• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of plntphysLink to Publisher's site
Plant Physiol. Sep 2010; 154(1): 410–422.
Published online Jul 8, 2010. doi:  10.1104/pp.110.157198
PMCID: PMC2938163

The Metabolic Network of Synechocystis sp. PCC 6803: Systemic Properties of Autotrophic Growth1,[C][W]

Abstract

Unicellular cyanobacteria have attracted growing attention as potential host organisms for the production of valuable organic products and provide an ideal model to understand oxygenic photosynthesis and phototrophic metabolism. To obtain insight into the functional properties of phototrophic growth, we present a detailed reconstruction of the primary metabolic network of the autotrophic prokaryote Synechocystis sp. PCC 6803. The reconstruction is based on multiple data sources and extensive manual curation and significantly extends currently available repositories of cyanobacterial metabolism. A systematic functional analysis, utilizing the framework of flux-balance analysis, allows the prediction of essential metabolic pathways and reactions and allows the identification of inconsistencies in the current annotation. As a counterintuitive result, our computational model indicates that photorespiration is beneficial to achieve optimal growth rates. The reconstruction process highlights several obstacles currently encountered in the context of large-scale reconstructions of metabolic networks.

Cyanobacteria are among the evolutionarily oldest organisms and are the only known prokaryotes capable of plant-like oxygenic photosynthesis. As primary producers in aquatic environments, they play an important role in global CO2 assimilation and oxygen recycling. Recently, cyanobacteria have also attracted growing attention for economic purposes, including drug discovery and as prolific producers of natural products (Sielaff et al., 2006; Tan, 2007). In particular their ability to directly convert atmospheric CO2 into biomass and organic compounds, driven by sunlight, offers considerable potential as a novel and renewable resource for bioenergy (Deng and Coleman, 1999; Atsumi et al., 2009; Mascarelli, 2009; Lindberg et al., 2010).

Among the diverse cyanobacterial strains, Synechocystis sp. PCC 6803 is one of the most extensively studied model organisms for the analysis of photosynthetic processes. With a rich compendium of genomic, biochemical, and physiological data available, Synechocystis sp. PCC 6803, therefore, offers an ideal starting point to obtain insights into the systemic properties of phototrophic metabolism. The prerequisite for such a systemic description is a detailed reconstruction of the metabolic network of the organism: that is, a reconstruction of the comprehensive set of enzyme-catalyzed reactions required to support cellular growth and maintenance. Once a metabolic reconstruction is available, the vast array of methods developed by computational systems biology over the past decades allows us to dissect the functioning and interplay of possible metabolic routes and biochemical interconversions. In this respect, constraint-based modeling, most notably flux-balance analysis (FBA), has become a quasi-standard in the field. FBA is increasingly utilized to elucidate and characterize large-scale network properties, to direct the discovery of novel or alternative pathways, to guide metabolic engineering, as well as for the conceptualization of high-throughput data (Oberhardt et al., 2009; Steuer and Junker, 2009). As one of its prime advantages, constraint-based modeling does not require knowledge of the kinetic parameters of individual metabolic reactions, making it applicable to large-scale, up to genome-scale, metabolic networks. Specifically, FBA allows the prediction of optimal steady-state fluxes that maximize a given objective function, usually the synthesis of biomass or biomass precursors required for growth. Although certainly not without pitfalls, the predictions of FBA have proven to reasonably reflect the modes of cellular operation, with manifold applications ranging from microorganisms to algae and plant metabolism (Varma and Palsson, 1994; Shastri and Morgan 2005; Boyle and Morgan, 2009, Feist et al., 2009; Grafahrend-Belau et al., 2009; Oberhardt et al., 2009; Poolman et al., 2009).

The aim of our contribution is to provide a high-quality stoichiometric reconstruction of the primary metabolic network of Synechocystis sp. PCC 6803 and a characterization of its functional properties during phototrophic growth. We identify inconsistencies in the current gene annotation and suggest several genes with putative novel or different enzymatic activity for further experimental validation. Our work significantly improves upon earlier reconstructions, which were either restricted to the primary carbon metabolism (Shastri and Morgan, 2005; Hong and Lee, 2007; Navarro et al., 2009) or incomplete with respect to several metabolic routes (Fu, 2008; Kun et al., 2008). During the reconstruction process, we identified various obstacles that currently hamper the large-scale construction of metabolic models. First, despite the rather detailed annotation of the genome sequence of Synechocystis sp. PCC 6803, reflecting the high level of curation for this model organism, our analysis reveals a number of glaring gaps even within primary metabolic pathways. For several essential metabolic precursors, no feasible and validated synthesis routes are known. Second, automated reconstruction and comparison of different resources is significantly compromised by heterogeneous naming conventions used in different databases, making a cross-database comparison a laborious and predominantly manual process. To facilitate model exchange and reuse, as well as to provide a resource for further studies, we therefore place specific emphasize on a consistent and standardized nomenclature. Our reconstruction complies with the minimal information requested for the annotation of biochemical models (MIRIAM) and is represented using the systems biology markup language (SBML), allowing researchers to uniquely identify network components and enabling automated reasoning and model comparison (Hucka et al., 2003; Le Novère et al., 2005).

The computational model is subsequently utilized to explore and characterize possible flux states of Synechocystis sp. PCC 6803. Making use of the framework of FBA, we identify essential reaction sets and optimal synthesis routes for growth and maintenance as well as the necessary reorganization of metabolic fluxes in the transition from storage-utilizing night metabolism to phototrophic growth. Remarkably, our results indicate that the seemingly wasteful oxygenation of ribulose-1,5-bisphosphate (RuBP) is required to achieve an optimal flux state. This unintuitive result is discussed in the light of ongoing work to elucidate and understand photorespiratory metabolism (Eisenhut et al., 2006, 2008).

RESULTS

A Metabolic Model of Synechocystis sp. PCC 6803

We aim at a comprehensive reconstruction of the primary metabolic pathways of Synechocystis sp. PCC 6803. The scope of our reconstruction encompasses the main metabolic pathways, including photosynthetic light reactions, the Calvin cycle, glycolysis, the (incomplete) TCA cycle, the pentose-phosphate pathway, oxidative phosphorylation, photorespiration, amino acid synthesis, as well as supply of nucleotides, lipid precursors, and cofactors. Our starting point is the annotated genome sequence of Synechocystis sp. PCC 6803 as obtained from the CyanoBase Web site (Nakamura et al., 1998; Nakao et al., 2010).

Following earlier strategies (Feist et al., 2009; Oberhardt et al., 2009), the reconstruction process comprises four steps. First, the annotated genome, together with pathway repositories such as the Kyoto Encyclopedia of Genes and Genomes (KEGG; http://www.kegg.com/), provide an initial draft network of the organism. Second, gaps and inconsistencies within the draft network are identified, and the network is manually curated based on the available biochemical literature. Third, the set of reactions is converted into a mathematical model, using consistent annotations as well as established exchange formats, such as SBML. The conversion into a computational model entails the addition of several pseudoreactions, such as ATP utilization for cellular maintenance or biomass formation, which do not correspond to genuine enzymatic reactions but reflect overall cellular processes. Finally, the model is evaluated using constraint-based analysis and other methods of computational systems biology. In particular, predictions about optimal flux states are validated using available phenotypic data. Details of the reconstruction process are given in “Materials and Methods.”

Our reconstruction results in a metabolic network of 380 reactions and 291 metabolic compounds. The set of reactions consists of 360 enzymatic reactions, including about 38 isoreactions, four spontaneous uncatalyzed reactions, five diffusion and transport reactions, as well as 11 auxiliary reactions to account for light absorption, photophosphorylation, biomass formation, and cellular maintenance. All processes are assumed to take place in a single compartment, with the exception of transport reactions. The reconstruction is focused on a functional description of phototrophic growth and utilization of storage compounds in the absence of light. Uptake and utilization of organic molecules is not considered. All enzymatic reactions annotated for Synechocystis sp. PCC 6803 in KEGG and other repositories but not included in the core reconstruction are provided as a separate file. An overview of the model is shown in Figure 1.

Figure 1.
An overview of metabolic pathways and reactions included in the metabolic reconstruction of Synechocystis sp. PCC 6803. The model consists of 380 reactions, including the primary metabolic pathways, transport reactions, light absorption, photophosphorylation, ...

Objectives for Growth and Maintenance

During the reconstruction process, the initial draft model obtained from the annotated genome sequence was iteratively refined using extensive manual curation. Our primary objective was to create a metabolically functional model capable of supporting cellular maintenance and growth: that is, a model that is capable of generating all relevant biomass components. In general, the biomass objective function (BOF), an equation that specifies all relevant biomass precursors in their appropriate molar ratios, is difficult to obtain experimentally. An estimation of the macromolecular biomass composition for Synechocystis sp. PCC 6803 under different growth conditions was given by Shastri and Morgan (2005) and afterward utilized in several studies. However, as the size of our reconstruction is considerably larger, the inclusion of additional components was necessary. To this end, we modified and extended the BOF given by Shastri and Morgan (2005) based on information on the macromolecular composition reported for the cyanobacterium Arthrospira (Spirulina) platensis (Cogne et al., 2003). The BOF is given in Table I and detailed in “Materials and Methods.”

Table I.
The BOF, adopted and modified from Cogne et al. (2003)

Knowledge of the BOF allows us to systematically check for available synthesis routes of biomass components. We performed iterative gap filling based on multiple pathway repositories, such as KEGG (http://www.kegg.com/), BioCyc (http://www.biocyc.com/), and BRENDA (http://www.brenda-enzymes.org/), as well as curation of primary biochemical literature and identification of putative enzymes via BLAST search. As expected, the initial draft network did not support the synthesis of all required intermediates. In particular, no complete synthesis routes for the amino acids Gly, Ser, Cys, Met, Asn, and His were annotated. In these cases, we proceeded according to the following scheme. First, we aimed to identify the missing steps to complete the synthesis pathways utilizing information from other cyanobacteria. For example, the enzyme for a methylenetetrahydrofolate reductase (EC 1.5.1.20) was not annotated for Synechocystis but is annotated for Cyanothece sp. PCC 8801, which enables the identification of a candidate gene (slr2141) for this enzyme in Synechocystis sp. PCC 6803. Second, if no satisfactory results could be obtained by direct comparison, we searched the primary literature for a description of the respective pathways, starting with cyanobacterial metabolism and then broadening the search to plants (mainly Arabidopsis [Arabidopsis thaliana]) and finally across all kingdoms. Our strategy resulted in complete synthesis routes for all amino acids except Asn and Met. For the latter, the synthesis steps from Microcystis aeruginosa (EC 2.3.1.31, EC 2.5.1.49) were adopted to allow for a metabolically functional network.

Similarly, the scheme described above was applied to test and complete the pathways for de novo synthesis of cofactors, including CoA, NADs (NAD/NADH), NADPs (NADP/NADPH), FADs (FAD/FADH), and tetrahydrofolate. Again, the initial draft network had to be supplemented with additional reactions. A complete list is provided in “Materials and Methods” and as Supplemental File S3. For three reactions, UDP-Glc phosphorylase (EC 2.7.7.9), aminodeoxychorismate synthase (EC 2.6.1.85), and hydrolase of 5-amino-6-(5′-phosphoribitylamino)uracil (EC 3.1.3.-), no reasonable candidate genes could be identified, although each step is necessary to allow for the synthesis of biomass precursors. Finally, it was verified that the model is able to replenish each intracellular metabolite upon dilution by growth.

Main resources during the reconstruction process were the CyanoBase Web site (Nakamura et al., 1998), KEGG (Kanehisa and Goto, 2000), MetaCyc and BioCyc (Caspi et al., 2010), BRENDA (Chang et al., 2009), the CyanoMutants Web site (Nakamura et al., 1999), and to a lesser extent UniProt (UniProt Consortium, 2010). The final network is provided as a MIRIAM-compliant SBML file (Supplemental File S1) and spreadsheet (Supplemental File S2).

Functional Properties of Phototrophic Growth

Based on the reconstructed network, the functional properties of phototrophic growth are evaluated using FBA. Specifically, the rate of change of the vector of metabolite concentrations S can be described by a differential mass-balance equation of the form

An external file that holds a picture, illustration, etc.
Object name is pp157198E01_LW.jpg

where N denotes the stoichiometric matrix and v indicates the vector of reaction fluxes. Using the steady-state hypothesis for the vector of metabolite concentrations, any stationary flux distribution v0 has to satisfy the steady-state constraint N v0 = 0. In addition to the constraint imposed by the mass-balance equation, FBA postulates that the distribution of fluxes is organized such that a given objective function, usually the synthesis of precursors required for cellular growth, is maximized. Mathematically, this assumption translates into a linear optimization problem for the stationary flux vector v0, such that:

An external file that holds a picture, illustration, etc.
Object name is pp157198E02_LW.jpg

under the conditions:

An external file that holds a picture, illustration, etc.
Object name is pp157198E03_LW.jpg

and:

An external file that holds a picture, illustration, etc.
Object name is pp157198E04_LW.jpg

for all metabolic fluxes vi0 (i = 1,…,m), where vimin and vimax denote optional upper and lower bounds for each stationary flux value. The constraint-optimization problem can be solved using standard techniques of linear programming (Becker et al., 2007; Klamt et al., 2007). The solution to the optimization problem results in a particular flux distribution that achieves a maximal growth yield, given predefined external conditions and exchange fluxes. We note that, owing to redundancy and flexibility within the network, the optimal solution is usually not unique.

Once the constraint-optimization problem is specified, FBA allows for a functional interrogation of the metabolic network with respect to (1) its ability to synthesize all metabolic precursors at their observed molar ratios, (2) essential reactions and prediction of gene deletion studies, (3) optimal routes for maximal growth, (4) flux variability analysis, as well as (5) qualitative reorganization of metabolic routes under different growth conditions.

In the following, we are mainly interested in phototrophic growth under conditions of either limiting light or limiting CO2 availability. Within the network, inorganic carbon is taken up from the environment as bicarbonate HCO3 by an active (ATP-binding cassette) transport (Omata et al., 1999). The intensity of light input is represented by flux of absorbed photons and translocation of H+ through the membrane, partially coupled to regeneration of NADPH. Photophosphorylation and the equations for the respiratory chain are adopted from Shastri and Morgan (2005). Figure 2 shows the predicted maximal growth rate as a function of input fluxes. Maximal growth is linearly dependent on either absorbed light or net CO2 availability, with an offset for small light intensities. In addition to the biomass function, we assume a basal ATP demand of 0.27 mmol g−1 dry weight h−1 for cellular maintenance; see “Materials and Methods.”

Figure 2.
The predicted maximal growth rate as a function of light intensity and total CO2 uptake. [See online article for color version of this figure.]

In the following, all simulations are performed at a specified light intensity, whereas the model is allowed to adjust HCO3 uptake as necessary for optimal growth yield. In this way, we circumvent the necessity to utilize a two-step optimization strategy, as compared with Shastri and Morgan (2005). Likewise, we do not make use of auxiliary optimization criteria, such as minimal total flux, to forestall variability within optimal flux distributions. Instead, the variability in the optimal solutions is described explicitly where appropriate. The constraint-optimization problem usually results in relative flux values. Typical solutions can be compared with observed growth rates. For example, Van Liere and Walsby (1982) report a typical carbon uptake of 0 to 13 μg carbon g−1 dry weight h−1, which translates into a flux of approximately up to 1.1 mmol CO2 g−1 dry weight h−1. If we assume an uptake of 0.22 mmol CO2 g−1 dry weight h−1, we obtain a growth rate of about 0.041 h−1, well within the range of values observed for Synechocystis under many conditions. The photosynthetic quotient, the ratio of moles of oxygen released per mole of CO2 fixed, is approximately 1.43. Overall, all quantitative results are in good agreement with values obtained from previous models (Shastri and Morgan, 2005; Fu, 2008), with exceptions mainly due to the different network stoichiometry.

A predicted optimal flux distribution of central metabolism consistent with maximal growth is shown in Figure 3. Flux values are reported relative to Rubisco activity (100%). The solution is not unique, but considerable variability is observed in the utilization of the transaldolase (talB, slr1793; EC 2.2.1.2), the Fru-1,6-bisP aldolases (cbbA and fda, sll0018 and slr0943; EC 4.1.2.13), and the Fru-1,6-bisPase II/sedoheptulose-1,7-bisphosphatase (fbpI and glpX, slr2094 and slr0952; EC 3.1.3.11 and EC 3.1.3.37), corresponding to alternative optimal routes within the network (Poolman et al., 2003).

Figure 3.
A predicted flux distribution to achieve maximal phototrophic growth yield (BOF). All flux values are given relative to the synthesis of RuBP (100%). The simulation was performed for a given light intensity, while carbon uptake was allowed to adjust as ...

Common to all solutions is that the highest (absolute) flux is allocated to the glyceraldehyde-3-phosphate dehydrogenase (gap2, sll1342; EC 1.2.1.12). The reaction is a key reaction of the Calvin-Benson cycle and essential for growth. The corresponding mutant is not viable under phototrophic conditions (Koksharova et al., 1998). Likewise common to all solutions is a dominance of ribulose-phosphate 3-epimerase (cfxE, sll0807; EC 5.1.3.1) over the Rib-5-P isomerase A (rpiA, slr0194; EC 5.3.1.6), with more than twice the flux allocated for optimal growth. Both reactions are essential within the network. At the level of glycerate-3-phosphate, the flux is split toward biosynthesis and regeneration of the Calvin-Benson cycle, with a ratio of approximately 1:8. The optimal ratio is significantly higher than the textbook ratio of 1:5, since precursors for biosynthesis are withdrawn during regeneration, such as d-Glc-6-P for synthesis of glycogen and nucleotide sugars and Rib-5-P for synthesis of folate, purines, and His.

Our model includes photorespiration, which competitively inhibits the photosynthetic carboxylation of RuBP by Rubisco. Interestingly, the predicted optimal flux pattern allocates approximately 4% to 5% of the flux through Rubisco to the allegedly disadvantageous oxygenation reaction, thereby producing glyoxylate as an essential precursor for Gly, Ser, and Cys formation. For the enzymes associated with the TCA cycle, the optimal flux distribution reflects the utilization of an incomplete cycle present in Synechocystis sp. PCC 6803. In particular, fumarate is mainly synthesized as a by-product of Arg and nucleotide biosynthesis and channeled back to oxaloacetate for reuse in the TCA cycle and to serve as a precursor for the synthesis of diverse amino acids. The distribution shown in Figure 3 is characteristic for photoautotrophic growth. We observed no qualitative changes with respect to different light intensities or different CO2 availability.

Gene Deletion Studies and Metabolic Robustness

The validity of any computational model can be examined using phenotypic data and information on metabolic knockouts. In particular, one of the prime accomplishments of FBA is its ability to predict the viability or the expected (relative) decrease in maximal growth rate upon deletion of one or more genes. In this respect, we can assess the essentiality of each gene based on two criteria. The first requirement is the persistent ability of the network to produce all precursors required for biomass formation, as specified by the BOF, even in the absence of a specific gene. Second is the requirement to regenerate all metabolites and cofactors at a specified rate in the absence of the respective gene. Although usually only the first requirement is considered in the literature, both criteria are essential for viability and persistent growth. From an algorithmic point of view, the essentiality of each gene is evaluated by removing its corresponding reactions from the network and reevaluating the constraint-optimization problem for maximal biomass production. Genes are classified as essential if removal of the respective reactions prohibits the formation of biomass. Nonessential genes are further classified according to whether the maximal growth rate is reduced or remains unaltered. Within our network, 126 (of 337) genes are classified as essential for biomass formation, the removal of 26 genes results in a decreased maximal growth yield, whereas the removal of the remaining 185 genes has no effect on the maximal growth yield. The latter set includes a set of 38 isoenzymes. Replacing biomass formation by the ability to regenerate all metabolites results in 177 essential genes, whereas 160 genes are classified as nonessential, including 31 genes that result in a less efficient regeneration of metabolites and cofactors. A comprehensive list of essential genes is provided as Supplemental File S5.

In total, 51 entries in CyanoMutants corresponding to mutational inactivation of distinct genes are directly related to our network reconstruction. Of these, 39 genes encode nonessential genes within the network, of which 35 can be confirmed by a successful knockout reported in CyanoMutants. The discrepant four genes correspond to Cys desulfurase (slr0077), the NADH dehydrogenase type I subunit 4L (sll0522), a polyphosphate kinase (sll0290), and a pyrophosphatase (slr1622). The latter two reactions are easily compensated in silico by pairs of reactions acting together in a cycle, a solution that might not be feasible in vivo under the conditions tested. Essential genes are more difficult to reconcile, as our network does not account for possible roles in regulatory processes or toxicity and other adverse effects upon possible accumulation of metabolites. Of the 51 entries from CyanoMutants, 12 genes are classified as essential within our model, but only for eight of those the construction of a fully segregated mutant failed. The inactivation of the remaining four genes each resulted in a viable genotype. A more detailed comparison, including a summary of discrepant findings, is given in “Materials and Methods.”

From Storage Utilization to Phototrophic Growth

In the absence of light, Synechocystis is able to temporarily utilize various storage compounds, such as glycogen and to a lesser extent cyanophycin and poly-β-hydroxybutyrate, to provide energy and precursors for cellular maintenance. In this respect, our in silico model allows one to simulate the necessary reorganization of metabolism during the transition from glycogen-utilizing night metabolism to phototrophic growth. Figure 4 shows the metabolic transitions on a two-dimensional plane with the flux of selected reactions color coded as a function of absorbed photons (x axis) and glycogen utilization (y axis). Within each subplot, the transition from night metabolism (glycogen utilization) to phototrophic growth corresponds to a path from the upper left to the lower right corner. Along such a path, we observe several characteristic transitions in the reorganization of metabolic flux.

Figure 4.
The transition from storage-utilizing night metabolism to phototrophic growth. Shown is the flux through selected reactions (color coded) as a function of absorbed photons (x axis) and glycogen utilization (y axis); see color bar for the respective ranges. ...

Under conditions of storage-utilization, corresponding to a point on the y axis in Figure 4, we assume glycogen as the only resource for cellular maintenance. In this case, the flux distribution is characterized by oxygen uptake and the absence of Rubisco-dependent carboxylation or oxygenation. Instead, we observe flux through the TCA cycle utilizing a bypass via the succinate-semialdehyde dehydrogenase (slr0370; EC 1.2.1.16) to circumvent the 2-oxoglutarate dehydrogenase not present in Synechocystis sp. PCC 6803. With increasing light intensity and concomitant decrease in glycogen mobilization, flux through the TCA cycle decreases. Instead, the photosynthetic light reactions start to generate the energy (ATP) requirements for cellular maintenance (fixed in the simulation at 0.27 mmol g−1 dry weight h−1), but no CO2 fixation takes place yet. We note that the observed in silico flux distribution during mixotrophic growth agrees with experimental results from Howitt and Vermaas (1998), who did not observe any effect on mixotrophic and phototrophic growth by the deletion of two oxidases.

Only for a light intensity above a certain threshold, flux through the Calvin-Benson cycle is initiated, resulting in an increasing rate of Rubisco-dependent carboxylation and oxygenation. Remarkably, oxygenation of RuBP occurs slightly before the start of CO2 fixation, even though within the in silico optimization a complete suppression of photorespiration is possible, if favorable for maximal ATP production or biomass formation. The occurrence of a nonzero photorespiration in the constraint-optimization problem points to a beneficial role of the seemingly wasteful side reaction as a supply of precursors for Gly formation. In particular, during heterotrophic growth, Gly is formed as a result of Pro degradation, which ceases in the presence of photorespiration. Consequently, the in silico model points to an ambivalent interpretation of the role of photorespiration during phototrophic growth, which is discussed in more detail below. We note that all results are contingent on the reconstructed network structure and the specified exchange (glycogen and light) fluxes imposed on the constraint-optimization problem and therefore may not necessarily replicate the in vivo reality. Nonetheless, the in silico simulation highlights the potential of cellular metabolism to adapt to changes in environmental conditions in complex, surprisingly subtle, and sometimes nonintuitive ways.

Photorespiration Revisited

One of the striking results of the in silico simulation is the occurrence of a nonzero rate of photorespiration during photoautotrophic growth. In Synechocystis sp. PCC 6803, as in most other photosynthetic organisms, the fixation of inorganic carbon is catalyzed by the bifunctional enzyme Rubisco, with the carboxylation of RuBP yielding two molecules of PG3. However, carboxylation competes with RuBP oxygenation, a process that generates one molecule of PG3 and one molecule of the two-carbon compound 2-phosphoglycolate (2PG; see Fig. 5). Photorespiration, the photosynthetic oxygenation of RuBP, is often described as one of the most wasteful processes on Earth and is widely considered to be mainly an evolutionary relic, tracing back to the evolution of the enzyme in a low-oxygen atmosphere.

Figure 5.
The key photorespiration pathways in Synechocystis species PCC 6803, taking into account the results of Eisenhut et al. (2006, 2008). Oxygenation of RuBP generates one molecule of PG3 and one molecule of 2PG. The latter is converted into glyoxylate and ...

In cyanobacteria, the role and relevance of photorespiration is still only partially understood. Cyanobacteria have evolved dedicated CO2-concentration mechanisms that significantly repress the oxygenase activity of Rubisco and vastly improve photosynthetic performance (Badger and Price, 2003). However, the presence of CO2-concentration mechanisms does not fully eradicate RuBP oxygenation, and Eisenhut et al. (2008) were recently able to show that an active photorespiratory 2PG metabolism is essential for cyanobacterial growth in the present oxygen-containing atmosphere. In particular, a nonnegligible rate of photorespiratory 2PG generation necessitates quick conversion to glyoxylate to avoid toxification and inhibition of the Calvin-Benson cycle.

In this respect, our metabolic reconstruction allows us to augment the current discussion on photorespiration from a network perspective. As shown above, the constraint-optimization problem assigns a nonzero flux to the oxygenation of RuBP under conditions of photosynthetic growth, suggesting a nontrivial role of photorespiration to achieve a maximal rate of biomass formation. Indeed, an intermediate of photorespiration, glyoxylate, plays an important role in cyanobacterial metabolism and can serve as a precursor for Gly synthesis. More specifically, within our reconstruction, the biosynthesis of the amino acids Gly, Ser, and Cys exclusively depends on the allocation of glyoxylate. Although it was suggested that cyanobacteria are capable of synthesizing Ser directly from 3-phosphoglycerate via 3-phosphohydroxy pyruvate and phosphoserine (Colman and Norman, 1997), no candidate genes for such a conversion are currently known for Synechocystis. The assumption that glyoxylate is an essential precursor for Ser synthesis is further supported by the fact that Eisenhut et al. (2006) were unable to obtain a completely segregated mutant with an inactivated Ser hydroxymethyltransferase (sll1931). The Ser hydroxymethyltransferase catalyzes interconversion of Ser and Gly and would be dispensable if an alternative synthesis pathway for Ser exists. Correspondingly, our simulations suggest that during phototrophic growth, the photorespiratory 2PG metabolism provides the required flux toward glyoxylate to allow for the formation of metabolic precursors and biomass. The computationally allocated rate of photorespiration is approximately 4.2% of the total Rubisco activity, a value that is in good agreement with the 4% to 5% photorespiration assumed for Synechocystis (M. Hagemann, personal communication). Prohibiting photorespiratory activity in silico results in a slightly decreased predicted maximal growth rate of 94% of the original value.

Consequently, in the context of our network reconstruction, photorespiration turns out to be the most efficient way to allocate glyoxylate during phototrophic growth. However, this superior efficiency does not hold for nonphototrophic (dark) metabolism, as already observed in Figure 4. In this case, the (rather low) demand for glyoxylate is met by degradation of Pro, in a series of five enzymatic steps. Despite its slightly lower carbon efficiency, the synthesis of Pro and subsequent conversion to glyoxylate yields additional units of NADH and ATP, as compared with photorespiration, making the pathway preferable in heterotrophic conditions when optimizing for total biomass formation. We emphasize that this difference highlights the fact that a criterion like metabolic efficiency is usually not a property of individual pathways, but it must be interpreted in the context of the metabolic status and the demands of the remaining network.

Our computational results allow us to formulate two different scenarios. First, the beneficial role of photorespiration may be merely an artifact of an incomplete network topology. In particular, our results depend on the absence of alternative pathways for the formation of Gly and Ser, most importantly the direct conversion of 3-phosphoglycerate to Ser, via 3-phosphohydroxy pyruvate and phosphoserine. Although no candidate genes are currently known, the pathway may be active in Synechocystis sp. PCC 6803. In this case, photorespiration would be completely suppressed in both phototrophic and heterotrophic conditions when optimizing for maximal growth yield (data not shown). Alternatively, the second scenario implies that the current annotation of the Synechocystis genome, augmented by our network reconstruction, does provide a correct picture of Gly metabolism. In this case, the metabolism of Synechocystis is adapted to make use of the inevitable supply of the side product 2PG under photoautotrophic conditions. In particular, there is no requirement for the organism to establish or maintain alternative and possibly more efficient pathways for Gly and Ser formation. We note that this view does not contradict earlier findings on gene essentiality and growth in a high-CO2 atmosphere. Within our simulation, photorespiration is not essential, and in the absence of photorespiration, all essential precursors can still be synthesized, albeit with slightly lower efficiency. In the context of our study, it is not possible to conclusively resolve which of the two scenarios mirrors the in vivo reality. Nonetheless, we emphasize that only a thorough reconstruction of cyanobacterial metabolism, to the best of one’s current knowledge, allows us to formulate and probe alternative hypotheses on the efficiency and optimality of reactions.

DISCUSSION AND PERSPECTIVES

Elucidation of the structure and function of metabolic networks is increasingly assisted by computational approaches that allow for a systemic view of possible interconversion routes. A prerequisite for such an analysis is the comprehensive reconstruction of the set of enzymatic reactions taking place in a single cell or organism. Here, we have presented a manually curated reconstruction of the metabolic network of Synechocystis sp. PCC 6803, an important unicellular model organism for phototrophic metabolism.

Network reconstruction is usually based on the annotated genome sequence, in combination with various reaction and pathway repositories. However, most databases and repositories primarily focus on the “parts lists” of enzymatic reactions: that is, they do not consider the resulting network as a functional entity. Conversely, network reconstruction specifically aims to take into account the multiple requirements that distinguish a list of reactions from a functioning network. Such requirements include the ability of the network to synthesize all essential intermediates and precursors for biomass production, to allow for energy (ATP) generation for cellular maintenance, to maintain the redox balance, as well as to be capable of regenerating metabolites and cofactors upon dilution by growth. Failure to fulfill one or more of these requirements usually points to errors or gaps in the network stoichiometry and necessitates revision of the reconstructed network accordingly. In this sense, network reconstruction is a prestage to metabolic modeling and recognizes that cellular metabolism operates as a highly integrated network.

Despite the great achievements over the last decade, resulting in more than a dozen reconstructions for several organisms, the systematic reconstruction of genome-scale metabolic networks is still in its infancy. While already considerable literature exists on algorithms for automated gap filling, these algorithms are usually exemplified only on highly domesticated examples, such as recovery of a previously known complete network upon artificial deletion of individual reactions. Therefore, as yet, high-quality reconstructions are predominantly based on extensive manual curation utilizing information from primary biochemical literature. A persistent problem is the diverse naming conventions used in different databases and existing reconstructions as well as the sometimes inconsistent or erroneous annotation of enzymes, often based only on sequence comparison. As pointed out previously (Stitt et al., 2010), a great challenge for the future is the development and enforcement of publishing standards and naming conventions. Other problems are rooted in interspecific and intraspecific variations and are less straightforward to overcome. For example, only recently it was shown than even within such a basic and exhaustively analyzed pathway as central carbon catabolism in Escherichia coli, previously unreported reactions occur (Nakahigashi et al., 2009). In this respect, we do not expect our reconstruction to be a faultless replicate of the in vivo reality; rather, the purpose is to start and foster a discussion about possible inconsistencies and shortcomings of our current understanding of cyanobacterial metabolism. In particular, a high-quality network reconstruction, even if erroneous or incomplete in parts, offers tremendous possibilities to probe and understand the functioning of metabolism.

Most current reconstructions are evaluated using FBA, a constraint-based method that allows one to predict and evaluate optimal flux patterns with respect to a given objective function. In our case, the application of FBA led to several conclusions about (1) the typical optimal reaction routes during phototrophic growth, (2) essential genes and synthesis pathways of photosynthetic metabolism, as well as (3) the necessary reorganization of metabolic flux during a transition from heterotrophic (glycogen-based) to phototrophic growth. The latter highlights the interconnection of cellular pathways as a system of supply and demand. We observe several characteristic switches that emphasize that the efficiency of cellular reaction pathways is not an intrinsic property of the respective pathways but depends on the current status and metabolic demand of the entire metabolism. The most prominent example for such a complex interdependency is the allocation of nonzero flux to the oxygenation reaction of RuBP. While for low light intensities the demand for Gly is met by alternative pathways, such as utilization of Pro, the Rubisco oxygenase reaction becomes more efficient with increasing light intensity and finally provides all glyoxylate required for cellular growth. This transition is not an effect of the efficiency of the pathway per se but rather reflects differences in the “sales value” of cellular commodities under different growth conditions. We note that the fact that the seemingly wasteful side reaction of Rubisco might be used in a constructive way is in good agreement with the known tradeoff between specificity and overall turnover rate of Rubisco (Tcherkez et al., 2006). The hypothesis resulting from our analysis might also explain the remarkably low specificity and high turnover rates observed in some cyanobacteria (Witzel et al., 2010).

Similar consideration with respect to the beneficial role of seemingly wasteful side reactions might also hold for other organisms and tissues. For example, in a recent computational study, Poolman et al. (2009) reported a nonzero Rubisco oxygenase reaction as a result of a constraint-optimization problem in heterotrophic Arabidopsis cells. While the scenario discussed therein is probably unlikely to occur in vivo, as also acknowledged by the authors (Poolman et al., 2009; Stitt et al., 2010), the prediction of flux through previously unrecognized reactions as a result of network optimization is not unusual (Schwender et al., 2004). In this sense, a thorough network reconstruction allows researchers to evaluate the inevitable functional consequences, given the current annotation and knowledge of enzymatic interconversions, and thus opens the possibility of specifically designing experiments to distinguish between alternative hypotheses.

MATERIALS AND METHODS

Abbreviations and Nomenclature

The following terms are used in the figures and supporting materials: 2PG (2-phosphoglycolate), ABC (ATP-binding cassette), AcCoA (acetyl-coenzyme A), BOF (biomass objective function), Cytb6f (cytochrome b6f complex), DHAP (dihydroxyacetone phosphate), DW (dry weight), E4P (d-erythrose 4-phosphate), F6P (d-fructose 6-phosphate), FBA (flux-balance analysis), FBP (d-fructose 1,6-bisphosphate), FNR (ferredoxin-NADP+ reductase), G1P (d-glucose 1-phosphate), PG2 (d-glycerate 2-phosphate), PG3 (d-glycerate 3-phosphate), G6P (d-glucose 6-phosphate), GAP (d-glyceraldehyde 3-phosphate), MTHF (5,10-methylenetetrahydrofolate), PEP (phosphoenolpyruvate), PHB (poly-3-hydroxybutyrate), PYR (pyruvate), R5P (d-ribose 5-phosphate), Ru5P (d-ribulose-5-phosphate), RuBP (d-ribulose 1,5-bisphosphate), S7P (d-sedoheptulose 7-phosphate), SBP (d-sedoheptulose 1,7-bisphosphate), THF (tetrahydrofolate), X5P (d-xylulose 5-phosphate).

Metabolic Network Reconstruction

Network reconstruction was performed as outlined in the main text. Starting point of the reconstruction is the cyanobacteria genome database CyanoBase (http://genome.kazusa.or.jp/cyanobase) containing the annotated nucleotide sequence of the genome of Synechocystis sp. PCC 6803 (Nakamura et al., 1998; Nakao et al., 2010). From the gene identifiers, an initial list of enzymes and their associated reactions was assembled using the KEGG database (Kanehisa and Goto, 2000).

Subsequently, the network was subjected to iterative gap filling. Incomplete pathways were found for the biosynthesis of Gly, His, Met, and Asn. To identify incomplete pathways, it was first ascertained whether the respective product can be produced under phototrophic conditions. Second, if no production was possible, it was established which set of reactions achieves production of this product in other organisms (other cyanobacterial species, and then broadening the search to plants). If applicable, candidate genes were identified in the Synechocystis genome using BLAST (http://blast.kazusa.or.jp/blast_search/cyanobase/genes). A list of all newly annotated candidate genes is provided as Supplemental File S3, along with details of the comparison and the respective sequence identities and E-values.

With respect to glyoxylate metabolism, we adopted the pathway suggested by Eisenhut et al. (2008). In particular, glyoxylate that is synthesized during photorespiration serves as a precursor for the amino acids Gly, Ser, and Cys. The substrate specificity for the transaminase (EC 2.6.1.-) coded by gene sll1559 remains unclear. Three scenarios are possible (Liepman and Olsen, 2001): the gene can code for a Ser-glyoxylate aminotransferase (EC 2.6.1.45), for a Ser-pyruvate aminotransferase (EC 2.6.1.51), or for an Ala-glyoxylate aminotransferase (EC 2.6.1.44), or any combination thereof. Again, the effects on the results of FBA are negligible, as there is no difference in the overall cofactor utilization for these reactions. Therefore, only the reaction for a Ser-glyoxylate aminotransferase was included in the network. Despite extensive literature research, several open questions and incomplete pathways remained. It was not possible to identify an enzyme for the amination of Asp to Asn, and no confirmed synthesis route for the amino acid Met could be found. As Met is an essential part of the BOF, we implemented a pathway via O-acetyl-homoserine. This pathway is postulated to exist in Microcystis aeruginosa NIES-843, which is the nearest relative of Synechocystis sp. PCC 6803 according to 16S rDNA analysis (Microbial Culture Collection at the National Institute for Environmental Studies, http://mcc.nies.go.jp/), but no candidate gene can be identified in Synechocystis sp. PCC 6803.

The reconstruction process continued with the identification of biosynthesis pathways for all compounds of the BOF, including the synthesis of purine and pyrimidine nucleotides, nucleotide sugars, and precursors of the lipid components of the biomass. Subsequently, the de novo biosynthesis routes of cofactors, namely NAD(P), CoA, glutathione, FAD, and tetrahydrofolate, were assessed. Transport mechanisms for the uptake of extracellular sulfate, hydrogen carbonate, and nitrate were added as well as for an exchange of oxygen and CO2. We assumed nitrate as the sole nitrogen source; uptake of urea or ammonia was not considered. Carbon uptake was limited to a single transport reaction. The reactions of photophosphorylation and the respiratory chain were adapted from Shastri and Morgan (2005) and Hong and Lee (2007). Finally, a catalase (sll1987) and a thioredoxin-disulfide reductase (slr0600) were added.

A persistent problem in network reconstruction is that for many enzymes the electron acceptors and donors are unknown (e.g. the Pro dehydrogenase [EC 1.5.99.8] or the glycolate dehydrogenase [EC 1.1.99.14]). In this case, it was assumed that NADH acts as the final electron donor and NAD as the final acceptor. Similarly, for many enzymes, the cofactor specificity with respect to NAD(H) and NADP(H) is unclear. In this case, the respective enzyme was assumed to catalyze both reactions. This choice has only marginal consequences on the results of FBA, since the network includes a NAD/NADP transhydrogenase (slr1239, slr1434), which allows a direct conversion between NADH and NADPH and vice versa. We note that FBA does not straightforwardly allow inclusion of quantitative differences in enzyme specificity but is primarily focused on stoichiometric information.

The reconstructed network is provided as Supplemental File S1 (SBML) and Supplemental File S2 (Excel). A complete list of reactions added during the reconstruction process is provided as Supplemental File S3. Supplemental File S6 provides a list of additional enzymes associated with Synechocystis sp. PCC 6803 metabolism available from KEGG but not included in our model. The final network model is compliant with MIRIAM standards (Le Novère et al., 2005) and contains links to CHEBI nomenclature (Degtyarenko et al., 2008), a standard endorsed in recent publications (Herrgård et al., 2008). For appropriate tools to visualize and interrogate the network, see the software guide at www.sbml.org.

BOF

FBA requires the definition of a BOF that specifies all cellular components in their appropriate molar ratios. A first BOF was provided by Shastri and Morgan (2005) and also used by Fu (2008). However, these BOFs were restricted mainly to precursors for amino acid compositions as constituents of the biomass. A considerably more elaborate function was given by Cogne et al. (2003) for a metabolic model of Arthrospira platensis, a blue-green photolithoautotroph cyanobacterium. The function is composed of five classes of macromolecules, namely carbohydrates, proteins, lipids, RNA and DNA, and the pigment chlorophyll. We utilized a modified version of the BOF of Cogne et al. (2003). Specifically, we neglected the flux to chlorophyll, corresponding to only 0.16% of total biomass. Additionally, the lipid demands were simplified to lipid precursors only, and the formation of carbohydrates was slightly simplified. The resulting BOF is given in Table I.

Reconciling Existing Models

Several previous reconstructions of Synechocystis sp. PCC 6803 are available. To our knowledge, the first reconstruction was provided by Shastri and Morgan (2005) and subsequently extended by Hong and Lee (2007). Both reconstructions are restricted to the primary metabolism, including glycolysis, the TCA cycle, and the Calvin-Benson cycle. The model of Shastri and Morgan (2005) encompasses 93 metabolic reactions (with reversible reactions counted twice), including 23 lumped reactions for the formation of amino acids and nucleotides.

A considerably larger genome-scale model was recently published by Fu (2008), encompassing about 831 metabolic reactions. However, the BOF was taken over from Shastri and Morgan (2005); therefore, the additional reactions cannot assume any functional role. Rather surprisingly, all previous reconstructions include a bacteria-like glyoxylate shunt, which was apparently suggested by Yang et al. (2002). We find no evidence for this assertion and assume that the inclusion of the glyoxylate shunt in previous reconstructions was mainly motivated by the omission of a C2 cycle.

FBA

FBA was based on available software tools and algorithms. All optimization studies were computed using the COBRA toolbox version 1.3.1 (Becker et al., 2007; available at http://gcrg.ucsd.edu/Downloads/Cobra_Toolbox) as well CellNetAnalyzer version 9.2 (Klamt et al., 2007; available at http://www.mpi-magdeburg.mpg.de/projects/cna/cna.html). The toolboxes were run under the MATBLAB 7.6.0 environment, including the MATLAB optimization toolbox as a Linear Program Solver. For the COBRA toolbox, a custom SBML file had to be generated (provided as Supplemental File S7). See also Oberhardt et al. (2009) and refs. therein for introductory tutorials on FBA.

For the constraint-optimization exchange fluxes, only the light input was preassigned, whereas nitrate, sulfate, and carbon (as hydrogen carbonate) uptake was allowed to adopt as necessary. In this way, we avoided the two-step optimization problem employed by Shastri and Morgan (2005), who first optimized for biomass formation using a fixed carbon uptake and free light flux, followed by a second optimization for minimal light flux, given the previously obtained biomass formation as an additional constraint. During our simulations, the uptake of hydrogen carbonate during phototrophic growth was constrained between 0 and 0.45 mmol HCO3 g−1 dry weight h−1, which is related to a maximum growth rate of 0.09 h−1. Limitation of growth by sulfate or nitrate was not considered. In addition to the BOF, we assumed a basal ATP demand (maintenance) for each FBA solution. As no reliable experimental data were available, the value was arbitrarily set to 10% of the maximal photosynthetic ATP generation used in Figure 2, corresponding to a value of 0.27 mmol g−1 dry weight h−1. We note that the basal ATP demand, contrary to the situation in heterotrophic organisms, has no influence on the FBA solution. Likewise, the FBA optimization problem does not take dilution of metabolites, including cofactors, into account.

Comparison of in Silico Knockout Mutants and in Vivo Measurements

A crucial test for network reconstruction and FBA is the ability to correctly predict lethal gene knockouts. In our case, as described in the main text, we observed a significant overlap between in silico mutants and in vivo experiments. Several of the in silico-predicted lethal genotypes can be confirmed by published data and information from the CyanoMutants repository (Nakamura et al., 1999). For example, the unavailability of a mutant lacking the folB gene (slr1626), coding for the dihydroneopterin aldolase, an important enzyme involved in the folate biosynthesis (see CyanoMutants Web site), indicates that there is only one possible pathway for folate biosynthesis. Correspondingly, the in silico gene deletion (folB) results in an inability of the model to produce the essential cofactor tetrahydrofolate.

Other examples include Quintero et al. (2000), who generated mutants showing an impaired Arg synthesis as a result of a knockout of gene sll0902, coding for an Orn carbamoyltransferase (ArgF). Also, an insertion in the gene (slr0661) coding a pyrroline-5-carboxylate reductase (ProC) resulted in impaired Pro synthesis. Both mutants had to be exogenously supplied by supplementation of the growth medium with either Arg or Pro, respectively. Autotrophic growth of either mutant could not be achieved. The phenotype of both mutants can be confirmed in silico. Under conditions of photoautotrophic growth, the genes coding for ArgF and ProC are classified as essential in the model but are dispensable if Arg and Pro are supplied in silico, respectively.

Nonetheless, discrepant findings were observed for four genes for which fully segregated mutants were reported, despite being essential in the model. On closer examination, two of these four genes correspond to the hydrogen carbonate transporter Cmp (slr0040, slr0041). In silico essentiality is due to Cmp being the only carbon transport mechanism present in our reconstruction. However, in Synechocystis sp. PCC 6803, more than one carbon uptake mechanism exists. A further gene (slr0847) encodes CoaD, a phosphopantetheine adenylyltransferase (EC 2.7.7.3) located in biosynthesis pathway of CoA. In our model, the enzyme catalyzes an essential step, the conversion from pantetheine 4′-phosphate to 3′-dephospho-CoA. Consequently, a mutant lacking this gene would be unable to synthesize CoA. The fact that it was possible to construct a viable mutant implies that the mutant was either not fully segregated or there exists an unknown alternative pathway for the synthesis of CoA in Synechocystis. The last discrepant gene (sll0934) is annotated as CcmA, a carboxysome formation protein, as well as a 3-deoxy-7-phosphoheptulonate synthase, which catalyzes the initial step of the aromatic amino acid synthesis. The latter is essential in our model. A fully segregated mutant, therefore, suggests the existence of alternative steps within the formation of aromatic amino acids (Ogawa et al., 1994).

Supplemental Data

The following materials are available in the online version of this article.

  • Supplemental File S1. Synechocystis network file, including identifiers for metabolites to the CHEBI and KEGG database and UniProt identifiers for genes (SBML format).
  • Supplemental File S2. List of reactions, associated enzymes, and genes implemented in the network with reference to literature (Excel file).
  • Supplemental File S3. Newly annotated genes for Synechocystis (Excel file).
  • Supplemental File S4. Comparison of gene knockout predictions in the model with in vivo data available from the CyanoMutants database (Excel file).
  • Supplemental File S5. Results of gene deletion study of all genes in the Synechocystis network (Excel file).
  • Supplemental File S6. List of additional enzymes associated with the Synechocystis metabolism available from the KEGG database (Excel file).
  • Supplemental File S7. Synechocystis network compatible for usage with COBRA toolbox version 1.3.1 (SBML format).
  • Supplemental File S8. Instruction and parameter settings to recalculate the results in Figures 2 to 4 using the COBRA toolbox version 1.3.1 (PDF file).

Supplementary Material

[Supplemental Data]

Acknowledgments

We thank N. Swainston for advice on SBML and metabolite nomenclature, Eva Grafahrend-Belau for advice on the BOF, and M. Hagemann and M. Eisenhut for advice on photorespiratory metabolism. We further thank all member of the FORSYS consortium Systems Biology of Cyanobacterial Biofuel Production for discussion and support.

References

  • Atsumi S, Higashide W, Liao JC. (2009) Direct photosynthetic recycling of carbon dioxide to isobutyraldehyde. Nat Biotechnol 27: 1177–1180 [PubMed]
  • Badger MR, Price GD. (2003) CO2 concentrating mechanisms in cyanobacteria: molecular components, their diversity and evolution. J Exp Bot. 54: 609–622 [PubMed]
  • Becker SA, Feist AM, Mo ML, Hannum G, Palsson BØ, Herrgard MJ. (2007) Quantitative prediction of cellular metabolism with constraint-based models: the COBRA toolbox. Nat Protoc 2: 727–738 [PubMed]
  • Boyle NR, Morgan JA. (2009) Flux balance analysis of primary metabolism in Chlamydomonas reinhardtii. BMC Syst Biol 3: 4. [PMC free article] [PubMed]
  • Caspi R, Altman T, Dale JM, Dreher K, Fulcher CA, Gilham F, Kaipa P, Karthikeyan AS, Kothari A, Krummenacker M, et al. (2010) The MetaCyc database of metabolic pathways and enzymes and the BioCyc collection of pathway/genome databases. Nucleic Acids Res 38: D473–D479 [PMC free article] [PubMed]
  • Chang A, Scheer M, Grote A, Schomburg I, Schomburg D. (2009) BRENDA, AMENDA and FRENDA the enzyme information system: new content and tools in 2009. Nucleic Acids Res 37: D588–D592 [PMC free article] [PubMed]
  • Cogne G, Gros JB, Dussap CG. (2003) Identification of a metabolic network structure representative of Arthrospira (Spirulina) platensis metabolism. Biotechnol Bioeng 84: 667–676 [PubMed]
  • Colman B, Norman EG. (1997) Serine synthesis in cyanobacteria by a nonphotorespiratory pathway. Physiol Plant 100: 133–136
  • Degtyarenko K, de Matos P, Ennis M, Hastings J, Zbinden M, McNaught A, Alcántara R, Darsow M, Guedj M, Ashburner M. (2008) CHEBI: a database and ontology for chemical entities of biological interest. Nucleic Acids Res 36: 344–350 [PMC free article] [PubMed]
  • Deng MD, Coleman JR. (1999) Ethanol synthesis by genetic engineering in cyanobacteria. Appl Environ Microbiol 65: 523–528 [PMC free article] [PubMed]
  • Eisenhut M, Kahlon S, Hasse D, Ewald R, Lieman-Hurwitz J, Ogawa T, Ruth W, Bauwe H, Kaplan A, Hagemann M. (2006) The plant-like C2 glycolate cycle and the bacterial-like glycerate pathway cooperate in phosphoglycolate metabolism in cyanobacteria. Plant Physiol 142: 333–342 [PMC free article] [PubMed]
  • Eisenhut M, Ruth W, Haimovich M, Bauwe H, Kaplan A, Hagemann M. (2008) The photorespiratory glycolate metabolism is essential for cyanobacteria and might have been conveyed endosymbiotically to plants. Proc Natl Acad Sci USA 105: 17199–17204 [PMC free article] [PubMed]
  • Feist AM, Herrgard MJ, Thiele I, Reed JL, Palsson BØ. (2009) Reconstruction of biochemical networks in microbial organisms. Nat Rev Microbiol 7: 129–143 [PMC free article] [PubMed]
  • Fu P. (2008) Genome-scale modeling of Synechocystis sp. PCC 6803 and prediction of pathway insertion. J Chem Technol Biotechnol 84: 473–483
  • Grafahrend-Belau E, Schreiber F, Koschützki D, Junker BH. (2009) Flux balance analysis of barley seeds: a computational approach to study systemic properties of central metabolism. Plant Physiol 149: 585–598 [PMC free article] [PubMed]
  • Herrgård MJ, Swainston N, Dobson P, Dunn WB, Arga KY, Arvas M, Blüthgen N, Borger S, Costenoble R, Heinemann M, et al. (2008) A consensus yeast metabolic network reconstruction obtained from a community approach to systems biology. Nat Biotechnol 26: 1155–1160 [PMC free article] [PubMed]
  • Hong SJ, Lee CG. (2007) Evaluation of central metabolism based on a genomic database of Synechocystis PCC6803. Biotechnology and Bioprocess Engineering 12: 165–173
  • Howitt CA, Vermaas WFJ. (1998) Quinol and cytochrome oxidases in the cyanobacterium Synechocystis sp. PCC 6803. Biochemistry 37: 17944–17951 [PubMed]
  • Hucka M, Finney A, Sauro HM, Bolouri H, Doyle JC, Kitano H, Arkin AP, Bornstein BJ, Bray D, Cornish-Bowden A, et al. (2003) The systems biology markup language (SBML): a medium for representation and exchange of biochemical network models. Bioinformatics 19: 524–531 [PubMed]
  • Kanehisa M, Goto S. (2000) KEGG: Kyoto Encyclopedia of Genes and Genomes. Nucleic Acids Res 28: 27–30 [PMC free article] [PubMed]
  • Klamt S, Saez-Rodriguez J, Gilles ED. (2007) Structural and functional analysis of cellular networks with CellNetAnalyzer. BMC Syst Biol 1: 2. [PMC free article] [PubMed]
  • Koksharova O, Schubert M, Shestakov S, Cerff R. (1998) Genetic and biochemical evidence for distinct key functions of two highly divergent GAPDH genes in catabolic and anabolic carbon flow of the cyanobacterium Synechocystis PCC 6803. Plant Mol Biol 36: 183–194 [PubMed]
  • Kun A, Papp B, Szathmary E. (2008) Computational identification of obligatorily autocatalytic replicators embedded in metabolic networks. Genome Biol 9: R51. [PMC free article] [PubMed]
  • Le Novère N, Finney A, Hucka M, Bhalla US, Campagne F, Collado-Vides J, Crampin EJ, Halstead M, Klipp E, Mendes P, et al. (2005) Minimum information requested in the annotation of biochemical models (MIRIAM). Nat Biotechnol 23: 1509–1515 [PubMed]
  • Liepman AH, Olsen LJ. (2001) Peroxisomal alanine:glyoxylate aminotransferase (AGT1) is a photorespiratory enzyme with multiple substrates in Arabidopsis thaliana. Plant J 25: 487–498 [PubMed]
  • Lindberg P, Park S, Melis A. (2010) Engineering a platform for photosynthetic isoprene production in cyanobacteria using Synechocystis as the model organism. Metab Eng 12: 70–79 [PubMed]
  • Mascarelli AL. (2009) Gold rush for algae. Nature 461: 460–461 [PubMed]
  • Nakahigashi K, Toya Y, Ishii N, Soga T, Hasegawa M, Watanabe H, Takai Y, Honma M, Mori H, Tomita M. (2009) Systematic phenome analysis of Escherichia coli multiple-knockout mutants reveals hidden reactions in central carbon metabolism. Mol Syst Biol 5: 306. [PMC free article] [PubMed]
  • Nakamura Y, Kaneko T, Hirosawa M, Miyajima N, Tabata S. (1998) CyanoBase, a www database containing the complete nucleotide sequence of the genome of Synechocystis sp. strain PCC 6803. Nucleic Acids Res 26: 63–67 [PMC free article] [PubMed]
  • Nakamura Y, Kaneko Y, Miyajima N, Tabata S. (1999) Extension of CyanoBase. CyanoMutants: repository of mutant information on Synechocystis sp. strain PCC 6803. Nucleic Acids Res 27: 66–68 [PMC free article] [PubMed]
  • Nakao M, Okamoto S, Kohara M, Fujishiro T, Fujisawa T, Sato S, Tabata S, Kaneko T, Nakamura Y. (2010) CyanoBase: the cyanobacteria genome database update 2010. Nucleic Acids Res 38: D379–D381 [PMC free article] [PubMed]
  • Navarro E, Montagud A, Fernandez de Cordoba P, Urchueguia JF. (2009) Metabolic flux analysis of the hydrogen production potential in Synechocystis sp. PCC 6803. Int J Hydrogen Energy 34: 8828–8838
  • Oberhardt MA, Palsson BØ, Papin JA. (2009) Applications of genome-scale metabolic reconstructions. Mol Syst Biol 5: 320. [PMC free article] [PubMed]
  • Ogawa T, Marco E, Orus MI. (1994) A gene (ccmA) required for carboxysome formation in the cyanobacterium Synechocystis sp. strain PCC 6803. J Bacteriol 176: 2374–2378 [PMC free article] [PubMed]
  • Omata T, Price GD, Badger MR, Okamura M, Gohta S, Ogawa T. (1999) Identification of an ATP-binding cassette transporter involved in bicarbonate uptake in the cyanobacterium Synechococcus sp. strain PCC 7942. Proc Natl Acad Sci USA 96: 13571–13576 [PMC free article] [PubMed]
  • Poolman MG, Fell DA, Raines CA. (2003) Elementary modes analysis of photosynthate metabolism in the chloroplast stroma. Eur J Biochem 270: 430–439 [PubMed]
  • Poolman MG, Miguet L, Sweetlove LJ, Fell DA. (2009) A genome-scale metabolic model of Arabidopsis and some of its properties. Plant Physiol 151: 1570–1581 [PMC free article] [PubMed]
  • Quintero MJ, Muro-Pastor AM, Herrero A, Flores E. (2000) Arginine catabolism in the cyanobacterium Synechocystis sp. strain PCC 6803 involves the urea cycle and arginase pathway. J Bacteriol 182: 1008–1015 [PMC free article] [PubMed]
  • Schwender J, Goffman F, Ohlrogge JB, Shachar-Hill Y. (2004) Rubisco without the Calvin cycle improves the carbon efficiency of developing green seeds. Nature 432: 779–782 [PubMed]
  • Shastri AA, Morgan JA. (2005) Flux balance analysis of photoautotrophic metabolism. Biotechnol Prog 21: 1617–1626 [PubMed]
  • Sielaff H, Christiansen G, Schwecke T. (2006) Natural products from cyanobacteria: exploiting a new source for drug discovery. IDrugs 9: 119–127 [PubMed]
  • Steuer R, Junker BH. (2009) Computational models of metabolism: stability and regulation in metabolic networks. Adv Chem Phys 142: 105–251
  • Stitt M, Sulpice R, Keurentjes J. (2010) Metabolic networks: how to identify key components in the regulation of metabolism and growth. Plant Physiol 152: 428–444 [PMC free article] [PubMed]
  • Tan LT. (2007) Bioactive natural products from marine cyanobacteria for drug discovery. Phytochemistry 68: 954–979 [PubMed]
  • Tcherkez GG, Farquhar GD, Andrews TJ. (2006) Despite slow catalysis and confused substrate specificity, all ribulose bisphosphate carboxylases may be nearly perfectly optimized. Proc Natl Acad Sci USA 103: 7246–7251 [PMC free article] [PubMed]
  • UniProt Consortium (2010) The Universal Protein Resource (UniProt) in 2010. Nucleic Acids Res 38: D142–D148 [PMC free article] [PubMed]
  • Van Liere L, Walsby AE. (1982) Interactions of cyanobacteria with light. Carr NG, Whitton BA., The Biology of Cyanobacteria, Ed 2 Botanical Monographs 19 Blackwell Scientific Publications, Oxford, pp 9–46
  • Varma A, Palsson BØ. (1994) Metabolic flux balancing: basic concepts, scientific and practical use. Bio/Technology 12: 994–998
  • Witzel F, Götze J, Ebenhöh O. (2010) Slow deactivation of ribulose 1,5-bisphosphate carboxylase/oxygenase elucidated by mathematical models. FEBS J 277: 931–950 [PubMed]
  • Yang C, Huaw Q, Shimizu K. (2002) Metabolic flux analysis in Synechocystis using isotope distribution from 13C-labeled glucose. Metab Eng 4: 202–216 [PubMed]

Articles from Plant Physiology are provided here courtesy of American Society of Plant Biologists
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...