Modeling of Glycolytic Pathway Using in Vitro Kinetics

Haluk Hamamci, Food Engineering, Middle East Technical University, Inonu Bulvari, Ankara, Turkey, Banu Ozturk-Temur, Biotechnology Department, Middle East Technical University, Inonu Bulvari, Ankara, Turkey, and A. Palazoglu, Department of Chemical Engineering and Materials Science, University of California, Davis, One Shields Avenue, Davis, CA 95616.

The simulation of the behavior of metabolic pathways using the in-vitro obtained kinetic expressions of the constituent enzymes is one of the on-going challenges in systems biology. One could build such a simulation model using basically the data obtained in-vitro and also employ a significant number of adjustable parameters to simulate the behavior of the living cell (or one of its pathways) but the results would be difficult to generalize or interpret. On the other hand, the desirable approach would be to use the kinetic data (again obtained in-vitro) but with a minimal (if any) inclusion of adjustable parameters and systematic assumptions. One such model is the model proposed previously by Teusink et al. (2000). This model is an extremely valuable contribution, in that most of the kinetic data are gathered from the same experiment and parallel samples, resulting in a consistent model with reasonable predictive capability. This is despite the fact that the authors, admittedly, were unsuccessful in simulating the glycolysis in baker's yeast.

While we are still far from predicting any new phenomena on the basis of biochemistry-based kinetic models in living cells, simulating their dynamic behavior with the minimal use of adjustable parameters should be the immediate aim. Once this is accomplished, one can go on building on the foundations established by the particular model and improve its parameters and structure or perform biochemical investigations to explore the reasons for discrepancies between observed behaviors and model predictions.

To accomplish this task, one has to test the models rigorously under varying biochemical and physiological conditions. Such tests could include simulating the cell response under conditions when some enzymes are not expressed, or when certain environmental conditions are changed or some mutants are employed.

Using the model proposed by Teusink et al. (2000), the production of glycerol offers such a possibility. Since glycerol is an important commodity as well as being an important part of the metabolism, there is a wealth of publications in the literature for its production. These vary from the use of chemicals to induce its production to the use of specific mutants. These studies are valuable for testing and validation of the model against such data and deciding on the suitability of the model in simulating if not exactly predicting the underlying biochemical phenomena.

In this study, we have modified and tested the aforementioned model against such conditions. We were able to show that this model, with the inclusion of some kinetic data from the literature and again with the minimal use of adjustable parameters, is capable of simulating phenomena related to the production of glycerol as a branch off the glycolytic pathway. The key modification in our model is the inclusion of the glycerol branching pathway that was not quantitatively considered by Teusink et al. (2000) with kinetic expressions of the enzymes involved and the dependence especially on the NAD and NADH concentrations in the cell.

The model of Teusink et al. (2000) provided all the kinetic expressions for most of the glycolytic enzymes, and we made use of the suggestions by Rizzi et al. (1997) and also used the kinetic expressions from Cronwright et al. (2002) for the enzymes glycerol-3-phosphate dehydrogenase and glycerol 3- phosphatase which include the effect of NAD and NADH as substrate and products.

The kinetic expressions result in a set of some 25 ordinary differential equations and these were integrated in FORTTAN and MATLAB environments. The simulation results showing the time-dependent concentration profiles were compared against the fermentation data obtained from the literature for different phenotypes of yeast. The results not only indicate the expected qualitative agreement as stated by Teusink et al. (2000), but indicate that for some systems quantitative simulations are also possible.

Another goal is to test the model for other well-known phenomena like the overflow or bottleneck hypothesis at the pyruvate branch point. We expect to provide the kinetic basis for the biomass vs ethanol production as part of the broader glycolytic pathway model.

As the metabolomics techniques develop, one will be able to test in vitro information with better in vivo data. However, one would also need better models to start with. We believe that the model proposed by Teusink et al. (2000) and their approach appear to be the best starting point for this purpose, as will be demonstrated by this study.

Cronwright GR. et al., Metabolic control analysis of glycerol synthesis in Saccharomyces cerevisiae, Applied and Environmental Microbiology, 68, pp. 4448-4456, (2002).

Rizzi M. et al., In vivo analysis of metabolic dynamics in Saccharomyces cerevisiae: II. Mathematical Model, Biotechnology and Bioengineering, 55, pp. 592-608, (1997).

Teusink B, et al., Can yeast glycolysis be understood in terms of in vitro kinetics of the constituent enzymes? Testing biochemistry, Eur. J. Biochem., 267, pp. 5313-5329, (2000).