- Research article
- Open Access
Long-term prediction of fish growth under varying ambient temperature using a multiscale dynamic model
© Bar and Radde; licensee BioMed Central Ltd. 2009
- Received: 13 March 2009
- Accepted: 10 November 2009
- Published: 10 November 2009
Feed composition has a large impact on the growth of animals, particularly marine fish. We have developed a quantitative dynamic model that can predict the growth and body composition of marine fish for a given feed composition over a timespan of several months. The model takes into consideration the effects of environmental factors, particularly temperature, on growth, and it incorporates detailed kinetics describing the main metabolic processes (protein, lipid, and central metabolism) known to play major roles in growth and body composition.
For validation, we compared our model's predictions with the results of several experimental studies. We showed that the model gives reliable predictions of growth, nutrient utilization (including amino acid retention), and body composition over a timespan of several months, longer than most of the previously developed predictive models.
We demonstrate that, despite the difficulties involved, multiscale models in biology can yield reasonable and useful results. The model predictions are reliable over several timescales and in the presence of strong temperature fluctuations, which are crucial factors for modeling marine organism growth. The model provides important improvements over existing models.
- Body Composition
- Atlantic Salmon
- Protein Metabolism
- Pyruvate Carboxylase
- Short Timescale
Efficient and accurate prediction of whole body growth is very important in animal husbandry, particularly in marine aquaculture, a rapidly developing and important source of food. Feed development is largely dependent on accurate growth prediction, and it has significant economical and environmental consequences. However, the growth of living organisms is a complex process, affected by both external and biotic factors. Thus, organism growth is extremely difficult to predict, particularly over long periods of time. Mathematical models of varying complexity often serve as tools for the prediction and simulation of animal growth [1, 2], although usually with a short prediction horizon of several days (see, for instance, [3, 4]). Earlier attempts to describe whole body growth were mainly conducted for the design of feed and were mostly based on empirical functions (see, for instance, [5–7]). These models were largely based on data sets derived from multiple experiments and were valid only under specific conditions. Thus, they could only be applied with confidence within a relatively small range of conditions. For example, growth in pigs, cattle, and chickens has been addressed using several dynamic models developed in recent years [8, 9]. These simple models effectively predicted growth and body composition under a narrow range of environmental conditions. A few mechanistic models were also developed to simulate the growth of fish [3, 10, 11]. These models did not attempt to describe protein metabolism or the regulatory mechanisms of the animal, and they could not predict features such as amino acid (AA) retention, an important nutritional and economic property. Moreover, the models did not take into account the dependence of protein metabolism on environmental factors, although these have been shown to have a strong impact on marine organisms. In particular, fluctuations in temperature strongly influence marine organisms. Most marine organisms are ectothermal, and their growth response is highly sensitive to variations in water temperature . Temperature affects protein synthesis in a linear manner [13–15]; however, despite a large number of studies, temperature's effect on protein degradation is still unknown, probably due to the difficulty of measuring in vivo rates of protein degradation . This temperature-degradation rate relationship is critical to any study of the growth of aquatic life forms.
Complex regulatory networks affect most inter- and intracellular processes, including cell growth . Nutrients from the feed intake are mobilized and utilized de facto on many levels of regulation, which also respond to external stimuli. Regardless of their framework, earlier predictive models incorporated neither cell regulation nor environmental effects, which are naturally not multiscaled. Conversely, structured mathematical models that do incorporate regulations recurrently have elucidated the link between the processes that affect homeostasis and growth . Such a framework also has practical applications due to the potential for the accurate prediction of growth and tissue composition, even in the presence of temperature variations.
The concept of the model presented here is based on a previous model of fish growth . Our research attempts to integrate the complete set of bodily functions (mainly growth), with inter- and intracellular regulation, to provide a multiscaled predictive tool for modeling fish in aquaculture rearing conditions. Our first-principles modeling approach includes the metabolic and regulative networks that are believed to have a large impact on growth. These are divided into functional compartments, namely, protein, lipid, and central metabolic processes, that are interconnected by mass flow. The input variables of the model are determined by the feeding process. Moreover, the output variables, i.e., the quantity of body lipids, proteins, carbohydrates, oxygen, CO2, and NH3, are all measurable (in principle), providing an experimental basis for model validation. The model setup was data-driven; the level of detail was chosen such that the values for model parameters were either known from the literature, or they could be estimated from available experimental data. In other words, almost all model parameters were identifiable. The parameters that had not been reported in the literature or that could not be estimated from experiments, particularly parameters involved in intracellular processes, were mainly estimated from data for other species of fish, or from mammal or yeast cell data. These processes (such as the tricarboxylic acid (TCA) cycle and AA catabolism) are, however, fundamentally similar in most species. As a consequence, like other models developed for similar purposes (such as the model for antibody production, see, e.g., [20, 21]), the overall model includes mechanistic descriptions at the level of chemical reactions (called a 'structured model' in ) as well as phenomenological functions (such as Michaelis-Menten and Hill-type kinetics describing cooperative enzymes), summarizing complex regulatory pathways that are not known in detail. Ho et al.  refers to the latter as an 'unstructured model'.
Here, we extended the model in  in several respects. First, we included the experimentally reported temperature dependence of protein metabolism (protein synthesis and degradation) in the modeling framework. To the best of our knowledge, the effect of temperature was modeled relatively accurately for the first time in this type of model. We approximate the relationship between temperature and protein degradation in fish as a parabolic function. Second, we integrated several feedback control mechanisms, such as ATP-regulated fluxes in the TCA submodel, to account for the maintenance of constant energy ATP/ADP ratios. Third, we accounted for the channeling of AAs produced by protein degradation into re-synthesis in the protein metabolism cycle. Homeostasis in the previous model was achieved by two functions, lumping together the complete regulatory mechanism of the TCA cycle. This, however, was highly inaccurate, and resulted in a large discrepancy in the rapid dynamics that produced strong oscillations in the long-term dynamics (see Figure 5 in ). To counter the problem of strong oscillations, the model was adapted to include most of the known regulatory mechanisms in the TCA cycle , particularly the effect of the α-ketoglutarate dehydrogenase and the pyruvate carboxylase enzymes. Moreover, several regulation mechanisms involved in protein metabolism were included in the new model. They allowed for a better description of the AA pathways, improving both long-term (muscle buildup) and short-term (energy accumulation due to AA breakdown) predictions. These modifications resulted in increased accuracy of the short timescale dynamics, and, consequently, more stable long-term predictions. We provide comparisons with experimental data on the long timescale of several months for overall growth and body composition, and on a shorter time scale of hours for the regulation of energy balance and AA dynamics.
The simulation results show that our model accurately predicts growth and body composition over a long time period of several months. Furthermore, despite the difficulties involved, we demonstrate that multiscale models are able to efficiently predict many dynamic properties for a wide variety of conditions (for instance, AA retention and the effect of AA profiles, fatty acids (FAs), and glucose on growth and body composition).
In comparison to the previous model, the refined model gives qualitatively and quantitatively more accurate predictions of whole body growth, especially under widely fluctuating temperature conditions. Due to improved regulation on the short timescale (i.e., the TCA cycle rate), energy balance is much better captured, which is also reflected in better and more stable predictions on the long time scale. As a consequence, we were able to make accurate predictions over a timespan of 5-6 months, whereas the old model only gave useful results for a span of less than two months. We also observed an improvement in the prediction of AA retention. Finally, we mention here that marine fish appear to be an important future source of food. Thus, small increases in the accuracy of predictions, by only a few percent, can already assist in finding optimal feeding strategies that will have large economical and environmental consequences.
The long timescale: Mass and body composition
The short timescale-Energy balance and amino acid dynamics
Figure 4 shows the behavior of ATP and acetyl-CoA (AcCoA) for the second experiment over 120 days (as explained in Figure 3). ATP concentrations were highly unstable in the old model, oscillating between values of 0 and 40 μ mol/g, probably due to faulty regulation of the TCA cycle. Energy balance was much improved in the new model, with ATP values that varied slightly within tight bounds of 7-9 μ mol/g. This value was slightly higher than the data reported in several experiments [24–26], but the model was not parametrized under the exact conditions in which these experiments were conducted. The response of AcCoA was greatly improved between the models, with smaller oscillations after each feeding instance (note the logarithmic scale). Data on mitochondrial AcCoA from experiments with fish under similar conditions was not available.
These discrepancies could have several sources. For example, the AA requirement profiles of the two species (salmon and trout) that were compared for protein building were different. These profiles are inserted into the model in the profile filter of the protein metabolism model (see the Protein Metabolism section). Unfortunately, we do not have the time-series data available for salmon. Another source of the discrepancy could be the composition and consistency of the diets used in the experiments. These diets can vary with respect to protein type (for instance the protein casein versus free AA diets, as in , or the commercial diet versus crystalline AA diets ), fat, and carbohydrates, as well as with respect to the processing to which the ingredients were subjected . These differences (diet ingredients, formulation, and processing) have a large effect on the digestion and absorption of the nutrients in the gut, particularly on the AAs. It is likely that the concerted effect of these unmodeled factors resulted in the observed experiment-to-experiment variation. Further investigation is necessary.
We also mention that the predicted peak values occurred earlier (5-10 hours) than those reported in the experiments (12 h) for almost all AAs. Two factors could explain this deviation: Nearly all fish in the experiments were force-fed and, therefore, stressed, which has a strong impact on the absorption and digestion rates of nutrients . Additionally, although the metabolic model presented here may have been accurate, the digestive model may have been too simple. The simplified digestive model used here did not account for many of the interactions between nutrients, e.g. competition of the AAs for the carriers during absorption, rate of digestion of the diet, nutrient breakdown in the gut due to enzyme interactions, density of the commercial diets as a result of processing, the quantity and type of lipids in the diet, and the general effects of lipids on absorption. All of these unmodeled factors (and other potential factors, listed in [12, 30]) may have resulted in earlier AA absorption, and, consequently, earlier simulated peaks compared to the experiments. A more comprehensive model of the digestive system is under preparation in order to increase the accuracy of the AA metabolism model.
Although not originally in the scope of this model, the short-term predictions (AA, ATP, ADP, and TCA components) were reasonably captured, in spite of the unmodeled dynamics. Because short-term and long-term predictions are interconnected (e.g. ATP regulates growth, and protein buildup is linked to the AA metabolism), a serious deviation in either of the timescales would be a matter of concern, as it would lead to faulty predictions on both timescales. By capturing the short timescale correctly, the model assures the credibility of the longer timescale.
Effect of temperature
Amino acid retention
The multiscale model presented here allowed prediction of growth, body composition, and AA retention under varying temperatures for a period of several months, with relatively low deviation from the experimentally measured quantities. First-principles (mechanistic) modeling, combined with empirical observations, appears to be crucial to achieving successful predictions over long periods. Earlier models could indeed predict growth over some timescales [5, 32, 33], but the predictions were almost always limited to only a few properties for each model, as is common in static models, such as the regression and bioenergetic models . Even dynamical models that predicted several properties simultaneously [3, 4, 19] could not guarantee accurate predictions for more than several weeks. Our model was able to predict over a long timespan and under variable temperatures. In this model, energy regulated most molecular processes and the flux balance between metabolic compartments (protein, lipid, and central metabolism). Maintaining homeostasis (constant steady-state ATP levels) in the model, via the regulation of metabolite fluxes from one compartment to another, assisted in the accurate prediction of growth and yielded better prediction of TAG and protein levels in tissue.
Temperature has a large impact on animal growth , but, to the best of our knowledge, most existing models (particularly bioenergetic models) that consider temperature only do so for a constant temperature (see, for instance, earlier dynamic models [5, 19]). The complex effect of temperature on the various processes in the body, and, in particular, on the protein degradation process, is mostly obscured and cannot always be measured , rendering the effect of temperature on growth difficult to model. Due to its dynamical characteristics, the model proposed in this study can account for instantaneous changes in the temperature, down to a resolution of seconds. Therefore, it can be applied to systems with considerable temperature fluctuations, such as those which naturally occur when the seasons change. In the model presented here, we used the measured effect of water temperature on protein synthesis and empirical data on temperature's concerted effect on growth to estimate the effect of temperature on protein degradation and lipid metabolism. The simulation results were reassuring: the simulated total growth correlated well with the growth data from experiments, and the effect of protein turnover was accurately captured (see Figure 6). This indicated that the concerted effect of temperature, its effect on the individual processes (e.g. the turnover process), and particularly on the protein degradation were all correctly modeled. It was suggested here that temperature influences protein degradation in a parabolic manner, as described by equation (7). This hypothesis was reasonable because the consensus of experimental data (reported in the literature) suggests that protein synthesis is linearly affected by temperature [13, 15, 34], and growth has a bell-shaped dependency [12, 15]. However, we assumed that temperature linearly affected other processes (lipid and central metabolism), which is not necessarily the case. Nevertheless, given that more than 80% of fish mass consists of protein and water , the remaining processes are not expected to exert a large perturbation on the overall growth model. For comparison, a linear protein degradation-temperature dependence (as modeled in ) appeared to yield a faulty response (Figure 6a). We emphasize that the parabolic dependence of degradation on temperature is an interesting result in its own right, which might motivate further experimental investigations.
The model described here is composed of processes that occur on very different timescales. Concentrations of TCA intermediates change within seconds and minutes (for instance Acetyl-CoA in Figure 4). AA concentrations (Figure 5), which rise rapidly within minutes of nutrient uptake, exhibit fast dynamics in the model. A large decrease in ATP concentration, within hours of feeding, occurs due to the large consumption of ATP during protein synthesis. ATP levels relax back nearly to steady-state conditions, by way of several regulatory mechanisms, after several hours (Figure 4). The consumption of AAs in response to feeding occurs on a timescale of hours. In contrast, the buildup of protein tissue and TAG reserves occurs over days and weeks, and considerable changes to the progress of growth occur over a timescale of many days or even months (Figure 1). Thus, all timescales within this wide range are relevant to model development, implying a variety of challenges for model assessment. First, errors in the short timescales of seconds and minutes can rapidly accumulate and render the long-term growth prediction unreliable, as we have seen in the previous simulations (Figures 3 and 4). Errors in the short timescales should be limited to ensure valid long timescale predictions, but the relevant intracellular variables are usually difficult to measure. Second, dimension reduction methods via separation of timescales, such as quasi-steady-state or rapid equilibrium approximations, probably fail here, because the relevant timescales vary over a wide range (Figures 4 and 5) and cannot be separated. The time needed for the ATP/ADP ratio to return to its steady-state may be longer than the time between two feeding scenarios (Figure 4), which can cause growing oscillations on the long timescale (Figure 1). Hence, we cannot apply a quasi-steady-state approximation for ATP. Third, interrelated processes with large differences in timescales usually translate into stiff differential equations, which require advanced numerical integration methods. This is also currently a limitation of the model. Because we found that implicit methods implemented in MATLAB failed here, we used a fixed step size that was manually adapted to the fast timescale, making the simulation calculations rather slow. Moreover, stiffness can lead to a variety of problems for methods such as bifurcation or sensitivity analysis. These issues constitute only some of the difficulties inherent in solving multiscale problems. Despite these challenges, omitting any of the regulatory mechanisms incorporated at different timescales (at the cellular and tissue levels) indeed degraded the prediction accuracy. We concluded, therefore, that multiscale modeling is often important for accurate predictions of complex systems.
We did not make use of any potentially available real-time measurements during long period experiments. Because fish in rearing conditions grow over a period of four years, we expect a significant discrepancy in the results, particularly because the feeding regime and composition changes during this time. Unmodeled dynamics, such as the effect of aging, are also a major concern for discrepancies when simulating long timespans (years). However, it is possible to increase the accuracy of the prediction by acquiring measurements of several variables (NH3, CO2 excretion, O2 consumption) and use this data to correct for discrepancies along the way. This procedure is equivalent to applying feedback from measurements in engineering control theory. It will also allow re-optimization of the feed to adjust for unmodeled variations due to, for instance, water salinity, stress in the rearing facility, and change in water pH values.
To conclude, the multiscale model presented here is able to predict growth, body composition, and AA retention relatively accurately over several months. Predictions hold even when strong temperature fluctuations occur, a key property for any model that is applied to the design of feeding experiments. We suggest that the protein degradation process in fish has a parabolic shaped dependency on temperature, but this hypothesis requires further investigation.
Our model is founded on a previous work described in . The main components of the model are the nutrients that enter the metabolic processes as AAs, glucose and fatty acids, the proteins composed of twenty AA, lipids, several metabolites of the central metabolism, and energy. In the following, we give an overview of the model, and the new components are described in detail.
Feed is transformed into a model input using the digestive model described in . According to this model, digestion has a smoothing effect by translating the discontinuous feeding process into an input function that is continuous over time. The nutrients absorbed in the digestive system are transformed in the various organism compartments to other metabolites or to energy. It is assumed that body mass is mainly composed of proteins, TAG in the adipose cells, water, and minerals, and growth is modeled in the same manner as described in .
Free amino acids (AA) concentration
Amino acids channeled to protein synthesis
Amino acids channeled to catabolism (break-down)
Tissue protein, given by its amino acids composition
Acetyl-CoA in the cytosol
Fatty acids (palmitate)
Citrate concentrations in the mitochondria
Isocitrate concentrations in the mitochondria
α-ketoglutrate concentrations in the mitochondria
Succinate concentrations in the mitochondria
Malate concentrations in the mitochondria
Oxaloacetate concentrations in the mitochondria
Malate concentrations in the cytosol
Oxaloacetate concentrations in the cytosol
Glucose concentrations (glycolysis process)
glucose-6-phosphate concentrations (glycolysis)
6-phosphogluconate concentrations (glycolysis)
glyceraldehyde 3-phospate concentrations (glycolysis)
The nutrients FA, glucose and 20 amino acids consumed
The output of the model, consists of protein (sum of its amino acids), storage lipids (TAG), NH3, O2 and CO2, all in units of gram
In the following, we describe the simplified processes of the functional compartments and the regulation of the mass flux between these compartments in more detail.
Mass flux regulation
with threshold value x c .
Protein metabolism - equations.
a(t) = (a a , a p , a b , p a ) T
a a = amino acids in the free AA pool
a p = amino acids channeled to protein synthesis
a b = amino acids channeled to catabolism
p a = tissue protein given by its amino acid composition
u aa = 20 amino acids that are recieved from the feed (input variable)
= -k aa a a - U ps k p a a + u aa
k aa = 1/s
= U ps k p a a - v s (t)a p + f p v d (t)p a
k p = 1/s
= k aa a a + f t v d (t)p a - ab, cat
= v s (t)a p - v d (t)p a
Protein metabolism - parameters.
f p , f t
Flux of amino acids post degradation process
Optimal operation temperature
Effect of temperature on protein synthesis rate
Parabolic temperature-degradation constant
Minimum degradation rate factor in f d
Genetically determined efficiency of rate of protein degradation
Concentrations of EAA in the pool
Protein buildup capacity constant
Effect of EAA on protein degradation
Gain on the protein buildup capacity. Estimated from large scale experiments and simulations, not a sensitive parameter 
Gain on the protein degradation. Estimated from simulations of nitrogen excretion and Comparison with data 
where U ps denotes the regulation of the buildup protein capacity described below, and is the flux of AAs available from protein degradation.
Regulation of protein metabolism
Regulation in the protein metabolism model is performed by modulating both the flow of AAs U ps and protein degradation U pd . Protein deposition in fish has an upper buildup capacity, which was found empirically to be approximately 18-19 g protein per 100 g body weight, depending on size and age . As protein deposition approaches its capacity, further synthesis is inhibited. This empirical evidence for upper protein capacity has not yet been fully explained and might incorporate several regulatory steps in cell growth. The upper capacity is modeled here by the piecewise linear regulation function, U ps = U pl (p(t); p c m w (t)).
Deficiency in the EAAs promotes protein degradation that supplies the missing EAAs from the body tissue.
This suggests a Hill-type model, U pd = U i (EAA j ; V pd , β a , ), with being the minimum required concentrations of the j ∈ [1, 2, ..., 20] intracellular EAAs (given, for instance, by [35, 43]). Because we assume the protein degradation process to be irreversible, the rate v d (T) in (6) must always be positive, and the value of c d should satisfy c d > V pd ≥1. Furthermore, the values for the parameters c d and V pd must be assumed because the regulation U pd is not associated with any specific enzyme but rather represents the cumulative effect of degradation regulation. However, we tested several values and observed no significant differences in our results. The model appears to be robust with respect to changes in these values.
Lipid network states and parameters.
State space equations
d(AcCoA cyt )/dt = -7kq 1AcCoA cyt ·ATP-kq 2AcCoA cyt ·malCoA·NADPH
d(malCoA)/dt = 7kq 1AcCoA cyt ·ATP - 7kq 2AcCoA cyt ·malCoA·NADPH
d(FA)/dt = kq 2AcCoA cyt ·malCoA·NADPH - 3U ls k ls FA + 3U ld k ld TAG - U xd k xd FA·NAD+ +
d(TAG)/dt = U ls k ls FA - U ld k ld TAG
Regulation of lipid metabolism
Lipid metabolism - parameters.
kq 1, k ls , k ld
rate constants for the lipid metabolism reactions
1, 1, 0.1/μ mol/s
kq 2, k xd
rate constants for the lipid metabolism reactions
100, 0.3/μ mol2/s
maximal velocity of FA synthesis, estimated to produce rapid Hill reaction, according to 
Hill coefficient FA synthesis
citrate concentration at half FA synthesis reaction
0.2 μ mol/g
capacity of TAG deposition in the body
exponential coefficient for TAG deposition, estimated from simulations producing slow Hill dynamics
gain of the lipid synthesis function, estimated from simulation
Hill coefficient of the β-oxidation process
concentration of ATP at half saturation of β-oxidation process
7 μ mol/g
maximal velocity of β-oxidation
0.3 μ mol
concentrations of ATP at half saturation in the TAG degradation process
where the constant is the capacity of a TAG deposit (percent of the whole body), ℳ TAG is the molecular mass of TAG, and V ls ≥1 is the maximal reaction velocity.
The rate of the β-oxidation process is regulated at several levels. Although a carnitine shuttle appears to be an important regulatory step for β-oxidation in the liver , it is not as important in skeletal muscles and is not modeled here. During muscle contraction, for which the primary function of β-oxidation is the provision of energy, the β-oxidation flux is regulated through the ATP levels and the ratio of NAD+ to NADH [52, 53]. ATP regulation of the process is modeled via U xd = U i ([ATP]m w ; V xd , β xd , m w ), where is the ATP level at half saturation, which varies between 6 and 10 μ mol·g-1, depending on the cell type (Table 5). The rate of TAG hydrolysis is regulated similarly to the regulation mechanism described in equation (8). The function U ld is assigned a slightly lower K A value (i.e., in Table 5).
Reactions of the central metabolism submodel.
pyruvate + NAD+
Acetyl-CoA + Oxaloacetate
citrate + NAD+
α-ketoglutarate +NADH + CO2
α-ketoglutarate + NAD+
succinate + NADH + CO2
succinate + ADP+
malate + NAD+
Central metabolism equations.
State space equations
= h2 + h3 - h1- h4
= h10 - h11
= h1 - h5
= h11 + h12 - h13 - h3
= h5 + h7 - h6- h8
= h13 + h4 - h12 - h5
= h6 - h7 - h9
= h15 - h2
= h9 - h10
= h8 - h15
h1 = Updhkr 1[pyr]·[NAD+]
h9 = Uketokr 9[iso]·[ATP]
h2 = kr 2[mal cyt ]·[NADP+]
h10 = k tca [keto]·[NAD+]
h3 = Umalkr 3[mal]·[NAD+]
h11 = k citr [suc]
h4 = Uoxalkr 4[pyr]·[ATP]
h12 = Umalkmal [oxal]·[NADH]
h5 = k tca [AcCoA]·[oxal]
h13 = koxal [mal]·[NAD+]
h6 = k citr [cit]
h14 = Uoxalkr 1[pyr]·[ATP]
h7 = k iso [iso]
h15 = kr 2[oxal cyt ]·[NADH]
h8 = k cyt [cit]·[ATP]
Central metabolism - parameters.
k r 1,3,4,9
rate constants for the network functions h1, h3, h4 and h9
k r 2
rate constant for the network function h2
maximal velocity of the reaction involving pyruvate dehydro-genase (PDH)
concentrations of AcCoA at half saturation of the reaction involving PDH
maximal velocity of isocitrate dehydrogenase reaction
ATP levels at half saturation of the isocitrate dehydrogenase reaction
maximal velocity of the malate-pyruvate pathway, must be smaller than koxal to maintain regulation. Estimated from 
0.1 μ mol
oxaloacetate levels at half saturation of the malate-pyruvate pathway
maximal velocity of the pyruvate-oxaloacetate pathway
5 μ mol
AcCoA levels at half saturation in the pyruvate-oxaloacetate pathway
rate constant in the TCA cycle
Half saturation constant in citrate aconitase reaction
rate constant of citrate aconitase reversible reaction
rate constant of malate to oxaloacetate reaction
rate constant of oxaloacetate to malate reaction
rate constant of citrate transport from the matrix to the cytosol (by tricarboxylate carriers), estimated from , must be slower than isocitrate dehydrogenase reaction
The second regulatory step in the TCA submodel regulates the rate of oxidation of isocitrate to α-ketoglutarate via the allosteric enzyme isocitrate dehydrogenase. This enzyme is composed of three distinguishable subunits [56, 57], suggesting the description, Uketo = U i ([ATP]; Vketo, βketo, m w ). This suppresses the flux of isocitrate to α-ketoglutarate when energy is abundant, i.e., ATP ≫ m w . The reversible reaction to citrate enables citrate accumulation and transport from the matrix to the cytosol to participate in the FA synthesis process. Vketo is a sensitive parameter, because the equilibrium may not tend towards α-ketoglutarate if ATP is deficient.
Additionally, two regulatory functions preserve homeostasis in the TCA cycle. These regulate the concentrations of malate and oxaloacetate to prevent excessive accumulation of intermediates. In the malate-pyruvate pathway (Figure 12), malate is actively transported from the matrix to the cytosol, where malate dehydrogenase oxidizes L-malate to oxaloacetate and finally to pyruvate. The transporter (dicarboxylate carrier) is assumed to limit the rate of this pathway in a mechanism that follows Hill-type dynamics. The second regulation involves the enzyme pyruvate carboxylase, which replenishes oxaloacetate when needed. AcCoA is a positive allosteric modulator of this enzyme that follows Hill-type dynamics . The regulations are modeled using the functions Umal = U a ([oxal]; Vmal, 1, ·m w ) and Uoxal = Uα ([acc]; Voxal, 1, ·m w ), for malate and oxaloacetate, respectively. As the concentration of AcCoA in the matrix exceeds the half saturation value, the flow of AcCoA into the cycle is enhanced by replenishing oxaloacetate.
Glycolysis. State space equations for the glycolysis process, with the state variables glu, g6p, pGt, g3p are glucose, glucose-6-phosphate, 6-phosphogluconate and glyceraldehyde 3-phosp.
State space equation
[ ] = - kg 1[glu] + ug, feed
[ ] = kg 1[glu] - kg 2[g 6p]·[NADP+] - U glyc k glyc [g 6p]
[ ] = kg 2[g 6p]·[NADP+] - kg 2[pGt]·[NADP+]
[ ] = U glyc k glyc [g 6p] + kg 2[pGt]·[NADP+] - kg 2[g 3p]·ATP
kg 1= 1/s
kg 2= 1/μ mol/s
k glyc = 1/μ mol/s
This research was supported by the Norwegian University of Science and Technology (NTNU), Norway. NR acknowledges support by the German Research Foundation (DFG) through the Cluster of Excellence in Simulation Technology/(EXC 310) at the University of Stuttgart. The authors would like to thank professor Sigurd Skogestad for his useful comments and BioMar for the data from a large scale experiment they provided.
- Birkett S, de Lange K: Limitations of conventional models and a conceptual framework for a nutrient flow representation of energy utilization by animals. Brit J Nutr. 2001, 86: 647-659.View ArticleGoogle Scholar
- Kitano H: International alliances for quantitative modeling in systems biology. Mol Syst Biol. 2005, 1: 2005.0007Google Scholar
- Conceiçâo L, Verreth J, Verstegen M, Huisman E: A preliminary model for dynamic simulation of growth in fish larvae: application to the African catfish Clarias gariepinus and turbot Scophthalmus maximus. Aquaculture. 1998, 163: 215-235.View ArticleGoogle Scholar
- Alver MO: Modeling, Instrumentation and Control in Marine Larviculture. PhD thesis. 2006, Norwegian University of Science and Technology, Trondheim, NorwayGoogle Scholar
- Lupatsch I, Kissil GW: Feed formulations based on energy and protein demands in white grouper (Epinephelus aeneus). Aquaculture. 2005, 248: 83-95.View ArticleGoogle Scholar
- Hua K, Bureau D: Modelling digestible phosphorus content of salmonid fish feeds. Aquaculture. 2006, 254: 455-465.View ArticleGoogle Scholar
- Hop H, Tonn WM, Welch HE: Bioenergetics of Arctic cod (Boreogadus saida) at low temperatures. Can J Fish Aquat Sci. 1997, 54: 1772-1784.View ArticleGoogle Scholar
- MacNamara J: A perspective on research and future of metabolic models of farm animals. Progress in research on energy and protein metabolism, 109. Edited by: Souffrant W, Metges C. 2003, 99-112. Wageningen Academic PublishersGoogle Scholar
- Vetharaniam I, McCall D, Fennessy P, Garrick D: A model of mammalian energetics and growth: model development. Agr Syst. 2001, 68: 55-68.View ArticleGoogle Scholar
- Jørgensen C, Fiksen O: State-dependent energy allocation in cod (Gadus morhua). Can J Fish Aquat Sci. 2006, 63: 186-199.View ArticleGoogle Scholar
- Olsen O, Balchen J: Structured modelling of Fish Physiology. Math Biosci. 1992, 112: 81-113.View ArticleGoogle Scholar
- Jobling M: Fish Bioenergetics. 1994, Chapman and Hall, 1Google Scholar
- McCarthy I, Houlihan D: The effect of temperature on protein metabolism in fish: The possible consequences for wild Atlantic salmon (Salmo salar L.) stocks in Europe as a result of global warming. Global Warming: Implications for Freshwater and Marine Fish, no. 61 in Society of Experimental Biology Seminar Series. Edited by: Wood C, McDonald D. 1997, 51-77. Cambridge, U.K.: Cambridge University PressView ArticleGoogle Scholar
- Owen S, McCarthy I, Watt P, Ladero V, Sanchez J, Houlihan D, Rennie M: In vivo rates of protein synthesis in Atlantic salmon (Salmo salar L.) smolts determined using a stable isotope flooding dose technique. Fish Physiol Biochem. 1999, 20: 87-94.View ArticleGoogle Scholar
- McCarthy I, Moksness E, Pavlov D, Houlihan D: Effects of water temperature on protein synthesis and protein growth in juvenile Atlantic wolffish (Anarhichas lupus). Can J Fish Aquat Sci. 1999, 56: 231-241.View ArticleGoogle Scholar
- Houlihan D, Carter C, McCarthy I: Protein synthesis in fish. Metabolic Biochemistry. Edited by: Hochachka P, Mommsen T. 1995, 191-220. Amsterdam, The Netherlands: ElsevierView ArticleGoogle Scholar
- Lodish H, Berk A, Matsudaira P, Kaiser CA, Krieger M, Scott MP, Zipursky SL, Darnell J: Molecular Cell Biology. 2004, 41 Madison Avenue, NY: W.H. Freeman and Company, 5Google Scholar
- Bajzer Z, Vuk-Pavlovic S: Modeling positive regulatory feedbacks in cell-cell interactions. Biosystems. 2005, 80: 1-10.View ArticleGoogle Scholar
- Bar NS, Sigholt T, Shearer KD, Krogdahl A: A Dynamic Model of Nutrient Pathways, Growth and Body Composition in Fish. Can J Fish Aquat Sci. 2007, 64: 1669-1682.View ArticleGoogle Scholar
- Ho Y, Varley J, Mantalaris A: Development and analysis of a mathematical model for antibody-producing GS-NS0 cells under normal and hyperosmotic culture conditions. Biotechnol Prog. 2006, 22: 1560-1569.View ArticleGoogle Scholar
- Kontoravdi C, Wong D, Lam C, Lee Y, Yap M, Pistikopoulos E, Mantalaris A: Modeling amino acid metabolism in mammalian cells - towards the development of a model library. Biotechnol Prog. 2007, 23: 1261-1269.View ArticleGoogle Scholar
- , : Principles of Biochemisty. 2005, W.H. Freeman and Company, 1Google Scholar
- Refstie S, Olli JJ, Standal H: Feed intake, growth and protein utilisation by post-smolt Atlantic salmon (salmo salar) in response to graded levels of fish protein hydrolysate in the diet. Aquaculture. 2004, 239: 331-349.View ArticleGoogle Scholar
- Ji H, Bradley TM, Tremblay GC: Atlantic Salmon (Salmo salar) Fed L-Carnitine Exhibit Altered Intermediary Metabolism and Reduced Tissue Lipid, but No Change in Growth Rate. J Nutr. 1996, 126 (8): 1937-1950.Google Scholar
- Waiwood B, Haya K, Eeckhaute L: Energy metabolism of hatchery-reared juvenile salmon (Salmo salar) exposed to low pH. Comp Biochem Physiol. 1992, 101C: 49-56.Google Scholar
- Haya K, Waiwood B, Eeckhaute LV: Disruption Of Energy Metabolism And Smoltification During Exposure Of Juvenile Atlantic Salmon (Salmo Salar) To Low PH. Comp Biochem Physiol. 1985, 82C (2): 323-329.Google Scholar
- Schuhmacher A, Wax C, Gropp JM: Plasma amino acids in rainbow trout (Oncorhynchus mykiss) fed intact protein or a crystalline amino acid diet. Aquaculture. 1997, 151: 15-View ArticleGoogle Scholar
- Walton MJ, Wilson RP: Postprandial changes in plasma and liver free amino acids of rainbow trout fed complete diets containing casein. Aquaculture. 1986, 51 (2): 105-115.View ArticleGoogle Scholar
- Murai T, Ogata H, Hirasawa Y, Akiyama T, Nose T: Portal Absorption and Hepatic Uptake of Amino Acids in Rainbow Trout Forced-Fed Complete Diets Containing Casein or Crystalline Amino Acids. Nippon Suisan Gakkaishi. 1987, 53 (10): 1847-1859.View ArticleGoogle Scholar
- Houlihan D, Boujard T, Jobling M, : Food Intake in Fish. 2001, Oxford, UK: Blackwell ScienceGoogle Scholar
- Fauconneau B, Mady MP, LeBail PY: Effect of growth hormone on muscle protein synthesis in rainbow trout (Oncorhynchus mykiss) and Atlantic salmon (Salmo salar). Fish Physiol Biochem. 1996, 15: 49-56.View ArticleGoogle Scholar
- Cacho O: Protein and fat dynamics in fish: A bioenergetic model applied to Aquaculture. Ecol Model. 1990, 50: 33-56.View ArticleGoogle Scholar
- Zhou Z, Xie S, Lei W, Zhu X, Yang Y: A bioenergetic model to estimate feed requirement of gibel carp, Carassius auratus gibeliom. Aquaculture. 2005, 248: 287-297.View ArticleGoogle Scholar
- Wood CW: Influence of feeding, exercise and temperature on nitrogen metabolism and excretion. Nitrogen Excretion. Edited by: Wright P, Anderson P. 2001, 201-238. London, UK: Academic PressView ArticleGoogle Scholar
- Shearer K, Åsgård T, Andorsdöttir G, Aas G: Whole body elemental and proximate composition of Atlantic salmon Salmo salar during the life cycle. J Fish Biol. 1994, 44: 785-797.View ArticleGoogle Scholar
- Koojiman S: Dynamic Energy and Mass Budgets in Biological Systems. 2000, Cambridge, UK: Cambridge University Press, 2View ArticleGoogle Scholar
- Machiels M, Henken A: A Dynamic Simulation Model for Growth of the African Catfish, Clarias gariepinus (Burchell 1822) I. Effect of Feeding Level on Growth and Energy Metabolism. Aquaculture. 1986, 56: 29-52.View ArticleGoogle Scholar
- Murray JD: Mathematical Biology, Volume 19 of Biomathematics. 1980, New York: Springer-VerlagGoogle Scholar
- Segel LA: Modeling Dynamic Phenomena in Molecular and Cellular Biology. 1984, Cambridge, UK: Cambridge University PressGoogle Scholar
- Gillooly J, Brown J, West G, Savage V, Charnov E: Effects of Size and Temperature on Metabolic rate. Science. 2001, 293: 2248-2251.View ArticleGoogle Scholar
- Bar N, Lale R: Modeling and Control of the Protein Synthesis process in Eukaryotic cells. IEEE Proceedings: Conference on Decision and Control (CDC), Cancun, Mexico. 2008, 179-184.Google Scholar
- Bar NS, Morris DR: Dynamic model of the process of protein synthesis in Eukaryotic cells. B Math Biol. 2007, 69: 361-393.View ArticleGoogle Scholar
- Halver JE, Hardy RW: Fish Nutrition. 2002, California, USA: Academic Press, 3Google Scholar
- Voet D: Biochemistry. 2004, [Wiley International Edition], Hoboken, NJ: John Wiley & Sons, inc, 3Google Scholar
- Halestrap AP, Denton RM: Hormonal regulation of adipose-tissue acetyl-Coenzyme A carboxylase by changes in the polymeric state of the enzyme. The role of long-chain fatty acyl-Coenzyme A thioesters and citrate. Biochem J. 1974, 142 (2): 365-0.PubMed CentralView ArticleGoogle Scholar
- Munday MR: Regulation of mammalian acetyl-CoA carboxylase. Biochem Soc T. 2002, 30 (Pt 6): 1059-1064.View ArticleGoogle Scholar
- Saddik M, Gamble J, Witters L, Lopaschuk G: Acetyl-CoA carboxylase regulation of fatty acid oxidation in the heart. J Biol Chem. 1993, 268 (34): 25836-25845.Google Scholar
- Chwalibog A, Thorbek G: Energy metabolism and lipogenesis in humans. Thermochim Acta. 2000, 349: 43-51.View ArticleGoogle Scholar
- Chwalibog A, Jakobsen K, Tauson AH, Thorbek G: Energy metabolism and nutrient oxidation in young pigs and rats during feeding, starvation and re-feeding. Comp Biochem Phys A. 2005, 140: 299-307.View ArticleGoogle Scholar
- Bendiksen E, Shearer K, Hillestad M: Whole body chemical composition of farmed Atlantic salmon (Salmo salar L.) from smolt to harvest. Proceedings of XII International Symposium on Fish Nutrition and Feeding, Biarritz, France: INRA. 2006, 121-Google Scholar
- Jobling M, Johansen S: The lipostat, hyperphagia and catch-up growth. Aquac Res. 1999, 30: 473-478.View ArticleGoogle Scholar
- Eaton S: Control of mitochondrial [beta]-oxidation flux. Prog Lipid Res. 2002, 41 (3): 197-239.View ArticleGoogle Scholar
- Oram JF, Bennetch SL, Neely JR: Regulation of Fatty Acid Utilization in Isolated Perfused Rat Hearts. J Biol Chem. 1973, 248 (15): 5299-5309.Google Scholar
- Behal RH, Buxton DB, Robertson JG, Olson MS: Regulation of the Pyruvate Dehydrogenase Multienzyme Complex. Annu Rev Nutr. 1993, 13: 497-520.View ArticleGoogle Scholar
- Harris RA, Bowker-Kinley MM, Huang B, Wu P: Regulation of the activity of the pyruvate dehydrogenase complex. Adv Enzyme Regul. 2002, 42: 249-259.View ArticleGoogle Scholar
- Anderson S, Schirf V, McAlister-Henn L: Effect of AMP on mRNA Binding by Yeast NAD+ -Specific Isocitrate Dehydrogenase. Biochemistry. 2002, 41 (22): 7065-7073.View ArticleGoogle Scholar
- Huang YC, Kumar A, Colman RF: Identification of the Subunits and Target Peptides of Pig Heart NAD-Specific Isocitrate Dehydrogenase Modified by the Affinity Label 8-(4-Bromo-2, 3-dioxobutylthio)NAD. Arch Biochem Biophys. 1997, 348: 207-218.View ArticleGoogle Scholar
- Libor SM, Sundaram TK, Scrutton MC: Pyruvate carboxylase from a thermophilic Bacillus. Studies on the specificity of activation by acyl derivatives of coenzyme A and on the properties of catalysis in the absence of activator. Biochem J. 1978, 169 (3): 543-558.PubMed CentralView ArticleGoogle Scholar
- Dobly A, Martin S, Blaney S, Houlihan D: Protein growth rate in rainbow trout (Oncorhynchus mykiss) is negatively correlated to liver 20S proteasome activity. Comp Biochem Physiol. 2004, 137: 75-85.View ArticleGoogle Scholar
- Mente E, Deguara S, Begona M, Houlihan D: White muscle free amino acid concentrations following feeding a maize gluten dietary protein in Atlantic salmon (Salmo salar L.). Aquaculture. 2003, 225: 133-147.View ArticleGoogle Scholar
- Moyes CD, West TG: Exercise Metabolism of Fish. Metabolic Biochemistry. Edited by: Hochachka P, Mommsen T. 1995, 368-392. Amsterdam, The Netherlands: ElsevierGoogle Scholar
- Bar NS: Dynamic Model of Fish Growth. PhD thesis. 2007, Department for Engineering Cybernetics, Norwegian University of Science and TechnologyGoogle Scholar
- Bendiksen E, Arnesen A, Jobling M: Effects of dietary fatty acid profile and fat content on smolting and seawater performance in Atlantic salmon (Salmo salar L.). Aquaculture. 2003, 225: 149-163.View ArticleGoogle Scholar
- Caldwell C, Hinshaw J: Nucleotides and adenylate energy charge as indicators of stress in rainbow trout (Oncorhynchus mykiss) subjected to a range of dissolved oxygen concentrations. Comp Biochem Physiol. 1994, 109B: 313-323.Google Scholar
- Jørgensen J, Mustafa T: The effect of hypoxia in Carbohydrate metabolism in Flounder (Platichthys flesus L.)-II. High energy phosphate compounds and the role of glycolytic and Gluconeogenetic enzymes. Comp Biochem Physiol. 1980, 67B: 249-256.Google Scholar
- Mommsen T, Hochachka P: The Purine Nucleotide Cycle as Two Temporally Separated Metabolic Units: A Study on Trout Muscle. Metabolis. 1988, 37: 552-558.View ArticleGoogle Scholar
- Ravindran S, Radke GA, Guest JR, Roche TE: Lipoyl Domain-based Mechanism for the Integrated Feedback Control of the Pyruvate Dehydrogenase Complex by Enhancement of Pyruvate Dehydrogenase Kinase Activity. J Biol Chem. 1996, 271 (2): 653-662.View ArticleGoogle Scholar
- Soundar S, Park JH, Huh TL, Colman RF: Evaluation by Mutagenesis of the Importance of 3 Arginines in alpha, beta, and gamma Subunits of Human NAD-dependent Isocitrate Dehydrogenase. J Biol Chem. 2003, 278 (52): 52146-52153.View ArticleGoogle Scholar
- LaNoue KF, Bryla J, Williamson JR: Feedback Interactions in the Control of Citric Acid Cycle Activity in Rat Heart Mitochondria. J Biol Chem. 1972, 247 (3): 667-679.Google Scholar
- Lindbladh C, Rault M, Hagglund C, Small WC, Mosbach K, Bülow L, Evans C, Srere PA: Preparation and kinetic characterization of a fusion protein of yeast mitochondrial citrate synthase and malate dehydrogenase. Biochemistry-Us. 1994, 33: 11692-11698.View ArticleGoogle Scholar
- Moya-Falcón C, Hvattum E, Tran T, Thomassen M, Skorve J, Ruyter B: Phospholipid molecular species, β-oxidation, desaturation and elongation of fatty acids in Atlantic salmon hepatocytes: Effects of temperature and 3-thia fatty acids. Comp Biochem Physiol. 2006, 145: 68-80.View ArticleGoogle Scholar
- Agrawal PK, Garg GK, Gollakota KG: Studies on two isozymes of aconitase from T. III. Enzymatic properties. Biochem Bioph Res Co. 1976, 70 (3): 987-996.View ArticleGoogle Scholar
- Hodges M, Yikilmaz E, Patterson G, Kasvosve I, Rouault TA, Gordeuk VR, Loyevsky M: An iron regulatory-like protein expressed in Plasmodium falciparum displays aconitase activity. Mol Biochem Parasit. 2005, 143: 29-38.View ArticleGoogle Scholar
- Kaplan RS, Mayor JA: Structure, function and regulation of the tricarboxylate transport protein from rat liver mitochondria. Journal of Bioenergetics and Biomembranes. 1993, 25 (5): 503-514.View ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.