Genome Informatics 14: 23–33 (2003)
23
MetaFluxNet, a Program Package for Metabolic Pathway Construction and Analysis, and Its Use in Large-Scale Metabolic Flux Analysis of Escherichia coli Sang Yup Lee1,2,3
Dong-Yup Lee1,2
Soon Ho Hong1,2
[email protected]
[email protected]
[email protected]
Tae Yong Kim1,2
Hongsoek Yun2
Young-Gyun Oh2
Sunwon Park2
[email protected]
[email protected]
[email protected]
[email protected]
1 2 3
Metabolic and Biomolecular Engineering National Research Laboratory Department of Chemical and Biomolecular Engineering and BioProcess Engineering Research Center Department of BioSystems and Bioinformatics Research Center, Korea Advanced Institute of Science and Technology, 373-1 Guseong-dong, Yuseong-gu, Daejeon 305701, Republic of Korea Abstract
We have developed MetaFluxNet which is a stand-alone program package for the management of metabolic reaction information and quantitative metabolic flux analysis. It allows users to interpret and examine metabolic behavior in response to genetic and/or environmental modifications. As a result, quantitative in silico simulations of metabolic pathways can be carried out to understand the metabolic status and to design the metabolic engineering strategies. The main features of the program include a well-developed model construction environment, user-friendly interface for metabolic flux analysis (MFA), comparative MFA of strains having different genotypes under various environmental conditions, and automated pathway layout creation. The usefulness and functionality of the program are demonstrated by applying to metabolic pathways in E. coli. First, a large-scale in silico E. coli model is constructed using MetaFluxNet, and then the effects of carbon sources on intracellular flux distributions and succinic acid production were investigated on the basis of the uptake and secretion rates of the relevant metabolites. The results indicated that among three carbon sources available, the most reduced substrate is sorbitol which yields efficient succinic acid production. The software can be downloaded from http://mbel.kaist.ac.kr/.
Keywords: metabolic flux analysis, MetaFluxNet, Escherichia coli, in silico simulation
1
Introduction
Understanding complex biological functions requires analysis or modeling of large metabolic reaction networks. Several available approaches for such analysis or modeling include structural (topological) pathway analysis, metabolic flux analysis (MFA), metabolic control analysis, and dynamic simulation. Among them, MFA is most widely adopted for rational design and in silico engineering of metabolic pathways: the only information required is the stoichiometry of metabolic reactions and mass balances around the metabolites under pseudo-steady state, or stationary, assumption [12, 18]. A major effort is required to establish a user-friendly computer program to implement MFA easily. Such an effort was initiated by LabVIEW [16]. Since then, appreciable progress has been made through the development of a number of programs. They include FBA [23], FluxAnalyzer [9], Fluxmap [24], INSILICO discovery [25], and Metabologica [27]. Nevertheless, most of them are not particularly attractive to date since the softwares available are not likely to be sufficient to have desirable features for MFA. Apart from flux calculations, they should be able to satisfy various purposes in MFA,
24
Lee et al.
strain.
Figure 1: Screen shot of the model composers for reactions (left) and metabolites (right). e.g., obtaining maximum theoretical yields, examining the influence of genetic modifications on the flux distributions [14], and classifying the metabolic system with measurements according to the system determinacy [18, 8]; moreover, it is highly desirable that the software would provide dynamic visualization of metabolic maps for interactive and comparative analysis of metabolic flux profiles under different conditions. Hence, MFA tool should provide an integrated environment satisfying the aforementioned various computational demands in the flux analysis. Consequently, we have developed a stand-alone program package, MetaFluxNet (version 1.6) which provides an easy and customized way for constructing a metabolic reaction system and performing MFA [10]. In this study, newly upgraded version (version 1.7) of MetaFluxNet is introduced and its application to a large-scale model is presented for in silico simulation of metabolic pathways in E. coli. What follows are the main features of MetaFluxNet. Subsequently, some important features and the usefulness of the program are demonstrated by applying to a large-scale in silico E. coli model. On the basis of secretion measurements of several metabolic products, the effect of various carbon sources on intracellular flux distributions and succinic acid production in E. coli is investigated and some noteworthy observations derived therefrom are discussed.
2
Methods
2.1
MetaFluxNet
MetaFluxNet (version 1.7) provides with an easy and customized environment for constructing a metabolic reaction network and for performing MFA and dynamic visualization of the MFA results [10]. Furthermore, it provides the graphical user interface (GUI) where calculated flux distributions or other computed results are analyzed dynamically as well as interactively. The main features of the MetaFluxNet are as follows. Construction of a metabolic reaction model MetaFluxNet provides a well-developed model construction environment. It allows the users to set up their own model systems in two ways: • Model composer: user-defined way by registering information on two object classes, Metabolites and Reactions, which are interactively linked in a metabolic system. Each class consists of several fields describing biological information (Fig. 1). For example, the entries of the Reactions class contain Enzyme Commission (EC) number, gene name, and substrates and products participating in the reaction. Data contents in the fields can be edited, stored and modified individually
MetaFluxNet
25
Figure 2: Query interface of the integrated metabolic database supported in MetaFluxNet. The current database system contains 3166 enzymatic information and 5039 compounds supporting the 121 pathway groups in 89 organisms. They are consolidated from three databases: ENZYME [22], LIGAND [26] and EcoCyc [21]. by the users. • Database query: reference retrieved way by taking up data from the Enzyme DB system. For this purpose, the original flatfiles of three main metabolic databases, ENZYME [22], LIGAND [26] and EcoCyc [21] are parsed and stored in our pathway database system according to the defined data structure. Thus, the reaction model can be easily constructed by importing the results of a wide range of SQL queries supported in the retrieval system and by mapping them into the current model project in MetaFluxNet. Fig. 2 shows the DataBase query interface where the list of metabolic reactions is retrieved as the result of a query, “find all reactions involved in the glycolysis pathway of E. coli K-12 strain”. Metabolic flux analysis Once the metabolic reaction model is constructed, a stoichiometric model is defined under pseudosteady state, or stationary, assumption on the basis of measured reaction rates or fluxes. In such a model, the relationships among all metabolites (intermediates) and reactions are balanced in terms of stoichiometry as follows: Sv = Sm vm + Sc vc = 0 (1) where S = stoichiometric matrix with K metabolites and J reactions [K×M] v = flux vector [J×1] Sm = stoichiometric matrix with K metabolites and M measured fluxes [K×M] Sc = stoichiometric matrix with K metabolites and C measured fluxes [K×C] vm = measured fluxes [M×1] vc = fluxes to be calculated (non-measured fluxes) [C×1] Then, the defined system can be classified according to Determinacy and Redundancy depending on the rank of Sc , which leads to four possible cases by the combination of the assignment of each classification property [8].
26
Lee et al. • Determinacy: determined (rank(S c )=C)/underdetermined(rank(Sc )< C) • Redundancy: nonredundant (rank(S c )=K)/redundant(rank(Sc )< K)
For each case, the corresponding procedure is applied to analyze the system and determine the flux distribution (Fig. 3). For example, in the case of the determined system (rank(S c )=C), a unique solution (Eq.(2)) for nonredundant case or a least-squares solution (Eq.(3)) for redundant case can be achieved by matrix operations if the system is observable. vc = −S−1 c Sm vm vc =
−S# c Sm vm
(2) (3)
where S# c denotes the Penrose pseudo-inverse [1]. Otherwise, gross measurement errors can be detected by a variance-covariance matrix, and subsequently measured fluxes are reconciled to remove the inconsistency in the case of the redundant system as delineated by Stephanopoulos et al. [18]. It is then followed by inspecting calculable fluxes which can be uniquely determined by the least-squares solution using the pseudo-inverse (see [8]). If the resultant balanced reaction model is underdetermined in calculating the flux distribution due to insufficient measurements or to constraints, the unknown fluxes within the metabolic reaction network are evaluated by means of flux balance analysis (FBA) based on linear programming (LP), subject to the constraints pertaining to mass conservation, reaction thermodynamics, and capacity as described elsewhere [15, 18, 20]. Comparative flux analysis MetaFluxNet has a capability of investigating the influences of gene addition or deletion, and of varying cultivation conditions on the metabolic flux distribution. The MFA results for such various scenarios under different conditions can be displayed in one window to compare them easily, where specified measurements and gene modification (addition or deletion) are represented by ‘measured’ and ‘added or deleted’, respectively, in the state field of fluxes, while the states of non-measured metabolic fluxes are categorized by ‘calculated’ and ‘bound’ (Fig. 3b). In addition, ‘plot view’ function visualizes the profile of internal flux distributions on different conditions. Using these features of MetaFluxNet, the sensitivity of the maximal growth flux to specific fluxes in the metabolic network can be investigated to quantify the relation between flux levels and optimal cellular growth [5]; the mutant objective function for each gene deletion or addition can be calculated interactively and compared with each other, thereby identifying essential genes required for the desired condition [4]; flux enhancements can be predicted via foreign gene recombination [3]. Thus, this scenario analysis or comparative flux analysis renders it possible to comprehend and interpret the metabolic and physiological changes of cell under different conditions, and consequently to design new metabolic engineering strategies to achieve desired goals. In the next upgrade, various uncertainty analysis methods, e.g., sensitivity analysis, Monte Carlo analysis and parameter variations will be implemented to evaluate effects of uncertain parameters or measurements in the model. Currently, this scenario-based analysis can be alternatively employed for such purpose by comparing the results of different parameter values. Visualization of reaction pathways and flux distributions MetaFluxNet provides an interactive and dynamic graphical user interface to display metabolic reaction pathways with flux distribution results (Fig. 3c). The pathways can be automatically and dynamically visualized by the spring embedder layout algorithm supported in the program. In the near future, this program will be upgraded to integrate the structural pathways analysis based on the graph theory [17] and dynamic simulation for comprehensive metabolic network modeling and simulation.
27
MetaFluxNet
conservation, reaction thermodynamics, and capacity as described elsewhere [2,16,17].
(a)
(b)
(c)
Figure 3: Screen shots of the flux analysis part of MetaFluxNet. The C. lipolytica model is from the reference [18]. (a) It provides interactive and customized environment for metabolic flux analysis of the model. (b) Metabolic flux profiles of interest under three different conditions (scenarios) can be compared through the plot view page. (c) Metabolic Flux distributions can be interactively determined and dynamically visualized via user-friendly interface.
28
2.2
Lee et al.
Application: the Effect of Various Carbon Sources on Succinic Acid Production in E. coli Using MetaFluxNet
Succinic acid, a member of C4 -dicarboxylic acid family, has been used in many industrial applications including surfactant, ion chelator, food additive, and supplement to pharmaceuticals, antibiotics and vitamins. Especially, it is known as an intermediate of several green chemicals and materials. Conventionally, succinic acid has been produced by chemical processes. Recently, much effort is being exerted for its production by microbial fermentation using renewable feedstocks, due to pollution problems associated with chemical processes [11]. E. coli produces several metabolic products by fermentation: acetic acid, ethanol, formic acid, lactic acid, and also small amount of succinic acid [7]. The ratio of these fermentation products varies depending on the culture condition employed. In this study, we examined the effect of various carbon sources on intracellular flux distributions and succinic acid production based on the secretion measurements of the relevant metabolites in recombinant E. coli. Carbon substrates examined include glucose, gluconate and sorbitol which have different reducing potentials. Thus, the effect of reducing power on the intracellular fluxes can be evaluated by MetaFluxNet. Model construction and metabolic flux analysis The in silico E. coli model was constructed using the model composer and database query supported in MetaFluxNet. Fig. 4 exhibits the central metabolic pathways of the E. coli model. This model incorporates 189 reversible and 126 irreversible reactions, and 280 metabolites. For the estimation of intracellular flux distribution, constraints were provided based on the fermentation results [6]. Experiments for various carbon sources were carried out under anaerobic condition to investigate the effect of each carbon source on the production of succinic acid. They include glucose, sorbitol and gluconate, for each of which secretion rates of metabolic products are given in Table 1. Thus, fueling of the metabolic network is rendered possible by a constrained amount of each carbon source, along with secretion measurements of relevant metabolites. System analysis of the constructed model by MetaFluxNet resulted in the underdetermined system. Accordingly, flux distribution can be determined by LP with the objective function of maximum growth rate [20] which is quantified by cellular contents of macromolecules [13]. Table 1: Constraints on the rates of metabolites uptake and excretion for different carbon sources. Constraints (mM/gDCW/h) Sorbitol Glucose Gluconate Substrate uptake 3.17 (1.00)* 6.76 (1.00) 16.64 (1.00) Succinic acid excretion 2.26 (0.71) 1.71 (0.25) 1.60 (0.10) Malic acid excretion 0.74 (0.23) 7.30 (1.08) 9.20 (0.55) Acetic acid excretion 0.85 (0.27) 0.00 (0.00) 11.72 (0.64) Lactic acid excretion 1.50 (0.47) 0.82 (0.12) 0.52 (0.03) ∗ Values in parenthesis are normalized with respect to the substrate uptake.
3
Results and Discussion
The effect of carbon substrate on the intracellular metabolic fluxes and succinic acid production in recombinant E. coli was evaluated by MFA. Metabolic fluxes under succinic acid producing conditions were determined for three different carbon substrates such as sorbitol, glucose and gluconate (Figs. 5 and 6). One mole of sorbitol produces six moles of [H] during its conversion to two moles of PEP while glucose and gluconate produce four and two moles of [H], respectively. When gluconate was
MetaFluxNet
29
Figure 4: Overview of the central metabolic pathways of E. coli model. Excretion rates of four metabolic products including succinic acid, acetic acid, malic acid, and lactic acid for each carbon source are given in Table 1.
Lee et al.
(a)
(b)
(c)
30
Figure 5: Normalized flux distributions of the central metabolic pathways in E. coli model for different carbon sources: (a) gluconate (b) glucose (c) sorbitol. The thickness of the arrow is proportional to the value of normalized flux.
31
MetaFluxNet
utilized as a carbon substrate, it was predicted that about 8.3% of total carbon flux was directed into the succinic acid pathway, and most of it was supplied through the malic enzyme. Next, the intracellular flux distribution in recombinant E. coli was evaluated for the glucose uptake. Significant amount of carbon flux (20.5%) was directed into the succinic acid pathway. Again, it is noteworthy that the malic enzyme flux was highly activated under the succinic acid producing condition, which is in accord with the available results suggesting pyruvate carboxylation as an optimal succinic acid production pathway [19]. Finally, flux analysis result for sorbitol uptake indicated that most of carbon flux (83.8%) was directed into succinic acid production pathway and most of other intracellular fluxes were severely reduced.
132465 . 2 & 7 & 2 ,
7 & 2 8.9,
!" #
$%'& () #
*+-,"./ ! #
0 './ () #
Figure 6: Yields of metabolites from anaerobic cultures with different carbon sources for the succinic acid producing strains. When comparing three flux analysis results, it was found that the succinic acid flux and yield were increased in the order of gluconate, glucose and sorbitol (Figs. 5 and 6). Moreover, consumption rates of NADH and FADH2 , which are required for the conversion of oxaloacetate to malate and fumarate to succinic acid, respectively, were also increased in the same order. This suggests that sorbitol can produce more reducing equivalents in the forms of NADH and FADH 2 than other carbon sources in the reducing power balance. Note that the stoichiometric coefficient, in conjunction with the flux vector, v, determined by FBA, leads to the formulation of the flux capacity for intermediate i in Eq. (4), whereby consumption rates of intermediates can be quantified and displayed in MFA result of MetaFluxNet (Fig. 3a). ci = where
X
|Sij vj |/2
∀i ∈ K
(4)
j∈J
i, j = metabolite and reaction, respectively K, J = sets of metabolites and reactions, respectively vj = flux of reaction j Sij = stoichiometric coefficient of metabolite i in reaction j ci = flux capacity of metabolite i Besides succinic acid production, production rates of other organic acids were also influenced by reducing power. Least reduced carbon substrate, gluconate, yielded highest production rate of acetic acid because reducing power is not required for the production of acetic acid from pyruvate (Fig. 5). When glucose was fueled as a carbon substrate, lactic acid, which requires NADH during conversion from pyruvate, was also produced as well as acetic acid. Whereas, the most reduced carbon substrate, sorbitol gave rise to highest production rate of the succinic acid as a major product: each mole of NADH and FADH2 is required for the conversion of one mole of succinic acid from pyruvate.
32
Lee et al.
Consequently, it can be mentioned that product profile can be manufactured by the control of reducing power, and succinic acid productivity could be enhanced if balance of reducing was achieved by supply of additional reducing power as predicted in previous study [6]. In this study, we have presented the main features of MetaFluxNet which is the integrated program package for managing information on the metabolic reaction network and for quantitatively analyzing metabolic fluxes in an interactive and customized way. The usefulness and functionality of the program were demonstrated through its application to the large-scale in silico E. coli model. The effects of carbon sources on intracellular flux distributions and succinic acid production were investigated on the basis of the uptake and secretion rates of the relevant metabolites. The results indicated that sorbitol is the most reduced carbon substrate leading to the efficient succinic acid production, which is in agreement with the experiments [2].
Acknowledgments This work was supported by the National Research Laboratory Program (2000-N-NL-01-C-237) of the Ministry of Science and Technology (MOST), the Advanced Backbone IT Development Project (IMT2000-C3-1) of the Ministry of Information and Communication (MIC) and MOST, and by the Brain Korea 21 project from the Ministry of Education.
References [1] Ben-Israel, A. and Greville, T.N.E., Generalized inverses: Theory and applications, John Wiley & Sons, New York, 1974. [2] Berr´ıos-Rivera, S.J., San, K.-Y., and Bennett, G.N., The effect of carbon sources and lactate dehydrogenase deletion on 1,2-propanediol production in Escherichia coli, J. Ind. Microbiol. Biotechnol., 30:34–40, 2003. [3] Burgard, A.P. and Maranas, C.D., Probing the performance limits of the Escherichia coli metabolic network subject to gene additions or deletions, Biotechnol. Bioeng., 74:364–375, 2001. [4] Edwards, J.S. and Palsson, B.O., The Escherichia coli MG1655 in silico metabolic genotype: Its definition, characteristics, and capabilities, PNAS, 97:5528–5533, 2000. [5] Edwards, J.S. and Palsson, B.O., Robustness analysis of the Escherichia coli metabolic network, Biotechnol. Prog., 16:927–939, 2000. [6] Hong, S.H. and Lee, S.Y., Importance of redox balance on the production of succinic acid by metabolically engineered Escherichia coli, Appl. Microbiol. Biotechnol., 58:286–290, 2002. [7] Hong, S.H. and Lee, S.Y., Metabolic flux analysis for succinic acid production by recombinant Escherichia coli with amplified malic enzyme activity, Biotechnol. Bioeng., 74(2):89–95, 2001. [8] Klamt, S., Schuster, S., and Gilles, E.D., Calculability analysis in underdetermined metabolic networks illustrated by a model of the central metabolism in purple nonsulfur bacteria, Biotechnol. Bioeng., 77:734–751, 2002. [9] Klamt, S., Stelling, J., Ginkel, M., and Gilles, E.D., FluxAnalyzer: exploring structure, pathways, and flux distributions in metabolic networks on interactive flux maps, Bioinformatics, 19:261–269, 2003. [10] Lee, D.-Y., Yun, H., Park, S., and Lee, S.Y., MetaFluxNet: the management of metabolic reaction information and quantitative metabolic flux analysis, Bioinformatics, in press.
MetaFluxNet
33
[11] Lee, P.C., Lee, W.G., Kwon, S., Lee, S.Y., and Chang, H.N., Succinic acid production by Anaerobiospirillum succiniciproducens: effects of the H2/CO2 supplying and glucose concentration, Enzyme Microb. Technol., 24:549–554, 1999. [12] Lee, S.Y. and Papoutsakis, E.T., Metabolic engineering, Marcel Dekker, New York, 1999. [13] Neidhardt, F.C., Curtiss, R., Ingraham, J.L., Lin, E.C.C., Low, K.B., Magasanik, B., Reznikoff, W.S., Riley, M., Schaechter, M., and Umbarger, H.E., Escherichia coli and Salmonella, ASM press, Washington D.C., 1996. [14] Nissen, T.L., Schulze, U., Nielsen, J., and Villadsen, J., Flux distributions in anaerobic, glucoselimited continuous cultures of Saccharomyces cerevisiae, Microbiology, 143:203–218, 1997. [15] Pramanik, J. and Keasling, J.D., Stoichiometric model of Escherichia coli metabolism: incorporation of growth-rate dependent biomass composition and mechanistic energy requirements, Biotechnol. Bioeng., 56:398–421, 1997. [16] Regan, L. and Gregory, M., Flux analysis of microbial metabolic pathways using a visual programming environment, J. Biotechnol., 42:151–161, 1995. [17] Seo, H., Lee, D.-Y., Park, S., Fan, L.T., Shafie, S., Bert´ok, B., and Friedler, F., Graph-theoretical identification of pathways for biochemical reactions, Biotechnol. Lett., 23:1551–1557, 2001. [18] Stephanopoulos, G.N., Aristidou, A.A., and Nielsen, J., Metabolic engineering, Academic Press, San Diego, 1998. [19] Stols, L. and Donnelly, M.I., Production of succinic acid through overexpression of NAD + dependent malic enzyme in an Escherichia coli mutant, Appl. Environ. Microbiol., 63(7):2695– 2701, 1997. [20] Varma, A. and Palsson, B.O., Metabolic flux balancing: basic concepts, scientific and practical use, Nat. Biotech., 12:994–998, 1994. [21] EcoCyc: http://ecocyc.org/ [22] ENZYME: http://www.expasy.ch/enzyme/ [23] FBA: http://gcrg.ucsd.edu/downloads/index.html [24] Fluxmap: http://www.biotecnol.com/ [25] INSILICO discovery: http://www.insilico-biotechnology.com/products_en.html [26] LIGAND: http://www.genome.ad.jp/ligand/ [27] Metabologica: http://www.metabologica.com/
