53(3), S6–S15.
6. B. J. Kudchodkar, M. J. Lee, S. M. Lee, N. M. DiMarco and A. G. Lacko, Effect of dietary protein on cholesterol homeostasis in diabetic rats, J. Lipid Res., 1988, 29(10), 1272–1287.
7. C.-C. Liao, Y.-L. Lin and C.-F. Kuo, Effect of High-Fat Diet on Hepatic Proteomics of Hamsters, J. Agric. Food Chem., 2015, 63(6), 1869–1881.
8. L. Pichon, J.-F. Huneau, G. Fromentin and D. Tomé, A high-protein, high-fat, carbohydrate-free diet reduces energy intake, hepatic lipogenesis, and adiposity in rats, J. Nutr., 2006, 136(5), 1256–1260.
9. D. A. Podolin, Y. Wei and M. J. Pagliassotti, Effects of a high-fat diet and voluntary wheel running on gluconeogenesis and lipolysis in rats, J. Appl. Physiol., 1999, 86(4), 1374–1380.
10. P. Satabin, B. Bois-Joyeux, M. Chanez, C. Y. Guezennec and J. Peret, Effects of long-term feeding of high-protein or high-fat diets on the response to exercise in the rat, Eur. J. Appl. Physiol., 1989, 58(6), 583–590.
11. S. Satapati, N. E. Sunny, B. Kucejova, X. Fu, T. T. He and A. Mendez-Lucas, et al., Elevated TCA cycle function in the pathology of diet-induced hepatic insulin resistance and fatty liver, J. Lipid Res., 2012, 53(6), 1080–1092.
12. C. Schindler and J. P. Felber, Study on the effect of a high fat diet on diaphragm and liver glycogen and glycerides in the rat, Horm. Metab. Res., 1986, 18(2), 91–93.
13. J. Schwarz, D. Tomé, A. Baars, G. J. E. J. Hooiveld and M. Müller, Dietary Protein Affects Gene Expression and Prevents Lipid Accumulation in the Liver in Mice, PLoS One, 2012, 7(10), e47303.
14. N. E. Sunny, E. J. Parks, J. D. Browning and S. C. Burgess, Excessive Hepatic Mitochondrial TCA Cycle and Gluconeogenesis in Humans with Nonalcoholic Fatty Liver Disease, Cell Metab., 2011, 14(6), 804–810.
15. L. P. Bechmann, R. Hannivoort, G. Gerken, G. S. Hotamisligi, M. Trauner and A. Canbay, The interaction of hepatic lipid and glucose metabolism in liver diseases, J. Hepatol., 2012, 56, 952–964.
16. S. Chevalier, S. C. Burgess, C. R. Malloy, R. Gougeon, E. B. Marliss and J. A. Morais, The greater contribution of gluconeogenesis to glucose production in obesity is related to increased whole-body protein catabolism, Diabetes, 2006, 55(3), 675–681.
17. R. J. Perry, V. T. Samuel, K. F. Petersen and G. I. Shulman, The role of hepatic lipids in hepatic insulin resistance and type 2 diabetes, Nature, 2014, 510(7503), 84–91.
18. K. Keane and P. Newsholme, Metabolic Regulation of Insulin Secretion, Vitamins & Hormones, ed. G. Litwack, Academic Press, 2014, ch. 1, pp. 1–33, http://www.sciencedirect.com/science/article/pii/B9780128001745000016.
19. P. Newsholme, V. Cruzat, F. Arfuso and K. Keane, Nutrient regulation of insulin secretion and action, J. Endocrinol., 2014, 221(3), R105–R120.
20. P. Newsholme, C. Gaudel and N. McClenaghan, Nutrient Regulation of Insulin Secretion and β-Cell Functional Integrity, The Islets of Langerhans, ed. M. S. Islam, Springer, Netherlands, 2010, pp. 91–114,  DOI:10.1007/978-90-481-3271-3_6.
21. P. Gual, Y. le Marchand-Brustel and J.-F. Tanti, Positive and negative regulation of insulin signaling through IRS-1 phosphorylation, Biochimie, 2005, 87(1), 99–109.
22. M. Lansard, S. Panserat, E. Plagnes-Juan, I. Seiliez and S. Skiba-Cassy, Integration of insulin and amino acid signals that regulate hepatic metabolism-related gene expression in rainbow trout: role of TOR, Amino Acids, 2010, 39(3), 801–810.
23. M. O. Weickert and A. F. H. Pfeiffer, Signaling mechanisms linking hepatic glucose and lipid metabolism, Diabetologia, 2006, 49(8), 1732–1741.
24. E. Breda and C. Cobelli, Insulin Secretion Rate During Glucose Stimuli: Alternative Analyses of C-Peptide Data, Ann. Biomed. Eng., 2001, 29(8), 692–700.
25. R. S. Sherwin, K. J. Kramer, J. D. Tobin, P. A. Insel, J. E. Liljenquist and M. Berman, et al., A model of the kinetics of insulin in man, J. Clin. Invest., 1974, 53(5), 1481.
26. G. Toffolo, A minimal model of insulin secretion and kinetics to assess hepatic insulin extraction, AJP: Endocrinology and Metabolism, 2005, 290(1), E169–E176.
27. G. Mingrone, Dietary fatty acids and insulin secretion, Food Nutr. Res., 2006, 50(1), 79–84.
28. T. K. T. Lam, H. Yoshii, C. A. Haber, E. Bogdanovic, L. Lam and I. G. Fantus, et al., Free fatty acid-induced hepatic insulin resistance: a potential role for protein kinase C-δ, Am. J. Physiol.: Endocrinol. Metab., 2002, 283(4), E682–E691.
29. S. Pereira, E. Park, Y. Mori, C. A. Haber, P. Han and T. Uchida, et al., FFA-induced Hepatic Insulin Resistance in vivo is mediated by PKC-δ, NADPH Oxidase, and Oxidative Stress, Am. J. Physiol.: Endocrinol. Metab., 2014, E34–E46.
30. D. A. Beard, Thermodynamic-based computational profiling of cellular regulatory control in hepatocyte metabolism, AJP: Endocrinology and Metabolism, 2004, 288(3), E633–E644.
31. P. Blavy, F. Gondret, H. Guillou, S. Lagarrigue, P. Martin and O. Radulescu, et al., A minimal and dynamic model for fatty acid metabolism in mouse liver. In: Journιes Ouvertes de Biologie, Informatique et Mathιmatique (JOBIM) [Internet], 2008, cited 2015 Jul 25, https://hal-agrocampus-ouest.archives-ouvertes.fr/hal-00729742/.
32. D. Calvetti, A. Kuceyeski and E. Somersalo, A mathematical model of liver metabolism: from steady state to dynamic, J. Phys.: Conf. Ser., 2008, 124, 012012.
33. E. Chalhoub, R. W. Hanson and J. M. Belovich, A computer model of gluconeogenesis and lipid metabolism in the perfused liver, AJP: Endocrinology and Metabolism, 2007, 293(6), E1676–E1686.
34. C. de Maria, D. Grassini, F. Vozzi, B. Vinci, A. Landi and A. Ahluwalia, et al., HEMET: Mathematical model of biochemical pathways for simulation and prediction of HEpatocyte METabolism, Comput. Meth. Programs Biomed., 2008, 92(1), 121–134.
35. J. T. Dean, M. L. Rizk, Y. Tan, K. M. Dipple and J. C. Liao, Ensemble Modeling of Hepatic Fatty Acid Metabolism with a Synthetic Glyoxylate Shunt, Biophys. J., 2010, 98(8), 1385–1395.
36. C. Gille, C. Bölling, A. Hoppe, S. Bulik, S. Hoffmann and K. Hübner, et al., HepatoNet1: a comprehensive metabolic reconstruction of the human hepatocyte for the analysis of liver physiology, Mol. Syst. Biol., 2010, 6(411), 1–13.
37. J. Hetherington, T. Sumner, R. M. Seymour, L. Li, M. V. Rey and S. Yamaji, et al., A composite computational model of liver glucose homeostasis. I. Building the composite model, J. R. Soc., Interface, 2012, 9(69), 689–700.
38. M. Konig, S. Bulik and H.-G. Holzhütter, Quantifying the Contribution of the Liver to Glucose Homeostasis: A Detailed Kinetic Model of Human Hepatic Glucose Metabolism, PLoS Comput. Biol., 2012, 8(6), e1002577.
39. M. A. Orman, J. Mattick, I. P. Androulakis, F. Berthiaume and M. G. Ierapetritou, Stoichiometry Based Steady-State Hepatic Flux Analysis: Computational and Experimental Aspects, Metabolites, 2012, 2(4), 268–291.
40. Z. Rausova, J. Chrenova, P. Nuutila, P. Iozzo and L. Dedik, System approach to modeling of liver glucose metabolism with physiologically interpreted model parameters outgoing from [18F]FDG concentrations measured by PET, Comput. Meth. Programs Biomed., 2012, 107(2), 347–356.
41. P. R. Shorten and G. C. Upreti, A mathematical model of fatty acid metabolism and VLDL assembly in human liver, Biochim. Biophys. Acta, Mol. Cell Biol. Lipids, 2005, 1736(2), 94–108.
42. K. Xu, K. T. Morgan, A. Todd Gehris, T. C. Elston and S. M. Gomez, A Whole-Body Model for Glycogen Regulation Reveals a Critical Role for Substrate Cycling in Maintaining Blood Glucose Homeostasis, PLoS Comput. Biol., 2011, 7(12), e1002272.
43. A. R. Sedaghat, A. Sherman and M. J. Quon, A mathematical model of metabolic insulin signaling pathways, Am. J. Physiol.: Endocrinol. Metab., 2002, 283(5), E1084–E1101.
44. V. K. Mutalik and K. V. Venkatesh, Quantification of the glycogen cascade system: the ultrasensitive responses of liver glycogen synthase and muscle phosphorylase are due to distinctive regulatory designs, Theor. Biol. Med. Modell., 2005, 2(1), 19.
45. P. K. U. Vinod and K. V. Venkatesh, Quantification of the effect of amino acids on an integrated mTOR and insulin signaling pathway, Mol. BioSyst., 2009, 5(10), 1163.
46. C. Dalla Man, R. A. Rizza and C. Cobelli, Meal Simulation Model of the Glucose–Insulin System, IEEE Trans. Biomed. Eng., 2007, 54(10), 1740–1749.
47. L. Wang, J. Yu and R. L. Walzem, High-carbohydrate diets affect the size and composition of plasma lipoproteins in hamsters (Mesocricetus auratus), Comp. Med., 2008, 58(2), 151.
48. F. Belfiore and S. Iannello, A formula for quantifying the effects of substrate cycles (futile cycles) on metabolic regulation. Its application to glucose futile cycle in liver as studied by glucose-6-phosphatase/glucokinase determinations, Acta Diabetol. Lat., 1990, 27(1), 71–80.
49. D. B. Jump, D. Botolin, Y. Wang, J. Xu, B. Christian and O. Demeure, Fatty acid regulation of hepatic gene transcription, J. Nutr., 2005, 135(11), 2503–2506.
50. R. J. Rafoth and G. R. Onstad, Urea synthesis after oral protein ingestion in man, J. Clin. Invest., 1975, 56(5), 1170.
51. M. C. G. van de Poll, S. J. Wigmore, D. N. Redhead, R. G. H. Beets-Tan, O. J. Garden and J. W. M. Greve, et al., Effect of major liver resection on hepatic ureagenesis in humans, Am. J. Physiol.: Gastrointest. Liver Physiol., 2007, 293(5), G956–G962.
52. F. Q. Nuttall, A. Ngo and M. C. Gannon, Regulation of hepatic glucose production and the role of gluconeogenesis in humans: Is the rate of gluconeogenesis constant?, Diabetes/Metab. Res. Rev., 2008, 24(6), 438–458.
53. C. Fromenti, D. Azzout-Marniche, M. Stepien, P. Even, G. Fromentin and D. Tomé, et al., Whole body amino acids are candidate precursors of postprandial hepatic neoglycogenogenesis in high protein fed rats, FASEB J., 2009, 23, 738.
54. M. A. Veldhorst, M. S. Westerterp-Plantenga and K. R. Westerterp, Gluconeogenesis and energy expenditure after a high-protein, carbohydrate-free diet, Am. J. Clin. Nutr., 2009, 90(3), 519–526.
55. M. Krssak, A. Brehm, E. Bernroider, C. Anderwald, P. Nowotny and C. Dalla Man, et al., Alterations in postprandial hepatic glycogen metabolism in type 2 diabetes, Diabetes, 2004, 53(12), 3048–3056.
56. M. Roden, Clinical Diabetes Research: Methods and Techniques, John Wiley and Sons, 2007.
57. Ulusoy E., Eren B., Histological changes of liver glycogen storage in mice (Mus musculus) caused by high-protein diets, 2006, cited 2015 Jul 25, Available from, https://digitum.um.es/xmlui/handle/10201/22701.
58. R. Taylor, I. Magnusson, D. L. Rothman, G. W. Cline, A. Caumo and C. Cobelli, et al., Direct assessment of liver glycogen storage by 13C nuclear magnetic resonance spectroscopy and regulation of glucose homeostasis after a mixed meal in normal subjects, J. Clin. Invest., 1996, 97(1), 126.
59. E. Krízová and V. Simek, Effect of intermittent feeding with high-fat diet on changes of glycogen, protein and fat content in liver and skeletal muscle in the laboratory mouse, Physiol. Res., 1996, 45(5), 379–383.
60. O. E. Owen, S. C. Kalhan and R. W. Hanson, The Key Role of Anaplerosis and Cataplerosis for Citric Acid Cycle Function, J. Biol. Chem., 2002, 277(34), 30409–30412.
61. A. Ferramosca, A. Conte, F. Damiano, L. Siculella and V. Zara, Differential effects of high-carbohydrate and high-fat diets on hepatic lipogenesis in rats, Eur. J. Nutr., 2014, 53(4), 1103–1114.
62. K. Petzke, A. Freudenberg and S. Klaus, Beyond the Role of Dietary Protein and Amino Acids in the Prevention of Diet-Induced Obesity, Int. J. Mol. Sci., 2014, 15(1), 1374–1391.
63. K. Nakano and K. Ashida, Effect of Dietary Carbohydrate and Fat on Amino Acid-degrading Enzymes in Relation to Their Protein Sparing Action, J. Nutr., 1969, 100, 208–2016.
64. S. C. Garcia-Caraballo, T. M. Comhair, F. Verheyen, I. Gaemers, F. G. Schaap and S. M. Houten, et al., Prevention and reversal of hepatic steatosis with a high-protein diet in mice, Biochim. Biophys. Acta, Mol. Basis Dis., 2013, 1832(5), 685–695.
65. M. Cahova, H. Dankova, E. Palenickova, Z. Papackova and L. Kazdova, The Opposite Effects of High-Sucrose and High-Fat Diet on Fatty Acid Oxidation and Very Low Density Lipoprotein Secretion in Rat Model of Metabolic Syndrome, J. Nutr. Metab., 2012, 2012, 1–10.
66. G. Silbernagel, D. Lütjohann, J. Machann, S. Meichsner, K. Kantartzis and F. Schick, et al., Cholesterol Synthesis Is Associated with Hepatic Lipid Content and Dependent on Fructose/Glucose Intake in Healthy Humans, Exp. Diabetes Res., 2012, 2012, 1–7.
67. F. Raymond, L. Wang, M. Moser, S. Metairon, R. Mansourian and M.-C. Zwahlen, et al., Consequences of Exchanging Carbohydrates for Proteins in the Cholesterol Metabolism of Mice Fed a High-fat Diet, PLoS One, 2012, 7(11), e49058.
68. S. D. Clarke, D. R. Romsos, A. C. Tsai, P. S. Belo and W. B. A. A. Leveille, Studies on-the Effect of Dietary Cholesterol on Hepatic Protein Synthesis, Reduced Glutathione Levels and Serine Dehydratase Activity in the Rat1, J. Nutr., 1976, 106, 94–102.
69. R. J. Rafoth and G. R. Onstad, Urea synthesis after oral protein ingestion in man, J. Clin. Invest., 1975, 56(5), 1170.
70. M. Wiechetek, G. Breves and H. Höller, Effects of increased blood ammonia concentrations on the concentrations of some metabolites in rat tissues, Q. J. Exp. Physiol., 1981, 66(4), 423–429.
71. A. D. Cherrington and B. Lecture, Control of glucose uptake and release by the liver in vivo, Diabetes, 1999, 48, 1198–1214.
72. J. Girard, The Inhibitory Effects of Insulin on Hepatic Glucose Production Are Both Direct and Indirect, Diabetes, 2006, 55(2), S65–S69.
73. J. I. Baum, D. K. Layman, G. G. Freund, K. A. Rahn, M. T. Nakamura and B. E. Yudell, A reduced carbohydrate, increased protein diet stabilizes glycemic control and minimizes adipose tissue glucose disposal in rats, J. Nutr., 2006, 136(7), 1855–1861.
74. C. Brøns, C. B. Jensen, H. Storgaard, N. J. Hiscock, A. White and J. S. Appel, et al., Impact of short-term high-fat feeding on glucose and insulin metabolism in young healthy men: High-fat feeding in young healthy men, J. Physiol., 2009, 587(10), 2387–2397.
75. P. H. Bisschop, J. de Metz, M. T. Ackermans, E. Endert, H. Pijl and F. Kuipers, et al., Dietary fat content alters insulin-mediated glucose metabolism in healthy men, Am. J. Clin. Nutr., 2001, 73(3), 554–559.
76. E. Henkel, M. Menschikowski, C. Koehler, W. Leonhardt and M. Hanefeld, Impact of glucagon response on postprandial hyperglycemia in men with impaired glucose tolerance and type 2 diabetes mellitus, Metabolism, 2005, 54(9), 1168–1173.
77. V. E. de Meijer, H. D. Le, J. A. Meisel, M. R. A. Sharif, A. Pan and V. Nosé, et al., Dietary fat intake promotes the development of hepatic steatosis independently from excess caloric consumption in a murine model, Metabolism, 2010, 59(8), 1092–1105.
78. P. Nguyen, V. Leray, M. Diez, S. Serisier, J. L. Bloc'h and B. Siliart, et al., Liver lipid metabolism, J. Anim. Physiol. Anim. Nutr., 2008, 92(3), 272–283.
79. D. G. Mashek, Hepatic Fatty Acid Trafficking: Multiple Forks in the Road, Advances in Nutrition: An International Review Journal, 2013, 4(6), 697–710.
80. K. F. Leavens and M. J. Birnbaum, Insulin signaling to hepatic lipid metabolism in health and disease, Crit. Rev. Biochem. Mol. Biol., 2011, 46(3), 200–215.
81. V. T. Samuel, K. F. Petersen and G. I. Shulman, Lipid-induced insulin resistance: unravelling the mechanism, Lancet, 2010, 375(9733), 2267–2277.
82. F. R. Jornayvaz and G. I. Shulman, Diacylglycerol Activation of Protein Kinase Cε and Hepatic Insulin Resistance, Cell Metab., 2012, 15(5), 574–584.
83. A. L. Birkenfeld and G. I. Shulman, Nonalcoholic fatty liver disease, hepatic insulin resistance, and type 2 diabetes, Hepatology, 2014, 59(2), 713–723.
84. E. Chalhoub, L. Xie, V. Balasubramanian, J. Kim and J. Belovich, A distributed model of carbohydrate transport and metabolism in the liver during rest and high-intensity exercise, Ann. Biomed. Eng., 2007, 35(3), 474–491.
85. P. R. Somvanshi and K. V. Venkatesh, A conceptual review on systems biology in health and diseases: from biological networks to modern therapeutics, Syst. Biol. Synth. Biol., 2014, 8(1), 99–116.
86. J. A. Calbet and D. A. MacLean, Plasma glucagon and insulin responses depend on the rate of appearance of amino acids after ingestion of different protein solutions in humans, J. Nutr., 2002, 132(8), 2174–2182.
87. J. C. Floyd, S. S. Fajans, S. Pek, C. A. Thiffault, R. F. Knopf and J. W. Conn, Synergistic effect of essential amino acids and glucose upon insulin secretion in man, Diabetes, 1970, 19(2), 109–115.
88. L. J. van Loon, W. H. Saris, H. Verhagen and A. J. Wagenmakers, Plasma insulin responses after ingestion of different amino acid or protein mixtures with carbohydrate, Am. J. Clin. Nutr., 2000, 72(1), 96–105.
89. E. Nyman, C. Brannmark, R. Palmer, J. Brugard, F. H. Nystrom and P. Stralfors, et al., A Hierarchical Whole-body Modeling Approach Elucidates the Link between in vitro Insulin Signaling and in vivo Glucose Homeostasis, J. Biol. Chem., 2011, 286(29), 26028–26041.
90. R. K. Dash, Y. Li, K. Jaeyeon, G. M. Saidel and M. E. Cabrera, Modeling Cellular Metabolism and Energetics in Skeletal Muscle: Large-Scale Parameter Estimation and Sensitivity Analysis, IEEE Trans. Biomed. Eng., 2008, 55(4), 1298–1318.
91. K. Keane and P. Newsholme, Metabolic Regulation of Insulin Secretion, Vitamins & Hormones, ed. G. Litwack, Academic Press, 2014, ch. 1, pp. 1–33, http://www.sciencedirect.com/science/article/pii/B9780128001745000016.
92. A. V. Matveyenko, D. Liuwantara, T. Gurlo, D. Kirakossian, C. Dalla Man and C. Cobelli, et al., pulsatile Portal Vein Insulin Delivery Enhances Hepatic Insulin Action and Signaling, Diabetes, 2012, 61(9), 2269–2279.
93. C. Gravena, P. Mathias and S. Ashcroft, Acute effects of fatty acids on insulin secretion from rat and human islets of Langerhans, J. Endocrinol., 2002, 173(1), 73–80.
94. M. Manco, M. Calvani and G. Mingrone, Effects of dietary fatty acids on insulin sensitivity and secretion. Diabetes, obesity and metabolism, Diabetes, Obes. Metab., 2004, 6, 402–413.
95. V. González-Vélez, G. Dupont, A. Gil, A. González and I. Quesada, Model for Glucagon Secretion by Pancreatic α-Cells, PLoS One, 2012, 7(3), e32282.
96. E. K. Ainscow and M. D. Brand, The responses of rat hepatocytes to glucagon and adrenaline, Eur. J. Biochem., 1999, 265(3), 1043–1055.
97. A. S. Pagliara, S. N. Stillings, M. W. Haymond, B. A. Hover and F. M. Matschinsky, Insulin and glucose as modulators of the amino acid-induced glucagon release in the isolated pancreas of alloxan and streptozotocin diabetic rats, J. Clin. Invest., 1975, 55(2), 244.
98. R. Henningsson and I. Lundquist, Arginine-induced insulin release is decreased and glucagon increased in parallel with islet NO production, Am. J. Physiol.: Endocrinol. Metab., 1998, 275(3), E500–E506.
99. K. N. Frayn, Metabolic Regulation: A Human Perspective, 2010.
100. S. Watterson, M. L. Guerriero, M. Blanc, A. Mazein, L. Loewe and K. A. Robertson, et al., A model of flux regulation in the cholesterol biosynthesis pathway: Immune mediated graduated flux reduction versus statin-like led stepped flux reduction, Biochimie, 2013, 95(3), 613–621.
101. J. A. Hanover, M. W. Krause and D. C. Love, The hexosamine signaling pathway: O-GlcNAc cycling in feast or famine, Biochim. Biophys. Acta, Gen. Subj., 2010, 1800(2), 80–95.
102. Iadicicco C., Mechanisms linking glucotoxicity to the development of insulin resistance: a role for the endoplasmic reticulum stress [Internet], Università degli studi di Napoli Federico II; 2010, cited 2014 Apr 29, Available from: http://www.fedoa.unina.it/7982/.
103. M. L. Halperin, B. H. Robinson and I. B. Fritz, Effects of palmitoyl CoA on citrate and malate transport by rat liver mitochondria, Proc. Natl. Acad. Sci. U. S. A., 1972, 69(4), 1003–1007.
104. K. Inoki, Y. Li, T. Zhu, J. Wu and K.-L. Guan, TSC2 is phosphorylated and inhibited by AKT and suppresses mTOR signaling, Nat. Cell Biol., 2002, 4(9), 648–657.
105. M. Laplante and D. M. Sabatini, mTOR Signaling in Growth Control and Disease, Cell, 2012, 149(2), 274–293.
106. S. Li, M. S. Brown and J. L. Goldstein, Bifurcation of insulin signaling pathway in rat liver: mTORC1 required for stimulation of lipogenesis, but not inhibition of gluconeogenesis, Proc. Natl. Acad. Sci. U. S. A., 2010, 107(8), 3441–3446.
107. P. Dalle Pezze, A. G. Sonntag, A. Thien, M. T. Prentzell, M. Godel and S. Fischer, et al., A dynamic network model of mTOR signaling reveals TSC-independent mTORC2 regulation, Sci. Signaling, 2012, 5(217), ra25.
108. T. Sumner, Sensitivity analysis in Systems biology modelling And its application to a Multi-scale model of blood Glucose homeostasis, Univ. College London, 2010.
109. S. Wullschleger, R. Loewith and M. N. Hall, TOR Signaling in Growth and Metabolism, Cell, 2006, 124(3), 471–484.
110. F. Tremblay and A. Marette, Amino acid and insulin signaling via the mTOR/p70 S6 kinase pathway A negative feedback mechanism leading to insulin resistance in skeletal muscle cells, J. Biol. Chem., 2001, 276(41), 38052–38060.
111. R. J. Shaw, LKB1 and AMP-activated protein kinase control of mTOR signaling and growth, Acta Physiol., 2009, 196(1), 65–80.
112. B. D. Manning, M. N. Logsdon, A. I. Lipovsky, D. Abbott, D. J. Kwiatkowski and L. C. Cantley, Feedback inhibition of AKT signaling limits the growth of tumors lacking Tsc2, Genes Dev., 2005, 19(15), 1773–1778.
113. M. S. Hanson, A. H. Stephenson, E. A. Bowles and R. S. Sprague, Insulin inhibits human erythrocyte cAMP accumulation and ATP release: role of phosphodiesterase 3 and phosphoinositide 3-kinase, Exp. Biol. Med., 2010, 235(2), 256–262.
114. M. Rendell, Y. Salomon, M. C. Lin, M. Rodbell and M. Berman, The hepatic adenylate cyclase system. III. A mathematical model for the steady state kinetics of catalysis and nucleotide regulation, J. Biol. Chem., 1975, 250(11), 4253–4260.
115. Q. Ni, A. Ganesan, N.-N. Aye-Han, X. Gao, M. D. Allen and A. Levchenko, et al., Signaling diversity of PKA achieved via a Ca²⁺-cAMP-PKA oscillatory circuit, Nat. Chem. Biol., 2011, 7(1), 34–40.
116. K. Omori and J. Kotera, Overview of PDEs and Their Regulation, Circ. Res., 2007, 100(3), 309–327.
117. B. Desvergne, Transcriptional Regulation of Metabolism, Phys. Rev., 2006, 86(2), 465–514.
118. K. J. Nadeau, Insulin Regulation of Sterol Regulatory Element-binding Protein-1 Expression in L-6 Muscle Cells and 3T3 L1 Adipocytes, J. Biol. Chem., 2004, 279(33), 34380–34387.
119. M. Feleischmann and P. Iynedjian, Regulation of sterol regulatory-element binding protein 1 gene expression in liver: role of insulin and protein kinase B/cAKT, Biochem. J., 2000, 349, 13–17.
120. M. Foretz, C. Guichard, P. Ferré and F. Foufelle, Sterol regulatory element binding protein-1c is a major mediator of insulin action on the hepatic expression of glucokinase and lipogenesis-related genes, Proc. Natl. Acad. Sci. U. S. A., 1999, 96(22), 12737–12742.
121. K. Iizuka, R. K. Bruick, G. Liang, J. D. Horton and K. Uyeda, Deficiency of carbohydrate response element-binding protein (ChREBP) reduces lipogenesis as well as glycolysis, Proc. Natl. Acad. Sci. U. S. A., 2004, 101(19), 7281–7286.
122. B. König, A. Koch, J. Spielmann, C. Hilgenfeld, F. Hirche and G. I. Stangl, et al., Activation of PPARα and PPARγ reduces triacylglycerol synthesis in rat hepatoma cells by reduction of nuclear SREBP-1, Eur. J. Pharmacol., 2009, 605(1–3), 23–30.
123. M. Rakhshandehroo, L. M. Sanderson, M. Matilainen, R. Stienstra, C. Carlberg and P. J. de Groot, et al., Comprehensive Analysis of PPARα-Dependent Regulation of Hepatic Lipid Metabolism by Expression Profiling, PPAR Res., 2007, 2007, 1–13.
124. W. J. Roesler, The role of C/EBP in nutrient and hormonal regulation of gene expression, Annu. Rev. Nutr., 2001, 21(1), 141–165.
125. D. N. Gross, A. P. J. van den Heuvel and M. J. Birnbaum, The role of FoxO in the regulation of metabolism, Oncogene, 2008, 27(16), 2320–2336.
126. D. Carling, The AMP-activated protein kinase cascade – a unifying system for energy control, Trends Biochem. Sci., 2004, 29(1), 18–24.
127. J. L. Owen, Y. Zhang, S.-H. Bae, M. S. Farooqi, G. Liang and R. E. Hammer, et al., Insulin stimulation of SREBP-1c processing in transgenic rat hepatocytes requires p70 S6-kinase, Proc. Natl. Acad. Sci. U. S. A., 2012, 109(40), 16184–16189.
128. F. Benhamed, P.-D. Denechaud, M. Lemoine, C. Robichon, M. Moldes and J. Bertrand-Michel, et al., The lipogenic transcription factor ChREBP dissociates hepatic steatosis from insulin resistance in mice and humans, J. Clin. Invest., 2012, 122(6), 2176–2194.
129. E. Morán-Salvador, M. López-Parra, V. García-Alonso, E. Titos, M. Martínez-Clemente and A. González-Périz, et al., Role for PPARγ in obesity-induced hepatic steatosis as determined by hepatocyte- and macrophage-specific conditional knockouts, FASEB J., 2011, 25(8), 2538–2550.
130. P. D. McMullen, S. Bhattacharya, C. G. Woods, B. Sun, K. Yarborough and S. M. Ross, et al., A map of the PPARα transcription regulatory network for primary human hepatocytes, Chem.-Biol. Interact., 2014, 209(1), 14–24.
131. L. Everett, J. Lay, S. Lukovac, D. Bernstein, D. Steger and M. Lazar, et al., Integrative genomic analysis of CREB defines a critical role for transcription factor networks in mediating the fed/fasted switch in liver, BMC Genomics, 2013, 14(1), 337.
132. S. Herzig, F. Long, U. S. Jhala, S. Hedrick, R. Quinn and A. Bauer, et al., CREB regulates hepatic gluconeogenesis through the coactivator PGC-1, Nature, 2001, 413(6852), 179–183.
133. N. Wang, M. Finegold, A. Bradley, C. Ou, S. Abdelsayed and M. Wilde, et al., Impaired energy homeostasis in C/EBP alpha knockout mice, Science, 1995, 269(5227), 1108–1112.
134. R. Matsushima, NHN. Effect of TRB3 on Insulin and Nutrient-stimulated Hepatic p70 S6 Kinase Activity, J. Biol. Chem., 2006, 281, 29719–29729.
135. K. D. Keyong Du, TRB3-A Tribble Homolog That Inhibits AKT/PKB Activation by Insulin in Liver, Science, 2003, 300, 1574–1577.
136. S.-H. Koo, H. Satoh, S. Herzig, C.-H. Lee, S. Hedrick and R. Kulkarni, et al., PGC-1 promotes insulin resistance in liver through PPAR-α-dependent induction of TRB-3, Nat. Med., 2004, 10(5), 530–534.
137. N. Hay, Interplay between FOXO, TOR, and AKT, Biochim. Biophys. Acta, Mol. Cell Res., 2011, 1813(11), 1965–1970.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ra18128c

---