Modeling the CalvinBenson cycle
 Jiri Jablonsky^{1}Email author,
 Hermann Bauwe^{2} and
 Olaf Wolkenhauer^{1}
DOI: 10.1186/175205095185
© Jablonsky et al; licensee BioMed Central Ltd. 2011
Received: 7 September 2011
Accepted: 3 November 2011
Published: 3 November 2011
Abstract
Background
Modeling the CalvinBenson cycle has a history in the field of theoretical biology. Anyone who intends to model this system will look at existing models to adapt, refine and improve them. With the goal to study the regulation of carbon metabolism, we investigated a broad range of relevant models for their suitability to provide the basis for further modeling efforts. Beyond a critical analysis of existing models, we furthermore investigated the question how adjacent metabolic pathways, for instance photorespiration, can be integrated in such models.
Results
Our analysis reveals serious problems with a range of models that are publicly available and widely used. The problems include the irreproducibility of the published results or significant differences between the equations in the published description of the model and model itself in the supplementary material. In addition to and based on the discussion of existing models, we furthermore analyzed approaches in PGA sink implementation and confirmed a weak relationship between the level of its regulation and efficiency of PGA export, in contrast to significant changes in the content of metabolic pool within the CalvinBenson cycle.
Conclusions
In our study we show that the existing models that have been investigated are not suitable for reuse without substantial modifications. We furthermore show that the minor adjacent pathways of the carbon metabolism, neglected in all kinetic models of CalvinBenson cycle, cannot be substituted without consequences in the mass production dynamics. We further show that photorespiration or at least its first step (O_{2} fixation) has to be implemented in the model if this model is aimed for analyses out of the steady state.
Background
There are two common approaches used in modeling the CalvinBenson cycle: kinetic modeling, e.g., [2, 3] and stoichiometric modeling, e.g., [4–6]. Kinetic modeling requires to obtain/have available kinetic properties of the enzymes involved. These are mostly known if one assumes conservative kinetic parameters among the species with several different ways of CO_{2} fixation. On the other hand, stoichiometric modeling allows an analysis of the entire carbon metabolism within the given constraints and without the need of kinetic properties. It would seem plausible to combine both approaches and overlap fluxes from the kinetic model of the CalvinBenson cycle with the stoichiometric analysis of the metabolic network [7, 8].
Even though it is a core part in a variety of photosynthetic models or projects modeling the entire photosynthesis [9, 10], kinetic modeling has mostly focused on the stability of the CalvinBenson cycle itself and continues to attract considerable scientific interest [11, 12]. To anyone entering this field, it is clear that the CalvinBenson cycle is the beststudied plant metabolic system. Since modeling the CalvinBenson cycle has a long history, the question arises how one should proceed with a further analysis of this system. The standard scientific approach is to build on existing knowledge and models. A natural progression would thus to adapt existing models, to refine and expand them for adjacent metabolic pathways.
What our present work reveals is that, without significant modifications and corrections of the wellestablished models, it is not possible to reuse existing models. Our analysis shows that one of the existing and widely used models produces an "accumulation" of starch in high negative concentration with consequences to the behavior of the whole system (apart from being biologically implausible). In another case an interchanged parameter leads to 100times faster equilibrium in one particular reaction, which significantly shifted the steady states of all metabolites. We furthermore found errors, which have propagated from one generation (publication) to another. Our study highlights difficulties with the reproducibility of modelbased results, provides solutions and suggestions on how to proceed with extensions to existing models of the CalvinBenson cycle.
In order to support the evolution of models for the CalvinBenson cycle and to improve the reproducibility of modeldriven results, we discuss ways to improve and correct existing models. We also open a discussion about how to connect the CalvinBenson cycle with other parts of the carbon metabolism in the case of cyanobacteria and in relation to the system's efficiency, regulation, mass production and implementation of photorespiration.
Methods
To investigate mathematical and computational models for their suitability and possible errors all models were encoded both in Matlab, using the SimBiology toolbox (MathWorks Inc., Natick, Massachusetts, USA) and at the same time in Copasi [13]. For transparency and to allow readers to use corrected models of the CalvinBenson cycle, we make all files and data required for the simulation publicly available in additional files (.xml).
Results and Discussion
Available kinetic models for the CalvinBenson cycle
There are studies that successfully use a simplified description of the CalvinBenson cycle, for instance the photosynthetic oscillations in the chlorophyll fluorescence [14]. However, if one analyses the regulation of carbon metabolism, such study cannot be done without a detailed kinetic model of the CalvinBenson cycle. Since kinetic modeling of the CalvinBenson cycle has a long tradition [2, 3, 15–22], the standard approach would be to adapt, refine and expand an existing model. Focusing on the CalvinBenson cycle and carbon metabolism, we have investigated several wellestablished models of this system, which are discussed in further detail below.
Hahn models
This fundamental model was presented by Hahn [15] who also published further related models of the CalvinBenson cycle [16, 17]. The kinetic parameters in all models were chosen on the basis of a realistic photosynthetic rate in order to give reasonable steadystate values for metabolites. One would assume that the earlier model [15] was extended by photorespiration in the next model generation [17]; together with the realistic response to the changes in the O_{2} and CO_{2} concentrations [17]. This assumption is based on the fact that the equations describing the CalvinBenson cycle are the same in both models [15, 17]; however, some values of the kinetic parameters differ significantly.
The most prominent example is the rate of the CO_{2} fixation expressed by the rate constant k_{1}. The value of this calculated parameter is in both models identical [15, 17], i.e. on the basis of estimated gross photosynthetic rate per unit area of leaf tissue $\left({P}_{g}^{A}\right)$. Since the $\left({P}_{g}^{A}\right)$ is the same in both models, it is not clear why value of k_{1} is 57.3  fold higher in the later model [17]. Hahn mentions that the value of k_{1} was, in order to compensate the neglected photorespiration in the earlier model [15], decreased (without any detail regarding the original value of k_{1}). However, it is unlikely that the original and unknown value of k_{1} in the earlier model [15] is 57.3  fold lower due to photorespiration because the author himself assumed in the later model [17] the ratio for the carboxylase/oxygenase activity of RuBisCO (Ribulose1,5bisphosphate carboxylase oxygenase) equal to 3.7.
Nevertheless, we have encoded the earlier Hahn model [15] into biochemical simulators according its original description. Having only the rate constants, we employed mass action kinetics as the author did [15]. For initial concentrations we used the steadystate values calculated by Hahn [15].
The reason for the drop in the ATP level in the certain time point (Figure 2) was a depletion of both intra and extracellular phosphates (data not shown). Finally, we achieved a stable solution by fixing (having constant concentration) the Pi_ext (extracellular phosphate). This approach prevents the suspension of the cycle (Figure 2) but the steadystate of the system is still far away from known values (e.g., the aforementioned range for PGA) or results presented by author (Figure 2 in [15]).
Pettersson model
Poolman and ZhuLaisk models
Although it is not a problem to reconstruct the established fundamental models from their description, a problem is that our simulations based on the Hahn and Pettersson models demonstrate totally different behaviors in comparison to what was presented by their authors [2, 15]. It is not clear what the problem is because the computational methodology used for running the simulations has not been described in the publications. But if one employs well established computational methods and tools (e.g., Matlab), the original results [2, 15] are not reproducible. Therefore, if one wants to study the cycle itself or expand the model by adding other part(s) of the carbon metabolism, the reasonable cause of action would be to look for available models for which one knows all details about the computational methods. Such models can be nowadays found in model databases (e.g., http://www.ebi.ac.uk/biomodelsmain) or they are made available as a supplement to a publication. For the CalvinBenson cycle, there are two main options, the Poolman model [21] and the Zhu model [3].
The Poolman model is available as a .xml file (http://www.ebi.ac.uk/biomodelsmain/) and can be therefore easily imported to any biochemical simulator supporting SBML (system biology markup language) level 2. We have employed both Copasi [13] and the SimBiology toolbox for Matlab (MathWorks) to study this model. The Poolman model is structurally very close to the Pettersson model [2] but with an altered kinetic parameters. One can speculate that the author encountered the same problems with Pettersson model as we have and as a result, employed different set of kinetic parameters. This model was used as an example to illustrate possible applications in the metabolic modeling [21] and stability analysis (e.g., number of steady states) rather [22] than for kinetic modeling of the particular metabolite(s). The initial conditions were exactly the same as in the original work [21].
Pilot simulations based on the Poolman model showed that the system reached the steady state at 0.6 s from the start of the simulation. If the initial concentration of PGA (3phosphoglycerate) is increased 10times, the steady state is reached after 1.4 s. This feature of the model is based on the assumption that metabolites in the system are maintained at a fast equilibrium [2]. Furthermore, to ensure the assumed fast equilibrium, rate constants in the Poolman model were set to extremely high values. For example, the rate constants for reactions catalyzed by transketolase (equations E7 and E10, [21]) are equal to 500 000 000 l·mmol ^{1}·s^{1} [21] and the flux through these reactions is 39.7 mmol·s^{1} (our calculation based on the unmodified Poolman model [21]). This flux is in sharp contrast to the proposed (0.1 mmol·s^{1} [19]) and measured (0.36 mmol·s^{1} [26]) maximal rates for these reactions.
The Poolman model reaches its equilibrium at 0.6 s of the simulation also in other cases (after dark → light switch or changes in CO_{2} level). On the other hand, it is known that the real system is limited by RuBisCO (ribulose1,5bisphosphate carboxylase/oxygenase) activation whose induction phase takes 120 s [27]  600 s [28]. We note that the activation of RuBisCO was also considered in a kinetic model with the time constant equals to 250 s [14].
The CalvinBenson cycle is an open system with many inputs and outputs (Figure 1). It is therefore important to analyze its output behavior as well. To that end, we have changed the settings for sinks and starch in the Poolman model from the fixed (constant) concentration to variable concentration (responding according the reactions and differential equations). We note that the original setting for other fixed metabolites, i.e., for inputs (CO_{2}, cytosolic Pi, NADPH, NADP^{+} and H^{+}) was not changed.
The Poolman model was designed for other purposes (stability analysis) but it is clear that this model cannot be used for the purpose of comparison with experimental timeseries data without changing both the equations and kinetic parameters.
The remaining option is the Zhu model [3]. Zhu and coworkers presented a model of C_{3} photosynthesis, extending the Laisk model [19] by addition of the photorespiratory pathway. Since we do not focus on the complete photorespiratory pathway in this study, we did not analyze the Laisk model separately, except for the difference in the case of phosphate translocator (see the section CalvinBenson cycle as a part of the metabolic network: PGA sink implementation).
The Zhu model is a natural starting point and suitable template for studying photorespiration, even if it is not easily accessible and understandable because it is encoded in 34 Matlab files. The metabolites in the Zhu model reach the equilibrium at 150 s of the simulation (data not shown), which is in agreement with previous findings [27] and in contrast to 0.6 s based on the Poolman model [21]. However, several discrepancies and serious errors in the Zhu model, particularly in the part describing the CalvinBenson cycle, can be found. The discrepancies in the Zhu model include interchanged values of k_{E13} and k_{I135} in Appendix C and the nowhere used constant K_{M8}, make the description of the model slightly confusing. The following section is dedicated to the analysis of the errors occurring in the Zhu model, as well as necessary modifications for this model.
Analysis of the Zhu model and problems occurring in kinetic modeling of the CalvinBenson cycle
The programming environments used for kinetic modeling of the CalvinBenson cycle include Turbo Pascal, used in [19]; Scampi, used in [21] or Matlab, used in [3]. Commandline programming is more prone to errors and also makes it more likely for errors to propagate into new models. For example, in the Laisk model [19] one error propagated into other models due the missing dimensional analysis in Pascal. This problem can also be found in the last generation of the model containing the CalvinBenson cycle, in the Zhu model [3]. The following problems occur in the Zhu model:

Two very different descriptions for reactions v7: S7P + GAP = Ri5P + Xu5P and v10: F6P + GAP = E4P + Xu5P), one in the paper (Appendix A) and another one encoded in the MATLAB file PSRate.m in supplement of the paper [3].

Wrong dimension for equations v7 and v10.

Kinetic parameters were incorrectly taken from kinetic characterization of transaldolase instead of transketolase.

The value of the equilibrium constant k_{E7} in equation v7 is 100 times higher than what is known from literature [1, 19].
In order to analyze these problems, we have rewritten the 2007 Zhu model, originally encoded in Matlab, for use with Copasi [13] and also rewrote it for the SimBiology. Since all errors occur in the description of the CalvinBenson cycle, the photorespiratory pathway was not considered in the rewritten Zhu model. Finally, we improved the Zhu model in a way that all metabolites have their own concentrations, e.g., we consider separated DHAP and GAP instead of T3P pool, in comparison to original 2007 Zhu model. We then compared the steadystate values of the metabolites in dependence of ATP · (ADP + ATP)^{1} ratio, reflecting the energy limitation [23]. Since all errors are related to the transketolase, we discuss the consequence only for two substrates of transketolase, F6P (fructose 6phosphate) and S7P (sedoheptulose7phosphate).
The case with two model versions gets more complicated because both of them suffer from other problems. The first one, related only to the version in Appendix A [3], is the consequence of a wrong dimension for equations describing the reactions catalyzed by transketolase (v7 and v10) and has its origin in the Laisk model [19]. The problem is that the modifiers E4P (erythrose 4phosphate) and Ri5P (ribose 5phosphate) were multiplied instead of added in one of the equations (TEMP1, [3]) employed in the aforementioned reactions. Even if this particular error has an negligible impact in the calculation itself (< ± 1%, data not shown), the incorrect rate equations do not allow a computation of the simulations in biochemical simulators that test the dimension of equations (e.g., SimBiology toolbox).
An alternative but not plausible approach was employed in the case of kinetic parameters for rate equation v10 in the Zhu model. Zhu and coworkers employed different kinetic parameters [29] in comparison to Laisk model [19]. The reason why this is not justified is that the new kinetic parameters were taken from transaldolase [29] instead of transketolase. The fact that the source of kinetic parameters is the nonphotosynthetic organism Dictyostelium discoideum might be justified. However, we cannot overlook the fact that the transaldolase catalyses the same reaction as transketolase but in opposite direction; it is localized in the pentose phosphate pathway.
One question that arises is why the original kinetic parameters for the rate equation v10 from the Laisk model [19] were not employed in the Zhu model. Strikingly, the original parameters from the Laisk model [19] block the cycle and lead to an accumulation of F6P and S7P, see Figure 5 (gray). This behavior is caused primarily by big difference between the original [19] and new [29] value of the parameter k_{M103}, 0.015 mmol·l^{1} and 0.46 mmol·l^{1}, respectively. In order to sustain a stable solution, the value of the parameter k_{M103} in the rate equation v10 must be equal or higher than 0.06 mmol·l^{1} (data not shown). This problem occurs, however, only for model version from Appendix A [3],. The Matlab model version [3] can employ the original parameter k_{M103} = 0.015 mmol·l^{1} from Laisk model [19] and sustains the stable solution sensitive to the changes of ATP level, see Figure 5 (violet).
Finally, and probably the most visible error in the Zhu model is an incorrect value for the equilibrium constants k_{E7}, that is, 10 [3] instead of 0.076 [1] or 0.1 [19]. The consequence is a theoretical 100fold increase in the forward rate of the reaction F6P + GAP = E4P + Xu5P, which is able to shift the equilibrium of entire model and reduce the starch synthesis. In order to analyze the possible impact, we performed the simulations based on both Appendix A (with k_{M103} = 0.06 mmol·l^{1}) and the Matlab version of Zhu model, both with the correct value of k_{E7} = 0.076 [1]. At first, we can compare original versions of Zhu model after correction on k_{E7}. The steady state concentrations of F6P and S7P are now 6 and 4.8fold higher, respectively, if we compare the original [3] and corrected Appendix A model (additional file 1). This is shown in Figure 5 in black and green. In the case of the original [3] and here corrected Matlab version of the Zhu model (additional file 2), the steady state concentrations of F6P is increased by 6.2fold but S7P is 9.9fold decreased, respectively; see Figure 5  red and blue. The quantitative differences in the behavior between these two versions of the Zhu model suggest that one should really speak about two different models rather than two versions of one model. This conclusion is supported by final comparison of both corrected models  the difference in steady state concentration of S7P between both Zhu models is 121fold as shown in Figure 5  green and blue.
CalvinBenson cycle as a part of the metabolic network: PGA sink implementation
Besides the approach employed in the Laisk model for the phosphate translocator, one can find another and simple one, introduced in the Pettersson model [2] and adopted in the next model generation [3, 21]. In this approach, the phosphate translocator is encoded in the MichaelisMenten kinetics and regulated, besides the PGA, by concentration of internal/external phosphate, DHAP and GAP. We note that, in comparison to Laisk model [19], the reaction PGA_{cytosol} → Sink is redundant in this case but was employed anyway [3]. Since this approach can be used only for chloroplast, we also tested a simplified version for cyanobacteria described with the simplest MichaelisMenten kinetics, regulated only by concentration of PGA and estimated k_{M}. To have an unregulated implementation of the PGA sink to compare with, we employed also irreversible kinetics using the mass action law without modifiers. Since we were interested in how the changes within the CalvinBenson can influence associated pathways of carbon metabolism and the efficiency of the cycle itself, we have focused our analysis on the content of metabolites within the CalvinBenson cycle as well as on the accumulation of key products, i.e., starch and PGA sink after 3 hours in the steady state conditions.
Analysis of Zhu model, versions from Matlab and Appendix A, in dependence of employed implementation of PGA sink in the model
PGA sink  Starch  CB pool  Total mass  

A model, MA  82.3 (89.5)  93.3 (61.8)  29.9 (16.8)  81.6 (86.2) 
A model, MM  96.4 (104.1)  102.5 (83.1)  162.4 (51.7)  98.4 (101.7) 
A model, MMcb  99.3 (103.8)  101.2 (85)  74.3 (63.1)  98.8 (101.8) 
M model, MA  87.3 (95.4)  72.8 (31.7)  17.3 (15.4)  84.9 (90.0) 
M model, MM  100.0 (106.1)  100.0 (69.2)  100.0 (55.0)  100.0 (102.8) 
M model, MMcb  101.4 (105)  98.2 (74.7)  59.6 (73.5)  100.2 (102.5) 
Minor sinks for adjacent pathways, photorespiration and system efficiency
The CalvinBenson cycle in cyanobacteria is directly linked to other parts of the carbon metabolism. The question therefore arises if all adjacent metabolic pathways have to be considered in the kinetic model. An associated problem to solve is if it is the number of sinks itself (Figure 1) or just the total flux out of the CalvinBenson cycle which has an impact on the efficiency of the mass production. Finally, even if photorespiration is considered only in the minority of models [3, 17], its flux out of the system is considered, usually in the phosphate translocator reactions, otherwise the CalvinBenson cycle cannot reach the steady state due to infinite grow of PGA concentration (data not shown). We therefore aimed our focus also in this direction.
The starting point for our analysis was the reconstruction of the metabolic network for cyanobacteria Synechocystis sp. PCC 6803, adjusted for highest efficiency of the mass production [5]. We have taken the model A (corrected Zhu model  version from Appendix A; additional file 1), particularly the modified subversion for cyanobacteria (sinks only for PGA and F6P) and used this model as a template for our further modeling efforts.
Photorespiration not directly implemented but considered within PGA sink
As previously mentioned, photorespiration does not have to be considered in the model but its flux out of the system cannot be neglected. Otherwise, the CalvinBenson cycle cannot reach the steady state due to an accumulation of PGA. In order to analyze the consequences of this approach, together with analysis of the minor sinks, we have developed models C1 and C2. Models C were "extended" by PGCA (phosphoglycolate) sink (photorespiration) in a way which was employed in the majority of models of the CalvinBenson cycle i.e., indirectly within the phosphate translocator. In our case, the PGCA sink was considered as a part of the PGA sink with the basic flux 4.2% of the RuBP (ribulose 1,5biphosphate) synthesis flux [5] multiplied by a stoichiometric factor 1.66 (RuBP has 5 carbons, PGA has 3 carbons).
Both models (C1 and C2) were adapted for the same RuBP synthesis flux in the steady state (Δ flux = 5 × 10^{6}) and flux through F6P sinks, 2.4% [5] of the RuBP (ribulose 1,5biphosphate) synthesis flux in the steady state within the CalvinBenson cycle. The key difference between models C1 and C2 is that in the case of the model C1 (additional file 3), sinks for DHAP, E4P and Ri5P were not encoded in the model but their basic fluxes were multiplied by a stoichiometric factor q (e.g., for Ri5P, q = 1.66), summed and added to the basic flux of the PGA sink which simulates the other sinks within the PGA sink (the same approach was used for PGCA). The basic fluxes for Ri5P, DHAP and E4P sinks are 0.7%, 0.32% and 0.31% of the RuBP synthesis flux (Henning Knoop, personal communication), respectively. On the other hand, model C2 (additional file 4) was extended by sinks for DHAP, E4P and Ri5P with basic fluxes according the above mentioned proportions (Henning Knoop, personal communication).
In the case of PGA sink, the maximal and applied stable flux (value after 3 hours) was found to be equal to 18.2% in the case of C1 model (after subtraction of other sinks, see above) and 19.3% (C2 model) instead of proposed theoretical 21.7% of the RuBP synthesis flux [5]. We note that the same flux through the PGA might be achieved without disbalancing the RuBP synthesis flux in both C models by using different set of parameters for each model but the parameters may be out of the physiological range. The initial concentrations for both C models (as well as for D models) were taken from the original Zhu model which has significantly different equilibrium in comparison to the corrected version and the system was set out of the steady state.
Photorespiration implemented directly as O_{2}fixation
However, our simulations showed that it is not necessary to use two separated complex equations for the carboxylase and oxygenase (see the equation above) activity of RuBisCO as it was done by Zhu and coworkers [3]. It is possible to encode only one equation (the one for the carboxylase activity) and the reaction for CO_{2} fixation can be extended by additional product PGCA: RuBP + CO_{2} → A × PGA + B × PGCA; in our case A = 1.916 and B = 0.084 (i.e., 4.2%, [5]). The differences between these two approaches are negligible both in and out of the steady state (data not shown). Moreover, our approach is easier for adjusting the level of photorespiration according our expectations and tested hypotheses and it was therefore employed in D models (see below).
Both models (D1 and D2) were again adapted for the same RuBP synthesis flux in the steady state (Δ flux = 3 × 10^{4}). The maximal and applied stable flux for PGA sink (value after 3 hours) was found to be equal to 17.1% in the case of D1 model (after subtraction of other sinks, see above) and 18.2% (D2 model) which shows slightly lower efficiency in comparison to C models.
The results based on D1 model (additional file 5) and D2 model (additional file 6) show totally different picture of what is going on during the response to perturbation until the steady state is reached, above all in the range of seconds  thousands of seconds, in comparison to C models, see Figure 7. If a model does not consider photorespiration described as oxygenase activity of RuBisCO, the difference in the export efficiency is reaching 10% (Figure 7, difference between gray lines). Furthermore, even if the time series data of particular metabolite shows the same qualitative pattern, see an example of E4P oscillation in Figure 7, the quantitative behavior demonstrates the differences in tenths of percentage. It is apparent that the content of metabolites within the CalvinBenson cycle is less sensitive to changes induced by minor sinks implementation (Figure 7) if the photorespiration is implemented as oxygenase activity of RuBisCO.
Conclusions
Due to its long history of modelling, the most studied plant metabolic system, the CalvinBenson cycle, provides a portfolio of kinetic models that can be used as a basis for further studies. What our present work reveals is that it is not possible to adapt existing and wellestablished models without significant modifications or corrections. We have shown that providing the basis for a look back at this history is important for further development in this field and essential to avoid the propagation of errors into new models generation.
In the present work we analyze two newer models, which are readily available and usable (the Poolman model and the Zhu model), as well as two fundamental models, referred to as the Hahn model and Pettersson model. Our study reveals that it is not possible to reproduce the simulations based on these fundamental models, probably due different (and unknown) computational methods employed in the original works [2, 15]. Furthermore, in the case of two newer models, we suggest corrections for these models and strongly recommend the use of two modeling tools, in our case Copasi and Matlab, as a standard approach. This approach significantly reduces the risk of systematic errors caused, e.g. by a lack of dimensional analysis.
We also discuss possible approaches for how the 3phosphoglycerate sink should be implemented in the model. We analyze it with respect to system efficiency, regulation and content of the pool of metabolites in the inner cycle. We further show that photorespiration or at least its first step (O_{2} fixation) has to be implemented in the model if this model is aimed for analyses out of the steady state  we suggest a simplified modification of the model, particularly of chemical equation for CO_{2} fixation without changing the rate equation or kinetic parameters. Finally, the minor adjacent pathways of the carbon metabolism have been analyzed and we show that they cannot be neglected and should be considered in the model.
Our study shows that modeling photosynthesis, in combination with a kinetic model and FBA together, constrains the model and we get more accurate results and better predictions. Further combinations of methods, for instance with nonsteady state metabolic control analysis in the case of detailed focus on dynamic response, can provide even more information about the properties of the model and above all about the biological system. This approach is a promising avenue for further research.
Declarations
Acknowledgements
We would like to thank to Henning Knoop from Humboldt University in Berlin for providing data for the minor sinks fluxes out of the CalvinBenson cycle on the basis of FBA for metabolic network of Synechocystic sp. PCC 6803.
This work presented here has been supported by the German Research Foundation (DFG) as part of PROMICS research group 1186.
Authors’ Affiliations
References
 Bassham JA, Krause GH: Free energy changes and metabolic regulation in steady state photosynthetic carbon reduction. Biochim Biophys Acta. 1969, 189: 207221. 10.1016/00052728(69)900486.View ArticlePubMedGoogle Scholar
 Pettersson G, RydePettersson U: A mathematical model of the Calvin photosynthesis cycle. Eur J Biochem. 1988, 175: 661672. 10.1111/j.14321033.1988.tb14242.x.View ArticlePubMedGoogle Scholar
 Zhu XG, de Sturler E, Long SP: Optimizing the distribution of resources between enzymes of carbon metabolism can dramatically increase photosynthetic rate: a numerical simulation using an evolutionary algorithm. Plant Physiol. 2007, 145: 513526. 10.1104/pp.107.103713.PubMed CentralView ArticlePubMedGoogle Scholar
 Boyle NR, Morgan JA: Flux balance analysis of primary metabolism in Chlamydomonas reinhardtii. BMC Syst Biol. 2009, 3: 4PubMed CentralView ArticlePubMedGoogle Scholar
 Knoop H, Zilliges Y, Lockau W, Steuer R: The Metabolic Network of Synechocystis sp. PCC 6803: Systemic Properties of Autotrophic Growth. Plant Physiology. 2010, 154: 410422. 10.1104/pp.110.157198.PubMed CentralView ArticlePubMedGoogle Scholar
 Montagud A, Navarro E, de Cordoba PF, Urchueguía JF, Patil KR: Reconstruction and analysis of genomescale metabolic model of a photosynthetic bacterium. BMC Syst Biol. 2010, 4: 15610.1186/175205094156.PubMed CentralView ArticlePubMedGoogle Scholar
 Fleming RMT, Thiele I, Provan G, Nasheuer HP: Integrated stoichiometric, thermodynamic and kinetic modelling of steady state metabolism. JTB. 2010, 264: 683692. 10.1016/j.jtbi.2010.02.044.View ArticleGoogle Scholar
 Tenazinha N, Vinga S: A Survey on Methods for Modeling and Analyzing Integrated Biological Networks. TCBB. 2011, 8: 943958.Google Scholar
 Laisk A, Eichelmann H, Oja V: C3 photosynthesis in silico. Photosynth Res. 2006, 90: 4566.View ArticlePubMedGoogle Scholar
 Safranek D, Cerveny J, Klement M, Pospisilova J, Brim L, Lazar D, Nedbal L: Ephotosynthesis: Webbased platform for modeling of complex photosynthetic processes. BioSystems. 2011, 103: 115124. 10.1016/j.biosystems.2010.10.013.View ArticlePubMedGoogle Scholar
 Zhu X, Alba R, de Sturler E: A simple model of the Calvin cycle has only one physiologically feasible steady state under the same external conditions. Nonlinear AnalReal. 2009, 3: 14901499.View ArticleGoogle Scholar
 Grimbs S, Arnold A, Koseska A, Kurths J, Selbig J, Nikoloski Z: Spatiotemporal dynamics of the Calvin cycle: Multistationarity and symmetry breaking instabilities. BioSystems. 2011, 103: 212223. 10.1016/j.biosystems.2010.10.015.View ArticlePubMedGoogle Scholar
 Hoops S, Sahle S, Gauges R, Lee C, Pahle J, Simus N, Singhal M, Xu L, Mendes P, Kummer U: COPASI  a COmplex PAthway SImulator. Bioinformatics. 2006, 22: 306774. 10.1093/bioinformatics/btl485.View ArticlePubMedGoogle Scholar
 Lazar D, Kana R, Klinkovsky T, Naus J: Experimental and theoretical study on high temperature induced changes in chlorophyll a fluorescence oscillations in barley leaves upon 2% CO2. Photosynthetica. 2005, 3: 1327.View ArticleGoogle Scholar
 Hahn BD: A mathematical model of leaf carbon metabolism. Annals of Botany. 1984, 54: 325339.Google Scholar
 Hahn BD: A Mathematical Model of the Calvin Cycle: Analysis of the Steady State. Annals of Botany. 1986, 57: 639653.Google Scholar
 Hahn BD: A mathematical model of photorespiration and photosynthesis. Annals of Botany. 1987, 60: 157169.Google Scholar
 Woodrow IE: Control of the rate of photosynthetic carbon dioxide fixation. Biochim Biophy Acta. 1986, 851: 181192. 10.1016/00052728(86)901246.View ArticleGoogle Scholar
 Laisk A, Eichelmann H, Oja V, Eatherall A, Walker DA: A mathematical model of the carbon metabolism in photosynthesis. Difficulties in explaining oscillations by fructose 2,6bisphosphate regulation. Proc R Soc Lond B Biol Sci. 1989, 237: 389415. 10.1098/rspb.1989.0057.View ArticleGoogle Scholar
 Woodrow IE, Mott KA: Modeling C3 photosynthesisa sensitivity analysis of the photosynthetic carbon reduction cycle. Planta. 1993, 191: 421432.View ArticleGoogle Scholar
 Poolman MG, Assmus HE, Fell DA: Applications of metabolic modelling to plant metabolism. J Exp Bot. 2004, 55: 11771186. 10.1093/jxb/erh090.View ArticlePubMedGoogle Scholar
 Poolman MG: Computer modeling applied to the Calvin cycle: PhD thesis. 1999, Oxford Brookes UniversityGoogle Scholar
 Kallas T, Castenholz RW: Internal pH and ATPADP pools in the cyanobacterium Synechococcus sp. During exposure to growthinhibiting low pH. Journal of Bacterology. 1982, 149: 229236.Google Scholar
 Portis AR, Chon CJ, Mosbach A, Heldt HW: Fructose and sedoheptulosebisphosphatase. The sites of a possible control of CO2 fixation by lightdependent changes of the stromal Mg 2+ concentration. Biochim Biophys Acta. 1977, 461: 313325. 10.1016/00052728(77)901815.View ArticlePubMedGoogle Scholar
 Stitt M, Wirtz W, Heldt HW: Metabolite levels during induction in the chloroplast and extrachloroplast compartments of spinach protoplasts. Biochim Biophys Acta. 1980, 593: 85102. 10.1016/00052728(80)900109.View ArticlePubMedGoogle Scholar
 Sprenger GA, Schorken U, Sprenger G, Sahm H: Transketolase A of Escherichia coli K12. Purification and properties of the enzyme from recombinant strains. Eur J Biochem. 1995, 230: 525532. 10.1111/j.14321033.1995.0525h.x.View ArticlePubMedGoogle Scholar
 Woodrow IE, Mott KA: Rate limitation of nonsteadystate photosynthesis by ribulose1,5bisphosphate carboxylase in spinach. Aust J Plant Physiol. 1989, 16: 487500. 10.1071/PP9890487.View ArticleGoogle Scholar
 Kana R, Kotabova E, Prasil O: Acceleration of plastoquinone pool reduction by alternative pathways precedes a decrease in photosynthetic CO2 assimilation in preheated barley leaves. Physiol Plant. 2008, 133: 794806. 10.1111/j.13993054.2008.01094.x.View ArticlePubMedGoogle Scholar
 Albe KR: Partial purification and kinetic characterization of transaldolase from Dictyostelium discoideum. Exp Mycol. 1991, 15: 25562. 10.1016/01475975(91)90027B.View ArticleGoogle Scholar
 Badger MR, Lorimer G: Interaction of sugar phosphate with the catalytic site of RuBPcarboxylase. Biochemistry. 1981, 20: 22192225. 10.1021/bi00511a023.View ArticlePubMedGoogle Scholar
 Farquhar GD: Models describing the kinetics of ribulose biphosphate carboxylaseoxygenase. Arch Biochem Biophys. 1979, 193: 456468. 10.1016/00039861(79)900523.View ArticlePubMedGoogle Scholar
 von Caemmerer S: Biochemical Models of Leaf Photosynthesis. 2000, Victoria, Australia, 128. Techniques in Plant Sciences Series. CSIRO PublishingGoogle Scholar
Copyright
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.