Modeling of Metabolic Networks
ID: 3933

Abstract: Andre Canelas

Existing modelling techniques for metabolic reaction networks are based almost exclusively on network stoichiometry so their application is somewhat limited and restricted to stationary conditions. The ultimate “virtual cell” model should also include knowledge on the thermodynamics (which determine the directionality) and kinetics (which determine the rate) of the individual reactions. Such a model would allow us to predict in vivo network behaviour in a wider range of stationary and dynamic conditions and constitute an decisive tool for the selection of modification targets for improved production.

Analysis of network thermodynamics requires knowledge of equilibrium constants, which can be determined in vitro. While for a few reactions ample literature is available, for most it can be quite scarce, or absent. Even if reliable data were available under well-defined in vitro conditions, that wouldn’t ensure accurate prediction of reaction directionality in vivo since the environment surrounding the enzymes (pH, ionic strength, etc) is in most cases largely uncharacterized. In the case of reaction kinetics the challenge is even greater since binding properties are enzyme-specific (hence species- and strain-dependent). Models constructed with kinetic parameters obtained in vitro have so far failed to accurately predict in vivo behaviour. In our experience, the in vitro kinetics can often not even explain flux changes qualitatively, much less quantitatively. What is needed is a methodology for accurate determination of in vivo thermodynamic and kinetic reaction parameters.

We shall present a novel approach we have taken to tackle these challenges in S. cerevisiae, an important model organism of major industrial relevance. By using chemostat cultures over a wide range of dilution rates it was possible to vary the flux through the metabolic network by more than 40-fold, and using improved rapid sampling techniques (1) and analytical methods (2,3) we could obtain an accurate, comprehensive metabolome dataset covering glycolysis, TCA cycle, PPP pathway and redox and energy metabolism. From these data we observed broadly two types of reaction rate behaviour. For reactions thought to operate near equilibrium the rate seems to be regulated mainly thermodynamically, that is, by the displacement of the mass action ratio in relation to the equilibrium constant. Furthermore, from the relation between fluxes and mass action ratios it was possible to determine the in vivo equilibrium constants for 16 reactions, some of which differ significantly from the corresponding in vitro values, and which henceforth may be used to better predict reaction directionality. With allosterically-regulated reactions we found that in most cases the changes in flux correlated with changes in known effectors, even though reaction rates could not be quantitatively predicted from in vitro kinetic parameters. This confirms the need of determining in vivo kinetic parameters and at the same time suggests that in many cases it will be possible to do so.

(1) Canelas et al (2008) Metabolomics, 4, 3, 226-239

(2) Canelas et al (2008) Biotech Bioeng, 100, 4, 734-743

(3) Cipollina et al, A comprehensive method for the quantification of the intracellular pentose phosphate pathway intermediates in Saccharomyces cerevisiae by GC-MS, submitted



Emrah Nikerel

Construction of large scale kinetically based dynamic models for microbial metabolic networks is one of the central issues in both understanding the complexity beyond intuitive comprehension and perform predictions for gene targets to improve the performance of the organism. However, construction of such models is often hampered by a number of challenges e.g. data availability, compartmentalization, parameter identification coupled to design of in vivo perturbations and scalability to genome scale models. Hence, an experimental and modeling approach that provides solutions to the mentioned challenges is of general interest.

Short-term (<300 sec) perturbation experiments [1] are becoming more and more common practice and have proven to provide insights on the in vivo kinetic properties of metabolic pathways. This approach consists of perturbing a well-characterized steady state by an external stimuli and subsequently monitoring the response of the intra- and extracellular metabolites over a short period of time. During this time, it may be assumed that enzyme concentrations and activities do not change, allowing the observed responses to be attributed to kinetic interactions at the metabolome level alone.

Considering the mentioned challenges, we have developed a modeling framework, to analyze such transient metabolome data which allows building large scale dynamic model for metabolic reaction networks. In short, the proposed modeling platform begins by determining the steady state flux distribution using the data prior to the stimuli and the known stoichiometry. For the rate expressions, approximative linlog kinetics is used, which allows the enzyme-metabolite kinetic parameters to be represented by an elasticity matrix. A first uniqe feature is that linlog kinetics allows perform model reduction, based on time-scale analysis. A second feature is a priori parameter identifiability analysis allowing the information content of the experimental data to be assessed, as described in [2]. The final values of the elasticities are estimated by fitting the model to the available short term kinetic response data described as an in silico study in [3].

In this presentation, we will firstly show the application of our modeling platform on experimental cases involving 2 industrially relevant organism: (i) the in vivo kinetic modeling of glycolysis in anaerobically cultivated Saccharomyces cerevisiae producing ethanol from glucose and (ii) the in vivo kinetic modeling central metabolism and penicillin production from glucose in aerobically cultivated Penicillin chrysogenum cells.

References

[1] Theobald U, Mailinger W, Baltes M, Rizzi M, Reuss M.., 1997, Biotech Bioeng, 55(2), 305-316

[2] Nikerel IE, van Winden WA, Verheijen PJT, Heijnen JJ, 2009, Metab. Eng., 11(1), 20-30

[3] Nikerel IE, van Winden WA, van Gulik WM, Heijnen JJ, 2006, BMC Bioinformatics, 7:540



Donald Rose

Metabolomics, unbiased global biochemical profiling, is an approach for obtaining a “snapshot” of the metabolism of a population of cells in a bioreactor. As such, it has had particular utility in the area of upstream bioprocess optimization of biopharmaceutical production. This presentation will discuss several case studies from biopharmaceutical collaborations, describing how metabolomics was used to optimize both cell line development and media composition. These case studies will show the monitoring of hundreds of biochemicals and metabolites, found both inside the cell and in the spent media. Using this approach, biopharmaceutical groups have been able to 1) rapidly identify media components which are rate limiting, 2) determine toxic metabolites which build up in either the cell or the spent media, and 3) optimize growth conditions for greatest yield.







Moderator: Douglas Lee, Metabolon, Inc (United States)

Presenter 1: In Vivo Thermodynamic and Kinetic Modeling of the Central Metabolism of Microbial Production Hosts
Andre Canelas, Kluyver Centre for Genomics of Industrial Fermentation, (Netherlands)  [Confirmed]

Presenter 2: Kinetic Network Modeling of Industrial Production Organisms Using Stimulus Response Data 
Emrah Nikerel, Kluyver Centre for Genomics of Industrial Fermentation, (Netherlands)  [Confirmed]

Presenter 3: Metabolomics: A Powerful Companion for Bioprocess Optimization  
Douglas Lee, Metabolon, Inc, (United States)  [Confirmed]

Presenter 4 (if necessary): {Presenter4PresentationTitle} 
{Presenter4FirstName} {Presenter4LastName}, {Presenter4CompanyName}, ({Presenter4Country})  [{Presenter4InviteStatus}]

Panel Organizer:
Matthew Carr, Biotechnology Industry Organization, (United States)

Why should your submission should be selected for this year’s program?