Báo cáo khoa học: Reconstruction of the central carbon metabolism of Aspergillus niger pot

Reconstruction of the central carbon metabolism of Aspergillus niger Helga David, Mats A ˚ kesson and Jens Nielsen Center for Process Biotechnology, BioCentrum-DTU, Building 223, Technical University of Denmark, DK-2800 Kgs. Lyngby, Denmark The topology of central carbon metabolism of Aspergillus niger was identified and the metabolic network reconstruc- ted, by integrating genomic, biochemical and physiological information available for this microorganism and other related fungi. The reconstructed network may serve as a valuable database for annotation of genes identified in future genome sequencing projects on aspergilli. Based on the metabolic reconstruction, a stoichiometric model was set up that includes 284 metabolites and 335 reactions, of which 268 represent biochemical conversions and 67 represent trans- port processes between the different intracellular compart- ments and between the cell and the extracellular medium. The stoichiometry of the metabolic reactions was used in combination with biosynthetic requirements for growth and pseudo-steady state mass balances over intracellular metabolites for the quantification of metabolic fluxes using metabolite balancing. This framework was employed to perform an in silico characterisation of the phenotypic behaviour of A. niger grown on different carbon sources. The effects on growth of single reaction deletions were assessed and essential biochemical reactions were identified for different carbon sources. Furthermore, application of the stoichiometric model for assessing the metabolic capabilities of A. niger to produce metabolites was evaluated by using succinate production as a case study. Keywords: Aspergillus niger; metabolic reconstruction; flux balance analysis; functional genomics. Filamentous fungi are important organisms for the pro- duction of industrial enzymes, speciality chemicals and pharmaceuticals. Moreover, they play a major role for human welfare as agents of biodegradation, spoilage and decay, and some filamentous fungi act as pathogens for humans, animals and plants, being responsible for a large number of deaths and substantial losses in the agricultural sector annually. For these reasons it has been decided recently to sequence several species of filamentous fungi (http://gene.genetics.uga.edu/white_papers/anidulans.html). In order to identify possible targets for drugs that may treat medical mycoses and to identify better fungicides that may prevent biodeterioration and against plant pathogenicity in agriculture, it will be of significant value to reconstruct the map of fungal metabolism, as this will give new insight into cellular function. A metabolic map will also be very useful for the design of improved producing strains that may then be constructed through metabolic engineering [1]. The budding yeast, Saccharomyces cerevisiae, is probably the best understood fungus and the eukaryotic model organism par excellence. Even though it represents a valuable starting point, yeast is not an adequate model for analysing overall cell function of filamentous fungi as the latter exhibit more genes and larger genomes, endowing them with more extensive metabolic capabilities. A key model system for filamentous fungi is Aspergillus nidulans, for which many specific mutants are available. Furthermore, several species of aspergilli are of industrial importance, as producers of a wide array of products that range from metabolites, such as organic acids (e.g. citric acid, A. niger [2]) and polyketides (e.g. statins, A. terreus [3]), to proteins, both homologous (e.g. a-amylase, A. oryzae [4]) and heterologous (e.g. human interferon, A. nidulans [5]). Besides their industrial relevance, some species of aspergilli can cause infections in humans and animals, namely allergic or invasive bronchopulmonary aspergillosis (A. fumigatus, A. terreus), pulmonary aspergilloma (A. niger) and sinusitis (A. flavus) ([6], http://www.Aspergillus.man.ac.uk). Further- more, powerful genetic, biochemical and molecular bio- logical techniques are available for analysis of cellular function in these organisms, and introduction of directed genetic modifications in aspergilli may hereby be used to design efficient cell factories through metabolic engineering for production of different industrially important products in the future [1,7,8]. Correspondence to J. Nielsen, Center for Process Biotechnology, BioCentrum-DTU, Building 223, Technical University of Denmark, DK-2800 Kgs. Lyngby, Denmark. Fax: + 45 4588 4148, Tel.: + 45 4525 2696, E-mail: jn@biocentrum.dtu.dk Enzymes: transketolase (EC 2.2.1.1); NADPH-dependent L -xylulose reductase (EC 1.1.1.10); glutamine-fructose-6-phosphate transaminase (isomerising; EC 2.6.1.16); and chitin synthase (EC 2.4.1.16); gluco- samine-phosphate N-acetyltransferase (EC 2.3.1.4); phosphoacetyl- glucosamine mutase (EC 5.4.2.3); UDP-N-acetylglucosamine pyrophosphorylase (EC 2.7.7.23); D -xylulose reductase (NADH- and NADPH-dependent) (EC 1.1.1.9); mannitol-2-dehydrogenase (NADP + -dependent) (EC 1.1.1.138); NADP + -dependent isocitrate dehydrogenase (EC 1.1.1.42); pyruvate decarboxylase (EC 4.1.1.1); ATP/citrate oxaloacetate-lyase (EC 4.1.3.8). (Received 4 June 2003, revised 14 August 2003, accepted 20 August 2003) Eur. J. Biochem. 270, 4243–4253 (2003) Ó FEBS 2003 doi:10.1046/j.1432-1033.2003.03798.x Several efforts have been made for the understanding and the quantitative description of the metabolism of A. niger during citric acid producing conditions. In concrete, experimental techniques, such as 13 C-NMR analysis [9], and modelling strategies, namely metabolic flux analysis [10] and other mass balance and energy balance techniques [11], as well as biochemical system theory [12], have been applied to quantitatively describe citric acid production and assist in the design of improved producing strains. However, no comprehensive model is available for the analysis of the central carbon metabolism of this microorganism, which is essential for a rational optimisation approach. Lately, detailed metabolic models, largely based on genomic sequence information, have been developed for microorganisms whose genomes have been sequenced and annotated. The modelled organisms include the prokary- otes Haemophilus influenzae [13], Escherichia coli [14], Helicobacter pylori [15] and most recently the eukaryote, Saccharomyces cerevisiae [16]. In spite of, or rather because of, its economic importance, there is currently no publicly available genome sequence for A. niger; however, although scattered, there is a considerable amount of biological knowledge in the literature. In this work, we present a comprehensive reconstruction of the central carbon metabolism of A. niger that served as a basis for the development of a detailed stoichiometric model consisting of 335 reactions and 284 metabolites distributed over three intracellular compartments (cytosol, mitochondria and glyoxysomes) and the extracellular medium. The metabolic model was used for the quanti- fication of fluxes through the branches of the metabolic network. Herein, metabolite balancing was applied in combination with linear optimisation methods to perform an in silico characterisation of the phenotypic behaviour of A. niger, under different environmental and genetic con- ditions, and to investigate its biochemical capabilities for metabolite production. Materials and methods Computational protocol The quantification of metabolic fluxes was accomplished using metabolite balancing. Reactions from the metabolic reconstruction were incorporated into a stoichiometric model that consisted of a set of algebraic equations representing material balances over intracellular metabolites in the metabolic network, assuming pseudo-steady state in the metabolite concentrations and negligible dilution effects from growth [17–19]. The stoichiometric model is conveni- ently represented in matrix form as S Á v ¼ 0 where, the matrix S contains the stoichiometric coeffi- cients and the vector v represents the fluxes in the metabolic reactions. As the number of reactions is typically greater than the number of intracellular metabolites, the system of equations comprising the stoichiometric model is underdetermined and an infinity of feasible flux distributions exists. In computational studies, a particular flux distribution can be found by formulating a suitable objective function and using linear optimisation [20], often referred to as flux balance analysis [21]. The linear programming problem was formulated as max z ¼ c T Áv s.t. SÁv ¼ 0 a i v i b i where, the vector c specifies the importance of the individual fluxes in the objective z. The linear inequali- ties (a i ,b i ) are used to define additional constraints on the individual fluxes, including information on reversi- bility, measured substrate uptake or product formation rates, etc. In what concerns reversibility, fluxes corres- ponding to reversible reactions are allowed to be positive or negative, whereas fluxes corresponding to irreversible reactions can only have positive values. Unless otherwise stated, growth was simulated by opti- misation of flux to biomass for a specified uptake rate of a selected carbon source. Other substrates, such as ammonia, sulphate, phosphate, oxygen, etc., could be taken up freely. All major metabolic products (carbon dioxide, organic acids, alcohols, amino acids, etc.) were allowed to be excreted. Linear programming calculations were performed using commercially available software, LINDO CALLABLE LIBRARY (Lindo Systems, Inc., Chicago, IL, USA) and OPTIMIZATION TOOLBOX in MATLAB 6.1 (The Mathworks, Inc.). Results and discussion Reconstruction process The metabolic reconstruction aims at depicting a detailed description of the central carbon metabolism of A. niger, namely of the metabolism of carbohydrates, organic acids, polyols and other alcohols, and amino-sugars, as well as the oxidative phosphorylation in the electron transport chain. Information was gathered through an extensive survey of literature, including scientific articles and biochemistry textbooks, and of on-line databases. Integration of different types of information, namely genomic, biochemical and physiological, and of data referring to related microorgan- isms, was crucial to carry out the reconstruction, as publicly available systematised information for A. niger is scarce. Therefore, in the reconstruction process, whenever there was physiological evidence for the presence of a reaction or pathway in A. niger, but no genomic nor biochemical data were available to support it, genomic or biochemical data referring to A. nidulans, other species of aspergilli, or other filamentous fungi, such as Penicillium chrysogenum,were considered. Moreover, some data were extrapolated from the recently developed genome-scale metabolic reconstruc- tion of S. cerevisiae [16]. Figure 1 depicts the process of reconstruction of A. niger’s central carbon metabolism and Table 1 presents a list of the on-line databases consulted. Presence of metabolic reactions. The presence of each reaction comprised in the metabolic network of A. niger was assessed based on genomic, biochemical or physiological data, with decreasing degrees of reliability. The trustwor- thiness in the asserted metabolic reactions or pathways also 4244 H. David et al. (Eur. J. Biochem. 270) Ó FEBS 2003 decreases when reactions were included based on informa- tion referring to other fungal species. ThegenomeofA. niger, being almost three times larger than the baker’s yeast genome (35.9 Mb), has been com- pletely sequenced by the Dutch company, DSM (Heerlen, the Netherlands), and 14 400 genes have been identified. About 40% of the genes have been annotated and classified into functional categories. The categories ‘Meta- bolism’ and ‘Energy’ account for 3111 genes (21.6%) and 354 genes (2.5%), respectively, and about 209 genes are involved in the metabolism of carbohydrates [21a]. How- ever, A. niger’s genome sequence and annotation are not available to the public and hence specific genomic data for this fungus was collected from a survey in literature, yielding about 20 reactions assigned to genes. This gap of informa- tion was to some degree supplemented with genomic data for A. nidulans, which is more abundant and systematised (http://www.gla.ac.uk/Acad/IBLS/molgen/aspergillus/index. html). The genomic data was complemented with reports on the presence of specific enzyme activities. For instance, the enzy- matic step in the pentose metabolism of A. niger catalysed by the NADPH-dependent L -xylulose reductase (EC 1.1.1.10) was included in the metabolic reconstruction, although the corresponding gene has not been cloned, as there are reports on the activity of this enzyme in A. niger [22,23]. Based on the metabolic networks of other fungal species, additional reactions or pathways were included in the metabolic reconstruction whenever there was physiological evidence for the consumption of a given substrate or formation of a given metabolic product in A. niger.For example, chitin is known to be a major component of the cell wall of most filamentous fungi, and in particular of A. niger [24], but only some enzymatic steps in the biosynthetic pathway leading to this polymer have been characterised in this species, namely those catalysed by the enzymes glutamine-fructose-6-phosphate transami- nase (isomerising; EC 2.6.1.16) [24a] and chitin syn- thase (EC 2.4.1.16) [25]. The remaining steps included in the reconstruction were based on the metabolic path- way in S. cerevisiae [PATHWAY (KEGG), Table 1] that involves the enzymes glucosamine-phosphate N-acetyl- transferase (EC 2.3.1.4), phosphoacetylglucosamine mutase (EC 5.4.2.3) and UDP-N-acetylglucosamine pyrophospho- rylase (EC 2.7.7.23). Stoichiometry, cofactor requirement and reversibi- lity. Once the presence of a reaction was confirmed, its stoichiometry was ascertained. This task was straightfor- ward for those reactions that had EC numbers ascribed, as queries could here be made directly in enzyme databases, whereas the stoichiometry of reactions without EC numbers assigned was determined through meticulous investigation in the literature and reaction or pathway databases. Some enzymes may catalyse several reactions, an example being transketolase (EC 2.2.1.1) in the pentose phosphate pathway that catalyses the transfer of two carbon groups in the two conversions D -ribose-5-phos- phate + D -xylulose-5-phosphate « sedoheptulose-7-phos- phate + D -glyceraldehyde-3-phosphate and D -erythrose- 4-phosphate + D -xylulose-5-phosphate « beta- D -fructose- 6-phosphate + D -glyceraldehyde-3-phosphate. In such cases, all asserted reactions for the enzyme in question were included in the metabolic reconstruction. A related issue refers to enzymes that can use more than one cofactor. For instance, D -xylulose reductase (EC 1.1.1.9) accepts both NADH and NADPH as cofactors. However, Fig. 1. Schematic representation of the reconstruction process. The arrows indicate the order in which the survey was accomplished and point towards decreasing reliability on the asserted reactions. The information provided by metabolic pathways databases (ERGO, WIT, PATHWAY) and enzyme databases (BRENDA) can also rely on genomic data. During the reconstruction process, protein databases were also consulted (not represented in the diagram). In ‘other fungi’ are included other species of aspergilli, P. chrysogenum and S. cere- visiae. Table 1. On-line databases consulted during the reconstruction process. Database Type of database Microorganism ERGO (http://wit.integratedgenomics.com/IGwit/) Metabolic pathways Emericella nidulans a WIT (http://wit.mcs.anl.gov/WIT2/) Metabolic pathways Aspergillus nidulans PATHWAY (KEGG) (http://www.genome.ad.jp/kegg/metabolism.html) Metabolic pathways Saccharomyces cerevisiae and other microorganisms BRENDA (http://www.brenda.uni-koeln.de/) Enzymes Aspergillus niger and other microorganisms PDSBSTR (http://www.genome.ad.jp/dbget-bin/www_bfind?pdbstr-today) Proteins aspergilli PIR (http://pir.georgetown.edu/) Proteins aspergilli PRF (http://www.prf.or.jp/en/) Proteins aspergilli SWISS-PROT (http://www.expasy.org/sprot/) Proteins aspergilli A. nidulans linkage map (http://www.gla.ac.uk/Acad/IBLS/molgen/ aspergillus/index.html) Genes Aspergillus nidulans a Sexual phase of the fungal life cycle of Aspergillus nidulans. Ó FEBS 2003 Reconstruction of Aspergillus niger metabolism (Eur. J. Biochem. 270) 4245 the exact cofactor requirements are often unknown, and in such cases both reactions involving NADH and NADPH were considered. By default, all the reactions were considered to be reversible, unless specific information indicating unidirec- tionality was available [e.g. hexokinase (EC 2.7.1.1)]. In the stoichiometric model, some reactions were subsequently assumed to be irreversible in the forward or the backward direction, in order to avoid artificial transhydrogenation cycles converting NADH into NADPH without net con- version of metabolites (refer to section Removing artificial transhydrogenation cycles). Compartmentation and localisation of reactions. In the metabolic reconstruction, intracellular compartmentation is considered and consequently reactions and metabolites are distributed among the extracellular medium and three intracellular compartments, namely cytosol, mitochondria and glyoxysomes. Thus, besides biochemical conversions, the metabolic network also includes transport processes between the different compartments and between the cell and the environment. By default, all the reactions were considered to occur in the cytosol, unless specific informa- tion on their localisation was available. In the reactions denoting transport processes across the cytoplasmic, mitochondrial and glyoxysomal membranes, neither protons nor ATP were considered, due to lack of specific information for aspergilli and for A. niger in particular. This assumption might have a profound effect when balancing protons or ATP for the calculation of fluxes using the stoichiometric model and this matter is further discussed in the section Energetic parameters. Reaction statistics Table 2 presents information on the number of biochemical transformations and metabolites that occur extracellularly and intracellularly, in the different compartments, as well as information on the number of transport processes, which are defined across the cytoplasmic, mitochondrial and glyoxysomal membranes, both for the reconstructed net- work and for the metabolic model subsequently developed (section Stoichiometric model for A. niger). The biochemical conversions comprising the metabolic reconstruction (both with and without EC numbers assigned) were classified into the six main classes of enzymes, according to the type of transformation implica- ted. The involvement of each class of enzymes in the carbohydrate metabolism, as well as energy metabolism proposed for A. niger, was assessed and compared to those in the metabolic reconstructions of S. cerevisiae [16] and of E. coli [14]. As shown in Fig. 2, the oxidoreduction reactions that are catalysed by oxidoreductases (class 1), represent the pre- dominant group of biochemical transformations (39%), being followed by reactions catalysed by transferases (class 2), which account for 26% of the total number of reactions in the part of A. niger’s metabolism under investigation. Hydrolases (class 3) and lyases (class 4) have lower contributions, corresponding to 15 and 11% of the total number of reactions considered, respectively. Iso- merases (class 5) and ligases (class 6) are involved to an even lesser extent, comprising 7% and 2% of the reactions under study, respectively. The relative contributions of the different classes of enzymes in the reconstructed carbohydrate and energy metabolisms in A. niger seem to follow the same trend of those in S. cerevisiae and a reasonable quantitative agree- ment is also observed. The same scenario does not apply for E. coli, where isomerases occupy the third position in abundance, in the reconstructed carbohydrate and energy metabolisms, and are followed by lyases, hydrolases and ligases (Fig. 2). Furthermore, the substrate specificity of the different groups of enzymes included in the metabolic reconstruction of A. niger was evaluated based on the ratio of the number of reactions to the number of enzymes in each category. Transferases appear to be the less substrate specific enzymes, followed by oxidoreductases, isomerases and lyases, whereas hydrolases and ligases seem to have high substrate speci- ficities, each of them catalysing only one reaction. Stoichiometric model for A. niger Following the phase of compilation of information con- cerning the structure of the central carbon metabolism of Table 2. Number of reactions and metabolites included in the metabolic reconstruction and in the stoichiometric model, and their localisation. Additionally to the reactions comprised in the metabolic reconstruction, the metabolic model also includes merged biochemical conversions. Processes Intracellular [number of reactions (%)] Extracellular [number of reactions (%)] Total [number of reactions (%)] Cytosol Mitochondria Glyoxysomes Metabolites 181 (63.7) 43 (15.1) 11 (3.9) 49 (17.3) 284 (100) Reactions Biochemical conversions Detailed 174 (80.9) 26 (12.1) 3 (1.4) 12 (5.6) 215 Lumped 52 (98.1) 1 (1.9) 0 (0.0) 0 (0.0) 53 Total 268 (80) Transport processes 46 (68.7) a 14 (20.9) a 7 (10.4) a – 67 (20) Total 355 (100) a Cytoplasmic, mitochondrial and glyoxysomal membrane. 4246 H. David et al. (Eur. J. Biochem. 270) Ó FEBS 2003 A. niger, a stoichiometric model was developed and subse- quently used to simulate growth and metabolite production as described in the section Model predictions. A list of the reactions that comprise the stoichiometric model is available as supplementary material. Anabolic reactions. To describe growth, biomass produc- tion was regarded as a drain of macromolecules and building blocks required to produce cellular components. The demands on each of these compounds were estimated based on the biomass composition. No drain of free metabolites or dilution of the metabolite pool due to biomass growth was considered [17]. The cellular compo- sition considered for A. niger was based on the contents of the main biomass components of A. oryzae determined in [26] (Table 3). The pathways considered for amino acid synthesis were based on a metabolic reconstruction of aminoacidbiosynthesisofA. nidulans from sequenced expressed sequence tag data (http://wit.mcs.anl.gov/WIT2), whereas the reactions for the anabolism of lipids, nucleic acids and other macromolecules were taken from a simplified model developed by Pedersen et al.[26]forthe central carbon metabolism of A. oryzae. Within the scope of this study, a single overall equation denoting formation of biomass was included in the model, even though the cellular composition varies with the specific growth rate [26]. The sensitivity of the biomass yield to perturbations in the biosynthetic demands has been assessed in different studies and some authors concluded that the biomass yield was not overly sensitive to changes in biosynthetic requirements [27], whereas others emphasised the importance of incorporating changes in biomass com- position with growth rate in flux estimation [28]. Removing artificial transhydrogenation cycles. As men- tioned previously, due to the lack of information, many reactions were, by default, represented as being reversible and/or accepting both NADH and NADPH as cofactors. When simulating the model, this would result in artificial transhydrogenation cycles converting NADH into NADPH without net formation of other metabolites. Such cycles may arise between pairs of reactions involving the same metabolites but different cofactors. As these cycles are not likely to be present under physiological conditions, one of the reactions involved was either constrained to be irrevers- ible or removed from the reaction set. As an example, we can refer to the potential cycle between the reactions catalysed by the enzymes D -xylulose reductase (NADH- and NADPH-dependent) and manni- tol-2-dehydrogenase (NADP + -dependent) (EC 1.1.1.138), interconverting D -xylulose and D -arabitol (Fig. 3). In this case, the transhydrogenation cycle was avoided by removing from the metabolic reconstruction the NADH-dependent reaction catalysed by D -xylulose reductase, which is equi- valent in assuming that the reduction of D -xylulose involves only the cofactor NADPH. A summary of the constraints considered to avoid artificial transhydrogenation cycles in the metabolic model proposed for A. niger is presented in Table 4. Energetic parameters. An advantage of using stoichiomet- ric models is that only a small number of parameters need to be determined. In addition to the biomass composition, the only parameters that had to be estimated were key energetic parameters: ATP requirement for nongrowth associated purposes (m ATP ), ATP yield on biomass (Y XATP )and operational P/O ratios. These parameters cannot be Fig. 2. Comparison of relative contributions of different enzyme classes in the reconstructed carbohydrate and energy metabolisms of A. niger (210 reactions), S . cerevisiae (143 reactions) and E. coli (119 reactions). Table 3. Cellular composition considered for determination of stoichiometric coefficients in biomass equation in the metabolic model of A. niger. Biomass component Molecular mass (g per mol of monomer in polymer) Content a (g per 100 g dry weight) Normalised b Stoichiometric coefficient c (mmol per g dry weight) Proteins 109.6 40.0 47.1 4.299 Carbohydrates 28.0 33.0 0.002 Glycogen 666.6 0.1 0.1 0.408 Chitin 203.2 7.0 8.3 1.515 Glucan 162.1 20.8 24.6 0.194 RNA 321.4 5.3 6.2 0.030 DNA 309.0 0.8 0.9 0.126 Lipids 634.9 6.8 8.0 0.213 D -Mannitol 182.2 3.3 3.9 0.090 Glycerol 92.1 0.7 0.8 Ash 15.1 – a For growth on glucose, using ammonia as the nitrogen source and for a specific growth rate of 0.1 h )1 [26]. b Without considering ash. c In the equation representing biomass formation (units: mmol of monomers in polymer per g dry weight). Ó FEBS 2003 Reconstruction of Aspergillus niger metabolism (Eur. J. Biochem. 270) 4247 determined independently, but if one of the parameters is known the others can be estimated from experimental data [17]. ATP and protons were in general not accounted for in the transport processes over the cellular membranes (refer to section Compartmentation and localisation of reactions). The only cases in which protons were explicitly considered were the reactions involved in the oxidative phosphorylation and electron transport chain, driving the proton motive force and generating ATP. In eukaryotes, many compounds are transported across the mitochondrial membrane by proton symport, resulting in an influx of protons into the mitochondrion that contributes to the incomplete coupling between the oxidation and phosphorylation processes in the oxidative phosphorylation, and consequently gives rise to lower P/O ratios than the theoretical values [17]. In order to account for this phenomenon in the model, the proton’s stoichiometric coefficient in the reaction catalysed by the enzyme H + -transporting ATP synthase (EC 3.6.1.34) was based on the operational P/O ratios observed in A. niger [29,30] (Table 5). The parameters m ATP and Y XATP (orrathertheATP requirements in the reaction denoting growth) were adjus- ted, so that the computed growth-yield matched experi- mentally observed biomass yields of A. niger, for different growth rates in glucose-limited continuous cultures [31]. The estimated values for the energetic parameters of A. niger are shown in Table 6, together with values found in the literature for the related filamentous fungus P. chrysogenum [29] and for S. cerevisiae [32]. The ATP requirement for nongrowth associated purposes calculated for A. niger is within the values presented for P. chrysogenum and S. cere- visiae, whereas the ATP yield on biomass is slightly lower than for P. chrysogenum and falls in the experimental range found in the literature for yeast. The former parameter was estimated to be 3.7 mmol ATP per g dry weight per h for A. niger, under citric acid production conditions [10]. Model predictions Once all relevant metabolic pathways of the central carbon metabolism of A. niger were identified and the model was further refined, the analysis of the system pursued with the Fig. 3. Representation of the artificial transhydrogenation cycle between the reactions catalysed by the enzymes D -xylulose reductase (NADH- dependent) and mannitol-2-dehydrogenase (NADP + -dependent), inter- converting D -xylulose and D -arabitol. Table 4. Potential artificial transhydrogenation cycles arising when simulating the metabolic model for A. niger and actions taken to avoid them. R, reversible reaction; I, irreversible reaction. Transhydrogenation cycle (NADH fi NADPH) Metabolites involved Added constraint NAD(H)-dependent D -Xylulose/xylitol NADP(H)-dependent reaction Xylitol dehydrogenase (R), considered to be irreversible in the L -Arabitol dehydrogenase (R), direction of reduction D -Xylulose reductase (EC 1.1.1.9) (R) NADP(H)-dependent D -Xylulose reductase (EC 1.1.1.9) (R) NAD(H)-dependent D -Xylulose/ D -arabitol NAD(H) not considered to act as D -Xylulose reductase (EC 1.1.1.9) (R) cofactor in the reaction catalysed by NADP(H)-dependent D -xylulose reductase D -Xylulose reductase (EC 1.1.1.9) (R), Mannitol 2-dehydrogenase (EC 1.1.1.138) (R) NAD(H)-dependent Glycerone/glycerol NADP(H) and NAD(H)-dependent Glycerol dehydrogenase (EC 1.1.1.6) (R) reactions considered to be NADP(H)-dependent irreversible in the direction of Glycerol dehydrogenase I and II (EC 1.1.1.72) (R) reduction and oxidation, respectively NAD(H)-dependent 2-Hydroxy-3-oxopropionate reductase (R) 2-Hydroxy-3-oxopropionate/glycerol Both NADP(H) and NAD(H)-dependent reactions NADP(H)-dependent considered to be irreversible in the 2-Hydroxy-3-oxopropionate reductase (R) direction of reduction NAD(H)-dependent Acetaldehyde/ethanol NADP(H)-dependent reactions Alcohol dehydrogenase I (EC 1.1.1.1) (R) considered to be irreversible in the NADP(H)-dependent D -Lactaldehyde dehydrogenase II (EC 1.1.1.78) (R), Glycerol dehydrogenase II (EC 1.1.1.72) (R) direction of reduction 4248 H. David et al. (Eur. J. Biochem. 270) Ó FEBS 2003 quantification of metabolic fluxes, using the framework of metabolite balancing in combination with linear program- ming algorithms. Flux distributions corresponding to opti- mal growth were calculated by maximising the flux of the reaction denoting biomass formation, while setting the substrate uptake rate to an appropriate value (Materials and methods). When simulating growth on glucose, it was observed that the model predicted zero flux through the pentose phos- phate pathway that is believed to be the major pathway for generation of NADPH. Using 13 C-labelling experiments, the pentose phosphate flux in a glucoamylase-producing recombinant strain of A. niger was estimated to be 58% and 72% of the glucose uptake rate, during batch (0.19 h )1 ) [33] and chemostat cultures (0.10 h )1 ) [34], respectively. For an a-amylase-producing strain and a wild-type strain of A. oryzae grown in chemostats at specific growth rates of approximately 0.10 h )1 , metabolite balancing was employed to calculate pentose phosphate pathway fluxes of 35% and 40%, respectively [26]. Carbon labelling analysis of glucose-limited continuous cultures of A. nidulans indicated that about 20 and 40% of the glucose is metabolised through this pathway, at low and high growth rates, respectively [7]. Besides the pentose phosphate pathway, other mechanisms have been proposed for the generation of NADPH in aspergilli, such as the mannitol cycle, the glycerol cycle and pyruvate/malate cycle, which involve transhydrogenation at the expense of ATP. However, these cycles seem to operate discontinuously and the studies accomplished provide no support for a significant contribution of these cycles in NADPH generation [7]. In the model simulations, NADPH is formed preferen- tially in the reaction catalysed by the cytosolic enzyme NADP + -dependent isocitrate dehydrogenase (EC 1.1.1.42). There is biochemical evidence for the presence of a NADP + -dependent isocitrate dehydrogenase in the cyto- plasm of A. niger [35,36], however, the activity of this enzyme seems to be very low, compared to the activity of the mitochondrial isoenzyme, when glucose is used as carbon source [37]. If the flux through this enzyme is constrained to zero in the model simulations, the pentose phosphate pathway becomes active (about 29% of the glucose is metabolised through this pathway for a growth rate of 0.09 h )1 ), and the computed biomass yield on glucose drops slightly [from 0.521 to 0.512 g (dry weight) per g glucose]. All flux distributions obtained using the model for simulation of growth involve secretion of fumarate in a rate that corresponds to 1–2% of the substrate uptake rate, and about to 3% of the specific growth rate on a carbon atom basis. However, there are no reports on the produc- tion of this organic acid by A. niger. Through investigation of the metabolic reconstruction for A. niger,itcanbe observed that fumarate is formed in the cytosol in reactions involved in the biosynthesis of amino acids and nucleotides, but there is no reaction for its consumption in this compartment. Unless a reaction in which cytosolic fumarate can be used as substrate or a transport process from the cytosol into the mitochondrion, where it can be consumed, are included in the metabolic network, secretion of fumarate to the extracellular medium is inevitable, as this compound is considered to be balanced in the stoichiometric model. The lack of evidence for a cytosolic fumarate dehydratase or a carrier for fumarate over the mitochondrial membrane, associated with a low predicted secretion rate of fumarate to the extracellular medium, seem to be reasonable reasons for accepting the simulated results. Similar effects on the computed flux distributions result from balancing of metabolites, such as CoA, NAD(P) + , FAD as well as one carbon compounds, for which there is no net formation and consumption. The proposed model predicts optimal metabolic beha- viour based on the stoichiometry of the reactions in the metabolic network and on the biomass composition con- sidered. However, there are other factors, such as kinetic or genetic regulation, that govern the metabolism and which are not accounted for in the model and therefore can explain the differences verified between simulated and experimental results, in some cases, such as the wrongly predicted flux through the pentose phosphate pathway discussed above. The example of fumarate secretion is indicative that the model needs to be further validated in order to predict reliable results and illustrates how the model can be used to guide experimental work, e.g. to identify the possible fate of fumarate produced in the cytosol. Biomass yields on different carbon sources. The maximum theoretical growth yield on different carbon sources was calculated for A. niger and compared to experimentally observed yields for A. oryzae for which a range of substrates have been investigated [4]. In order to account for the relative effect of maintenance for nongrowth associated purposes, all computations were performed considering the corresponding experimental substrate uptake rates. Table 5. P/O ratios considered for the computations and experimental range observed for A. niger. P/O ratios Considered value Experimental range for A. niger a NADH mit 2.46 2.3–2.7 NADH 1.64 1.4–1.8 Succinate 1.64 1.5–1.8 a Values taken from [29,30]. Table 6. Energetic parameters estimated for A. ni ger and comparison values found in the literature for other fungi. Energetic parameters Considered value Experimental value P. chrysogenum a S. cerevisiae b m ATP (mmol ATP per g dry weight per h) 1.9 2.9 < 1 Y XATP (mmol ATP per g dry weight) 71.4 75.0 71–91 a For an operational P/O ratio of 1.5 and Y sx ¼ 0.5 gÆg )1 [29]. b Values taken from [32]. Ó FEBS 2003 Reconstruction of Aspergillus niger metabolism (Eur. J. Biochem. 270) 4249 The predicted optimal growth yields are generally in good agreement with the experimentally observed values (Fig. 4). First of all, it can be seen that the biomass yield on glucose for A. oryzae is slightly lower than that of A. niger against which the model was calibrated. When fructose is used as the sole carbon source, the simulated flux distribution is as expected, very similar to that obtained for growth on glucose (results not shown) leading to an identical predicted biomass yield. The largest deviations between predicted and experimental result are observed for glycerol and acetate, which may be explained by a less optimised metabolic network for these uncommon substrates, i.e. futile cycles may operate in vivo, but they are not predicted by the model. Interestingly, the predicted growth yield on mannitol is higher than that the one for growth on glucose, a trend that is also observed experimentally. Theoretically, this is due to the fact that compared to glucose, each mole of mannitol converted to fructose-6-phosphate generates an additional mole of NADPH that can be used for biosynthesis. Essential reactions. In order to study the importance of the biochemical reactions in the metabolic reconstruction, each individual reaction was deleted from the metabolic network and optimal growth for the corresponding mutant was simulated for different carbon sources, namely glucose, xylose, glycerol and acetate. Table 7 shows that only a small Fig.4.Computedandexperimentalgrowthyields,duringgrowthon different carbon sources. Experimental data refers to A. oryzae [4]. Table 7. Essential and growth-retarding reactions for growth of A. niger on different carbon sources and pathways in which they take part. mit, Mitochondrial reaction; gly, glyoxysomal reaction (the remaining reactions are cytosolic). 4250 H. David et al. (Eur. J. Biochem. 270) Ó FEBS 2003 number of biochemical reactions are essential for growth on the carbon sources under study, reflecting the flexibility of the metabolic network to meet the biosynthetic require- ments, as well as the fact that many of the reactions are not involved during growth on these carbon sources. The removal from the metabolic network of reactions that are essential for growth on some carbon sources may have a retardantorhavenoeffectongrowthonothercarbon sources. The reactions that are essential for growth on the different carbon sources studied are mainly involved in the major catabolic pathways, namely tricarboxylic acid cycle (all carbon sources), pentose phosphate pathway (pentose), gluconeogenesis (glycerol and acetate) and glyoxylate shunt (acetate), as well as in the oxidative phosphorylation (all carbon sources). Furthermore, the model predicts that the elimination of certain reactions in the pathways of synthesis of biomass components (chitin, glucan, glycogen and mannitol) has a lethal effect on A. niger, for all the carbon sources investigated. Metabolite yields. A. niger is an important organism for metabolite production, in particular for organic acids. By maximising the excretion flux of a desired product instead of the biomass flux, the stoichiometric model can be used to assess the maximum theoretical yield for a given pair of substrate and product. The optimisation also results in one possible optimal flux distribution corresponding to the optimal yield, although it does not necessarily give information on how it could be achieved. An efficient process would typically require optimisation of both environmental conditions and microorganism, e.g. using genetic manipulations. Some of these considerations can, however, also be investigated using the described modelling framework as will be seen in the example below. Considering the production of succinate from glucose, the maximum theoretical yield for A. niger is 1.5 mol succinate per mol glucose (0.98 g per 1 g glucose) corres- ponding to 100% carbon yield. Unless the cells are forced to produce succinate, this outcome is not immediately physio- logically meaningful. Normally, succinate is observed as a by-product in fermentation, and although A. niger is a strictly aerobic organism one could imagine a production phase under microaerobic conditions. In the simulations, this could then be mimicked by constraining the specific oxygen uptake to be below a certain value. Figure 5 shows how the maximum and minimum succinate yields vary with the biomass yield on glucose, under fully aerobic conditions and ‘microaerobic’ condi- tions. The lighter shaded area of the figure represents the possible combinations of yields of biomass and succinate on glucose for fully aerobic conditions (unconstrained oxygen uptake rate), whereas the darker shaded area was obtained by constraining the specific oxygen uptake rate to be below 0.5 mmol O 2 per g dry weight per h (‘microaer- obic’ conditions). The highlighted points indicate the cases optimal growth and optimal succinate production, under fully aerobic conditions, as well as optimal growth, at ‘microaerobic’ conditions. These results suggest that a restricted oxygen supply does not necessarily imply pro- duction of succinate, and, at growth rates close to the optimal, the main fermentation by-product predicted is ethanol. Thus, to enforce production of succinate, one might have to consider inactivation (or addition) of specific metabolic reactions, for instance using genetic manipulations or starvation for important ‘cofactors’. The effects of such actions can also be investigated using the described frame- work simply by restricting the flux of the chosen reaction to zero or by adding a new reaction to the model. It is, for example, possible to search for optimal deletions that give high product formation at optimal growth. This can be elegantly formulated as a bi-level optimisation problem for any number of deletions [37a], but for the purpose of this study it is enough to consider direct search of optimal single and double deletions. Figure 6 shows the simulated results for the wild-type (darker shaded area), together with the theoretically optimal single (intermediate shaded area) and double (lighter shaded area) deletion mutants at ‘microaerobic’ conditions. The optimal single deletion found was the disruption of pyruvate decarboxylase (EC 4.1.1.1), preventing extensive channel- ling of pyruvate towards ethanol and acetate. For this mutant, several optimal flux distributions exist. At specific growth rates close to the optimal, the succinate yield on glucose is at least 0.47 mol succinate per mol glucose (0.31 g per 1 g glucose), and the other fermentation prod- ucts are either glycerol or L -arabitol and ethanol. When two disruptions are allowed, the highest succinate production is achieved by combining deletion of pyruvate decarboxylase with deletion of ATP:citrate oxaloacetate-lyase (EC 4.1.3.8), corresponding to a yield of at least 1.12 mol succinate per mol glucose (0.74 g per 1 g glucose), at optimal growth, being also produced as by-products glycerol and either ethanol or oxalate. These results suggest that the gene(s) encoding the mentioned enzyme(s) may be potential targets for metabolic Fig. 5. Computed succinate production limits of wild-type A. niger, under different conditions. Fully aerobic conditions (unconstrained qO 2 ) (lighter shaded area) and ‘microaerobic conditions’ (qO 2 con- strained to be below 0.5 mmol O 2 per g dry weight per h) (darker shaded area). The highlighted points indicate: (j) optimal growth yield and (r) optimal succinate yield on glucose, under fully aerobic conditions, and (m) optimal growth yield on glucose, under ‘micro- aerobic’ conditions. Ó FEBS 2003 Reconstruction of Aspergillus niger metabolism (Eur. J. Biochem. 270) 4251 engineering, however, mutant strains do not necessarily grow optimally [38]. Recent results suggest however, that it is possible to evolve microorganisms exhibiting suboptimal growth to the theoretically predicted properties [39]. Conclusions The reconstruction of the central carbon metabolism of A. niger presented here provides the first detailed descrip- tion of the central carbon metabolism of this microorgan- ism, namely in what concerns carbohydrates, organic acids, alcohols, and amino-sugars, and thereby covers the wide variety of carbon compounds that can be used by this fungus as a single carbon source for growth. As A. niger’s genomic sequence and annotation are not publicly available, the reconstruction process involved compilation and integration of different types of informa- tion concerning A. niger as well as data regarding other species of aspergilli and other fungi. Thus, the metabolic reconstruction presented here embodies a comprehensive database of reactions, resulting from a multitude of information sources, and accordingly may be used as a platform for reconstructing the metabolism of other related microorganisms. Although detailed to some extent, the metabolic recons- truction does not intend to provide a complete description of A. niger’s metabolism of carbohydrates, organic acids, alcohols, and amino-sugars; it represents instead an endeavour to provide systematic information in order to understand fungal metabolism. A thorough stoichiometric model was developed, based on the reconstructed metabolic network, and used to determine the metabolic capabilities of A. niger, under different genetic and environmental conditions, by employ- ing the framework of metabolite balancing in combination with linear programming methods. The model predicts optimal metabolic behaviour, and hence upper limits to the experimental data, and in some cases close agreement between experimental and simulated results can only be achieved by incorporating additional constraints related to the regulatory mechanisms governing the metabolism. On the other hand, the model requires further validation and here the availability of experimental data plays an important role. Once validated, the model can be used as a tool for the analysis, interpretation and prediction of metabolic beha- viour and hence guide the design of improved producing strains through metabolic engineering. Furthermore, the model can play a role in functional genomics, through identification of metabolites or reactions for which there is no interconnectivity in the metabolic network, and thereby suggesting missing metabolic reactions. Acknowledgements The authors thank Jochen Fo ¨ rster for extending his experience in reconstructing and analysing metabolic networks to this project, Marlene Leong for software development and George Ruijter for sharing his knowledge of A. niger’s metabolism. Financial support was provided in part by Fundac¸ a ˜ oparaaCieˆ ncia e a Tecnologia, Portugal, through a research fellowship for H. D. M. A ˚ . acknowledges Alf A ˚ kerman foundation, Sweden, and the Danish Biotechnology Instrument Center, Denmark. The research work on metabolite production by Aspergillus was financed by Vinnova, Sweden, and Erhvervsfremmestyrelsen, Denmark, via the Øresund Center Contract ETIF. References 1. Nielsen, J. (2001) Metabolic engineering. Appl. Microbiol. Bio- technol. 55, 263–283. 2. Kubicek, C. & Ro ¨ hr, M. (1986) Citric Acid Fermentation. Crit. Rev. Biotechnol. 3, 331–373. 3. Manzoni, M. & Rollini, M. (2002) Biosynthesis and biotechno- logical production of statins by filamentous fungi and application of these cholesterol-lowering drugs. Appl. Microbiol. Biotechnol. 58, 555–564. 4. Carlsen, M. & Nielsen, J. (2001) Influence of carbon source on alpha-amylase production by Aspergillus oryzae. Appl. Microbiol. Biotechnol. 57, 346–349. 5. Gwynne, D.I., Buxton, F.P., Williams, S.A., Garven, S. & Davies, R.W. (1987) Genetically engineered secretion of active human interferon and a bacterial endoglucanase from Aspergillus nidulans. Bio/Technology 5, 713–719. 6.Denning,D.W.,Anderson,M.J.,Turner,G.,Latge,J.P.& Bennett, J.W. (2002) Sequencing the Aspergillus fumigatus gen- ome. Lancet Infect. Dis. 2, 251–253. 7. Hondmann, D.H.A. & Visser, J. (1994) Carbon metabolism. In Aspergillus: 50 Years on (Martinelli, S.D. & Kinghorn, J.R., eds), pp. 61–139. Elsevier Science B, V., Amsterdam. 8. Ward, M. (1991) Aspergillus nidulans and other filamentous fungi as genetic systems. In Modern Microbial Genetics (Streips, V.N. & Yasbin, R.E., eds), pp. 455–496. Wiley-Liss, New York. 9. Peksel, A., Torres, N.V., Liu, J., Juneau, G. & Kubicek, C. (2002) 13 C-NMR analysis of glucose metabolism during citric acid production by Aspergillus niger. Appl. Microbiol. Biotechnol. 58, 157–163. 10. Guebel, D.V. & Darias, N.V.T. (2001) Optimization of the citric acid production by Aspergillus niger through a metabolic flux balance model. Electronic J. Biotechnol. 4, 1–17. 11. Verhoff, F.H. & Spradlin, J.E. (1976) Mass and energy balance analysis of metabolic pathways applied to citric acid production by Aspergillus niger. Biotechnol. Bioeng. 18, 425–432. Fig. 6. Computed succinate production limits of A. niger, under ‘microaerobic’ conditions (qO 2 constrained to be below 0.5 mmol O 2 per g dry weight per h). Wild-type (darker shaded area); single deletion mutant DEC 4.1.1.1 (intermediate shaded area); and double deletion mutant DEC 4.1.1.1 + DEC 4.1.3.8 (lighter shaded area). 4252 H. David et al. (Eur. J. Biochem. 270) Ó FEBS 2003 [...]... FEBS 2003 Reconstruction of Aspergillus niger metabolism (Eur J Biochem 270) 4253 12 Alvarez-Vasquez, F., Gonzalez-Alcon, C & Torres, N.V (2000) Metabolism of citric acid production by Aspergillus niger: model definition, steady-state analysis and constrained optimization of citric acid production rate Biotechnol Bioeng 70, 82–108 13 Edwards, J.S & Palsson, B.O (1999) Systems properties of the Haemophilus... Classification of fungal chitin synthases Proc Natl Acad Sci USA 89, 519–523 26 Pedersen, H., Carlsen, M & Nielsen, J (1999) Identification of enzymes and quantification of metabolic fluxes in the wild type and in a recombinant Aspergillus oryzae strain Appl Environ Microbiol 65, 11–19 27 Varma, A & Palsson, B.O (1993) Metabolic capabilities of Escherichia coli II Optimal growth patterns J Theor Biol 165,... J 236, 549–557 36 Muller, H.M & Frosch, S (1975) Oxalate accumulation from citrate by Aspergillus niger II Involvement of the tricarboxylic acid cyclase Arch Microbiol 104, 159–162 37 Jaklitsch, W.M., Kubicek, C.P & Scrutton, M.C (1991) Intracellular location of enzymes involved in citrate production by Aspergillus niger Can J Microbiol 37, 823–827 37a Maranas C.D (2002) In silico pathway analysis and... Vries, R.P & Visser, J (2001) The Aspergillus niger D-xylulose kinase gene is co-expressed with genes encoding arabinan degrading enzymes, and is essential for growth on D-xylose and 1-arabinose Eur J Biochem 268, 5414–5423 23 Witteveen, C.F.B., Busink, R., Van de Vondervoort, P., Dijkema, C., Swart, K & Visser, J (1989) L-Arabinose and D-xylose catabolism in Aspergillus niger J General Microbiol 135,... model of Escherichia coli metabolism: incorporation of growth-rate dependent biomass composition and mechanistic energy requirements Biotechnol Bioeng 56, 398–421 29 Nielsen, J (1997) Physiological engineering aspects of Penicillium chrysogenum World Scientific Publishing Co, Inc., Singapore 30 Watson, K & Smith, J.E (1967) Oxidative phosphorylation and respiratory control in mitochondria from Aspergillus. .. respiratory control in mitochondria from Aspergillus niger Biochem J 104, 332–339 31 Schrickx, J.M., Krave, A.S., Verdoes, J.C., van den Hondel, C.A., Stouthamer, A.H & van Verseveld, H.W (1993) Growth and product formation in chemostat and recycling cultures by Aspergillus niger N402 and a glucoamylase overproducing transformant, provided with multiple copies of the glaA gene J General Microbiol 139, 2801–2810... Verduyn, C., Postma, E., Scheffers, W.A & van Dijken, J.P (1990) Energetics of Saccharomyces cerevisiae in anaerobic glucose-limited chemostat cultures J Gen Microbiol 136, 405–412 33 Pedersen, H., Christensen, B., Hjort, C & Nielsen, J (2000) Construction and characterization of an oxalic acid nonproducing strain of Aspergillus niger Metab Eng 2, 34–41 34 Schmidt, K., Norregaard, L.C., Pedersen, B.,... Villadsen, J (1999) Quantification of intracellular metabolic fluxes from fractional enrichment and 13C)13C coupling constraints on the isotopomer distribution in labeled biomass components Metab Eng 1, 166–179 35 Meixner-Monori, B., Kubicek, C.P., Harrer, W., Schreferl, G & Rohr, M (1986) NADP-specific isocitrate dehydrogenase from the citric acid-accumulating fungus Aspergillus niger Biochem J 236, 549–557... Zakowska, Z., Gabara, B & Kusewicz, D (1997) Cell wall analysis in Aspergillus niger strains characterized by different tolerance to toxic compounds of beet molasses Acta Microbiol Pol 46, 27–36 24a Damveld R.A., (2002) Cell wall remodeling in A niger (6th European Conference on Fungal Genetics, 6–9 April 2002, Pisa, Italy Institute of Molecular Plant Sciences, Leiden University.) p 78 (Vannacci, G... ¨ Genome-scale reconstruction of the Saccharomyces cerevisiae metabolic network Genome Res 13, 244–253 17 Stephanopoulos, G.N., Aristidou, A.A & Nielsen, J (1998) Metabolic Engineering – Principles and Methodologies Academic Press, San Diego, USA 18 Nielsen, J & Villadsen, J (1994) Bioreaction Engineering Principles Plenum Press, New York 19 Aiba, S & Matsuoka, M (1979) Identification of metabolic model: . pathways of the central carbon metabolism of A. niger were identified and the model was further refined, the analysis of the system pursued with the Fig discussion Reconstruction process The metabolic reconstruction aims at depicting a detailed description of the central carbon metabolism of A. niger, namely of the