Logo of plntphysLink to Publisher's site
Plant Physiol. 2010 Dec; 154(4): 1871–1885.
Published online 2010 Oct 25. doi:  10.1104/pp.110.166488
PMCID: PMC2996019

C4GEM, a Genome-Scale Metabolic Model to Study C4 Plant Metabolism1,[W][OA]


Leaves of C4 grasses (such as maize [Zea mays], sugarcane [Saccharum officinarum], and sorghum [Sorghum bicolor]) form a classical Kranz leaf anatomy. Unlike C3 plants, where photosynthetic CO2 fixation proceeds in the mesophyll (M), the fixation process in C4 plants is distributed between two cell types, the M cell and the bundle sheath (BS) cell. Here, we develop a C4 genome-scale model (C4GEM) for the investigation of flux distribution in M and BS cells during C4 photosynthesis. C4GEM, to our knowledge, is the first large-scale metabolic model that encapsulates metabolic interactions between two different cell types. C4GEM is based on the Arabidopsis (Arabidopsis thaliana) model (AraGEM) but has been extended by adding reactions and transporters responsible to represent three different C4 subtypes (NADP-ME [for malic enzyme], NAD-ME, and phosphoenolpyruvate carboxykinase). C4GEM has been validated for its ability to synthesize 47 biomass components and consists of 1,588 unique reactions, 1,755 metabolites, 83 interorganelle transporters, and 29 external transporters (including transport through plasmodesmata). Reactions in the common C4 model have been associated with well-annotated C4 species (NADP-ME subtypes): 3,557 genes in sorghum, 11,623 genes in maize, and 3,881 genes in sugarcane. The number of essential reactions not assigned to genes is 131, 135, and 156 in sorghum, maize, and sugarcane, respectively. Flux balance analysis was used to assess the metabolic activity in M and BS cells during C4 photosynthesis. Our simulations were consistent with chloroplast proteomic studies, and C4GEM predicted the classical C4 photosynthesis pathway and its major effect in organelle function in M and BS. The model also highlights differences in metabolic activities around photosystem I and photosystem II for three different C4 subtypes. Effects of CO2 leakage were also explored. C4GEM is a viable framework for in silico analysis of cell cooperation between M and BS cells during photosynthesis and can be used to explore C4 plant metabolism.

Many of the most productive crops in agriculture, such as maize (Zea mays), sorghum (Sorghum bicolor), and sugarcane (Saccharum officinarum), possess the Kranz leaf anatomy directly and causally associated with the C4 photosynthetic pathway (Laetsch, 1974; Hatch and Kagawa, 1976). Unlike C3 plants, where photosynthetic CO2 fixation proceeds in a single tissue, the mesophyll (M), in C4 plants, this process is distributed between M and bundle sheath (BS) cells (Edwards and Walker, 1983). In the Kranz anatomy, a ring of M, containing a few small chloroplasts concerned with the initial fixing of carbon dioxide, surrounds a sheath of parenchyma cells (BS), which have large chloroplasts involved in the Calvin-Benson cycle. BS cells have thick cell walls and contain centrifugally arranged chloroplasts with large starch granules and unstacked thylakoid membranes, whereas the M cells contain randomly arranged chloroplasts with stacked thylakoids and little or no starch (Kennedy et al., 1977; Edwards and Walker, 1983; Spilatro and Preiss, 1987). The fully differentiated BS and M chloroplasts each accumulates a distinct set of photosynthetic enzymes and proteins that enable them to cooperate in carbon fixation.

All C4 species operate on the same basic theme of pumping CO2 via C4 acids from M tissue, where phosphoenolpyruvate carboxylase (PEPC) activity is enhanced, to the BS layer, where Rubisco is localized and C4 acids are decarboxylated. Other than this, the only common feature shared by C4 plants is a reduction in the ratio of M to BS cells when compared with C3 plants. Because diffusion of organic acids between the M and BS must be relatively rapid, M cells in the C4 plants are rarely more than one cell distant from BS cells. As a result, M-to-BS ratios are between 1 and 2 in C4 plants, while they are over 4 in most laminate C3 leaves. Beyond these common features, C4 plants exhibit substantial variation in how they accomplish CO2 concentration. These variations result from three distinct decarboxylation modes and multiple patterns of anatomical modification. Biochemically, the major distinguishing feature of C4 subtypes is the enzyme used for the decarboxylation step in BS (Hatch, 1974, 1992; Hatch and Kagawa, 1976). The three carboxylation modes are NADP-malic enzyme (ME; present in maize, sugarcane, and sorghum), NAD-ME (present in Panicum miliaceum and Amaranthus spp.), and phosphoenolpyruvate carboxykinase (PCK; present in Panicum maximum and Amaranthus spp.). These different carboxylation modes are represented in Figure 1. It is important to note that these different mechanisms lead to different patterns of intercellular movements of C3 and C4 compounds. During C4 photosynthesis, atmospheric CO2 is initially fixed in M cells by PEPC to form C4 dicarboxylic acids. The movement of photosynthetic intermediates between M and BS cells is restricted largely or entirely to the plasmodesmata and transpirational water movement through the cell walls (Evert et al., 1977). The C4 dicarboxylic acids are transferred to Kranz cells (BS cells) and decarboxylated there. The released CO2 is refixed via ribulose-1,5-bisphosphate carboxylase of Kranz cell chloroplasts and incorporated into the Calvin-Benson cycle (Hatch and Kagawa, 1973, 1974, 1976). The transfer of atmospheric CO2 to BS cells is unique to C4 photosynthesis, since it leads to concentration of CO2 at the site of the reaction of Rubisco (Furbank et al., 1989, 1990b). This CO2-concentrating mechanism, together with specific properties of photosynthetic enzymes, configurations of intracellular and intercellular transporters, and modifications in leaf anatomy, allows C4 plants to achieve higher photosynthetic efficiency than C3 plants (Bull, 1969). C4 photosynthesis allows fast biomass accumulation with high nitrogen and water use efficiency (Smith, 1976; Sage, 2004).

Figure 1.
Schematic representation of the photosynthetic metabolism of three C4 subtypes distinguished according to the decarboxylating enzyme. Numbers refer to enzymes: (1) PEPC, (2) NADP-malate dehydrogenase, (3) NADP-ME, (4) pyruvate-Pi dikinase, (5) Rubisco, ...

Attempts have been made to engineer C4 metabolism in C3 plant species, such as rice (Oryza sativa), by overexpressing C4 enzymes (from maize; Suzuki et al., 2000, 2006; Ku et al., 2001; Agarie et al., 2002). However, significant increase in photosynthetic efficiency in the C3 mutants has yet to be achieved.

With the genome sequences available for sorghum (Paterson et al., 2009) and maize (Schnable et al., 2009), it is possible to perform a whole-genome-level exploration of C4 pathway evolution by comparing key photosynthetic genes in sorghum and maize (Wang et al., 2009; Yilmaz et al., 2009). These genome projects also facilitate the annotation of other important C4 crops for which incomplete genome sequences are available, such as sugarcane (also a NADP-ME type; Dufour et al., 1997; Guimarães et al., 1997). Along with the increase in genomic information, diverse data sets, including genomic, transcriptomic, proteomic, and metabolomic data, are becoming readily available for plants (Morgenthal et al., 2006; Thompson and Goggin, 2006; Dai et al., 2007; Yonekura-Sakakibara et al., 2008; Minic et al., 2009; Sawada et al., 2009), including C4 plant species (Majeran et al., 2005; Papini-Terzi et al., 2005; Majeran et al., 2008; Vicentini and Menossi, 2009).

The annotated genomes, along with literature data, define the known biological components of organisms. From these components, biological networks can be reconstructed and the properties of the system explored. In silico genome-scale metabolic reconstruction remains the most tangible and applicable systems biology approach for metabolic engineering (Blazeck and Alpert, 2010; Dietz and Panke, 2010; Tyo et al., 2010). A well-curated genome-scale network reconstruction is a common denominator for those studying systems biology of an organism. For these metabolic networks, a stoichiometric formalism (Edwards and Palsson, 2000; Thiele and Palsson, 2010) is natural and underpins a powerful set of analytical tools for exploring relationship between genotype and phenotype, predicting product yields and growth rates under different environments, and discovering global effects of genetic manipulations (Edwards and Palsson, 2000; Famili et al., 2003; Price et al., 2003; Quek and Nielsen, 2008; Oberhardt et al., 2009; Thiele and Palsson, 2010).

The most represented domain for genome-scale models is bacteria, with 25 species reconstructed (Oberhardt et al., 2009). Genome-scale metabolic networks have also been reconstructed for mouse (Sheikh et al., 2005; Quek and Nielsen, 2008), human (Mo et al., 2007; Sigurdsson et al., 2009), and more recently for Arabidopsis (Arabidopsis thaliana; Poolman et al., 2009; de Oliveira Dal’Molin et al., 2010). These reconstructions vary both in detail and in scope. For example, the Arabidopsis reconstruction of Poolman et al. (2009) considered two organelles (cytosol and mitochondria) to represent the plant metabolic network and hence was limited to describing heterotrophic cell culture. In contrast, our model (AraGEM; de Oliveira Dal’Molin et al., 2010) describes metabolism across cytosol, plastid, mitochondrion, peroxisome, and vacuole, which enabled the exploration of metabolic differences in photosynthetic and nonphotosynthetic cells and the examination in greater metabolic detail of cells undergoing photosynthesis, photorespiration, β-oxidation, or respiration. While both models covered the full range of known Arabidopsis reactions, validation was limited to primary metabolism (i.e. the synthesis of major biomass components, such as amino acids, nucleotides, lipid, starch, cellulose, and some vitamins). Nevertheless, both models are important contributions redressing the conspicuous lack of plant metabolic reconstructions (Oberhardt et al., 2009).

Genome-scale model reconstruction and validation of eukaryotes is certainly more challenging than in prokaryotic organisms, given the multiple organelles. However, such an approach applied for complex, multitissue plant systems would be of potentially even greater value, since direct measurement of fluxes using pulse-chase studies is practically impossible in plants due to signal dispersion in the large network as well as metabolite and pathway replication in multiple organelles. In this work, we attempted to use this systems-based framework to understand the interplay of metabolic networks in complex systems, like C4 plant metabolism. This, to our knowledge, is the first reconstruction used to formulate an in silico genome-scale model describing the interactions between the two tissue types, M and BS, involved in C4 metabolism. The predicted differences in flux distribution between M and BS cells were in agreement with observed proteomic studies and C4 metabolism features reported in the literature, suggesting that the model can be used to further explore C4 plant metabolism.


Characteristics of the Genome-Scale Model

The metabolic network reconstruction includes associations between genes, enzymes, and reactions to represent C4 plant metabolism based on the best available online resources (Table I).

Table I.
Online resources for the reconstruction of the metabolic network of C4 plants

Draft models were constructed from all reactions supported in the Kyoto Encyclopedia of Genes and Genomes (KEGG) for the three C4 plants sorghum, maize, and sugarcane. Inconsistencies in the KEGG database (e.g. multiple identifiers for single metabolites and unbalanced reactions) were resolved as described previously (Quek and Nielsen, 2008; de Oliveira Dal’Molin et al., 2010).

Subcellular compartmentalization plays an important role in plant metabolism. An initial attempt to use WoLF PSORT (http://wolfpsort.org/), a protein subcellular localization predictive tool (Horton et al., 2007), to assign compartments for each gene failed, as most calls were ambiguous or inconsistent across the three species. Instead, AraGEM (de Oliveira Dal’Molin et al., 2010) was used to make the compartmentalization calls for each reaction. AraGEM is manually curated for subcellular localization (cytosol, mitochondrion, plastid, peroxisome, or vacuole), and we have not identified any inconsistencies between localization in Arabidopsis and the limited number of proteins in C4 plants for which experimental localization data are available. The compartmentalized draft model covers 3,557 genes and 13,114 gene-reaction association entries for sorghum, 11,623 genes and 38,829 associations for maize, and 3,881 genes and 13,593 associations for sugarcane (Table II). The large number of gene-reaction associations reflects that, in the absence of localization information, genes associated with a reaction found in multiple compartments must be associated with each compartment.

Table II.
Characteristics of C4GEM based on the genomic information of sorghum, sugarcane, and maize

A functional model of primary metabolism, C4GEM, was derived from the draft model through manual curation. The model describes 1,588 unique reactions involving 1,755 metabolites (Table II). Forty-seven biomass drain equations describe the accumulation of carbohydrates, amino acids, fatty acid (palmitic acid only), cellulose, and hemicellulose, representing the major biomass drains for a plant cell, as well as some vitamins and cofactors (Table III). Eighteen intercellular exchange reactions (cytoplasm extracellular) have been included to describe the uptake of light (photons), assimilation/secretion of inorganic compounds (CO2, H2O, O2, NO3, NH3, H2S, SO42–PO43–), translocation of sugars (Suc, Glc, Fru, and maltose), and amino acids (Gln, Glu, Asp, Ala, and Ser). Together with biomass drains, the intercellular exchangers define the broad physiological domain of the model, that is, the curated aspects of primary C4 plant metabolism captured by C4GEM. A total of 83 interorganelle transporters were introduced in the model to achieve the desired functionality. The active, validated scope of C4GEM includes glycolysis (plastidic and cytosolic), pentose phosphate pathway (plastidic and cytosolic), tricarboxylic acid (TCA) cycle, light and dark reactions (Calvin cycle), NADH/NADPH redox shuttle between the subcellular compartments, fatty acid synthesis, β-oxidation, and glyoxylate cycle.

Table III.
List of biomass components, major drains

Although genome annotation can be used to derive a majority of plant metabolic reactions, there are missing reactions where the genome has not been fully elucidated. During the reconstruction, gaps were found in the primary metabolism for the three C4 models: 131, 135, and 156 gaps for sorghum, maize, and sugarcane, respectively. Some of these gaps could be filled by combining all of the C4 plant models. The remaining 100 essential reactions were all found in AraGEM (for a list of covered reactions, see Supplemental Table S1). The reconstruction allowed a functional annotation of the C4 plant models to ensure that reaction gaps are filled in the context of metabolic functions of nucleotide, carbohydrate, amino acids, and nitrogen metabolism, citrate cycle (TCA), carbon fixation, and pentose phosphate pathway.

The final model has 446 singleton or dead-end metabolites (i.e. internal metabolites only used in a single reaction; Supplemental Table S2). A total of 512 reactions are linked to the use or production of dead-end metabolites. These reactions will by definition have zero flux in flux balance analysis. There are several causes of singletons. Most are linked to vitamins, cofactors, and secondary metabolites not currently included in the model (i.e. not described by biomass drains or intercellular transporters). Some result from true gaps in the network, where one or more essential reactions have not yet been assigned. Finally, some result from KEGG’s automatic annotation, in which every reaction known to be catalyzed by a particular Enzyme Commission enzyme in some organism will be included by default, whether or not other reactions required for functionality of the particular reaction are present. Continued curation efforts will focus on resolving these singletons.

In its current form, the model has 183 degrees of freedom. Of these, 47 are associated with the production of biomass components and reduce to a single degree of freedom (growth rate) when a fixed biomass composition is assumed. Another 18 degrees of freedom are associated with intercellular transport. The remaining 118 degrees of freedom represent the maximum cellular scope for using alternative pathways to achieve identical outcomes in terms of growth rate and net transport (e.g. the use of cytosolic or chloroplastic glycolysis). The real scope is further limited by irreversibility constraints as well as regulatory constraints. C4GEM is available in System Biology Markup Language (SBML) format for maize, sorghum, and sugarcane (Supplemental Files S1–S3).

Two-Tissue Model

The active scope of C4GEM is similar to AraGEM (de Oliveira Dal’Molin et al., 2010). We have previously demonstrated how AraGEM can be used to explore metabolism in a single tissue (e.g. photorespiration in C3 M cells). The purpose of this study, however, is to explore the interactions between two distinct tissues, M and BS. Significantly, it represents the first step to represent a plant multitissue model.

C4GEM encapsulates the full complement of primary metabolic reactions available to any cell in a C4 plant. Tissue specificity is realized through differential gene expression (i.e. by constraining what part of the network is available to the individual tissues). Thus, M and BS tissue will be described as two instances of C4GEM differing only in terms of the constraints applied to the models (see below). The Kranz anatomy is modeled by connecting these two instances, M and BS, through transporters representing plasmodesmata. The movement of photosynthetic intermediates between M and BS cells (Fig. 1) is largely or entirely limited to the plasmodesmata (Evert et al., 1977). While plasmodesmata are capable of transporting a broad spectrum of low molecular mass (less than 800–900 D) metabolites, significant transport requires high concentration, and transporters were only added for major substrate carriers involved in the C4 photosynthetic pathway (subtypes NADP-ME, NAD-ME, and PCK), such as malate, pyruvate, phosphoglycerate, triose phosphates, Asp, Ala, and phosphates (pyrophosphate and orthophosphate; Furbank et al., 1989, 1990b; Fig. 1).

The final C4 model is described pictorially in Figure 2A. The model is based on the annotated genome for sorghum but uses AraGEM for compartmentalization and the 131 missing, essential reactions in primary metabolism. The model has been extended with 11 intercellular transporters (plasmodesmata), including Suc translocation and gases efflux.

Figure 2.
Elements of the reconstructed C4GEM network and the two-tissue model. ORFs, Open reading frames.

Mathematically, the large network is broken down into a stoichiometric matrix, S, with a row for each intracellular metabolite and a column for each reaction. Component balances governing a metabolic network of N = {1,.., n} chemical transformations and M = {1,.., m} chemical entities are

An external file that holds a picture, illustration, etc.
Object name is pp166488E01_LW.jpg

where Ci denotes the concentration of chemical entity i, Sij is the stoichiometric coefficient of chemical entity i in transformation j, and rj is the corresponding flux of transformation j. The time scale of metabolic reactions is short compared with regulation and growth dynamics, allowing us to invoke the steady-state assumption

An external file that holds a picture, illustration, etc.
Object name is pp166488E02_LW.jpg

In addition, some reactions are irreversible, so

An external file that holds a picture, illustration, etc.
Object name is pp166488E03_LW.jpg

With two tissues communicating, the model is duplicated and the metabolites in plasmodesmata, which are external (unbalanced) for the individual tissues, become internal and balanceable in the two-tissue model. In matrix form, the two-tissue model reads

An external file that holds a picture, illustration, etc.
Object name is pp166488E04_LW.jpg

where PM and PBS have a row for each of the 11 metabolites transported via plasmodesmata and a “1” entry in the column for the corresponding transporter. Thus, the third row dictates that the sum of metabolites exported into the plasmodesmata is 0, or, in other words, that whatever is exported from M must be taken up by BS (negative export).

The two-tissue model describes the network topology connecting inputs (e.g. photons, inorganic compounds) to outputs (e.g. biomass). The network is redundant (i.e. there are several paths through the network connecting any given input to any given output). This redundancy is inherent in all biological networks and confers flexibility and robustness to biological systems. The actual path chosen is defined by enzyme kinetics and regulation, and the current model does not consider these.

Equations 3 and 4 define the feasible flux space (i.e. the combinations of fluxes in M and BS that satisfy flux balance and irreversibility constraints). By careful formulation of additional assumptions and, in some instances, an optimality criterion, it is possible to explore the feasible space and formulate testable biological hypotheses, such as the following. Given a certain photon supply, what is the maximum rate of photosynthate production? What is the minimum number of photons required to support plant metabolism, and what is the optimal flux distribution? Is a particular gene deletion expected to affect growth and/or photosynthate production?

It is also possible to incorporate expression data from transcriptomics or proteomics to reduce the feasible solution space by deleting nonexpressed reactions. In the following sections, we will illustrate the use of constraint-based analysis.


In this study, we are interested in exploring (1) if the observed differences between M and BS metabolism in C4 plants are consistent with an assumption that the plant has evolved to use the network in an optimal manner, and (2) if the different C4 subtypes differ in their potential efficiency.

For a photosynthetic system, a logical optimality criterion is photon efficiency (i.e. a plant network will operate with the feasible flux distribution that minimizes photon uptake for fixed rates of biomass synthesis and export of carbon assimilates to other tissues). While no proof exists for the validity of this criterion, we observed for AraGEM that this criterion accurately predicted the classical photorespiration pathway as the most photon-efficient way to handle the effect of the Rubisco oxygenation reaction and that the predicted cost of photorespiration compared with pure photosynthesis was a 40% increase in photon usage for a 3:1 carboxylation:oxygenation ratio, which is consistent with experimental data (de Oliveira Dal’Molin et al., 2010).

Assumptions and Constraints

As a minimum definition of tissue specificity, Rubisco and PEPC activity are constrained to BS and M tissues, respectively, based on gene expression, enzyme activity, and proteome studies (Ghosh et al., 1994; Gutteridge and Gatenby, 1995; Dong et al., 1998; Patel and Berry, 2008; Table IV). These constraints are common to all three C4 subtypes. Mathematically, the tissue-specific activities are captured by equality constraints:

An external file that holds a picture, illustration, etc.
Object name is pp166488E05_LW.jpg
Table IV.
Set of constraints used for C4 subtypes based on the literature

To represent each C4 subtype, the enzyme used to release CO2 from the C4 acids was specified (NADPH-ME, NAD-ME, or PCK; Table IV). Similarly, the active C3-C4 transport systems have been specified based on subtypes. Finally, the drain of biomass components, starch accumulation, and export of carbon assimilates (Suc) was fixed based on average measured values.

It should be stressed that this is a minimum definition of C4 subtype metabolism used in this study, because we seek to explore questions around optimality. More detailed constraints (e.g. based on transcriptome analysis) may be used when exploring other questions.

Preferential Starch Accumulation in BS Cannot Be Explained by Photon Efficiency

In C4 plants, starch is principally synthesized in BS, although the amount of M starch biosynthesis is species dependent (Black and Mollenha, 1971; Spilatro and Preiss, 1987). In maize, for example, starch biosynthesis occurs only in the BS during normal photosynthesis, but plants growing in continuous light do accumulate starch in the M, suggesting that photosynthate availability is the limiting factor (Downton and Hawker, 1973).

One explanation for preferential starch biosynthesis in BS could be that it is “less expensive” (i.e. requires less photons) to accumulate starch in BS than in M, and hence BS stores are filled before M stores of starch. We tested this hypothesis by comparing the minimum photon requirement predicted with all starch produced in M to the minimum photon requirement for the standard model with all starch produced in BS. We did not observe an energy efficiency advantage for producing starch in BS rather than in M; thus, preferential starch production in BS cannot be argued to be the result of an evolutionary pressure for higher energy efficiency. Starch accumulation in BS only was added as an experimentally based constraint for subsequent analysis.

Suc Translocation

To sustain other plant tissues, fixed carbon must move out of the photosynthetic cells and into phloem (Braun and Slewinski, 2009). For most plants, this occurs by loading Suc into the phloem and transporting it from source tissues (net exporters) to sink tissues (net importers), where Suc is unloaded. Suc drain from BS to vascular parenchyma has been added to represent the translocation of carbon for the sustenance of the sink tissues. It is unclear if Suc produced in M or BS is transported through plasmosdesmata during C4 photosynthesis or if malate and pyruvate (for NADP-ME subtypes like maize) are the main carbon flow during the carbon assimilation. We tested this hypothesis in silico by allowing Suc exchange between both cells during C4 photosynthesis. According to the flux variability analysis (see “Materials and Methods”), Suc can be exchanged during this cooperative process, but malate and pyruvate show a much greater flux in magnitude (about 100-fold). Suc exchange between both cells did not affect the main contrasts in metabolic activity by comparing BS and M.

In Silico Flux Profile Correlates with Maize Chloroplast Proteome Studies

The optimal flux distribution predicted by the two-tissue model shows a number of reactions through which there is a significant difference in the flux observed in BS and M (Fig. 3). While the magnitude of flux is not directly linked to the amount of enzyme present, one might reasonably expect that a significant increase in flux in a tissue in most cases is matched by a significant increase in the expression of the corresponding enzyme. We compared metabolic flux predictions with a large data set comparing BS and M chloroplast proteomes in maize (Majeran et al., 2005, 2008; Friso et al., 2010). Since the optimality criterion does not necessarily yield a unique flux distribution, flux variability analysis (see “Materials and Methods”) was used to establish the range of values for each flux for which optimal photon usage could be achieved. All but four fluxes (ATP/ADP and oxaloacetate/malate transporters, glyceraldehyde 3-phosphate dehydrogenase, and triose phosphate isomerase) were uniquely defined by the optimality criterion.

Figure 3.
Flux map of the central metabolism. Relative flux distributions (BS/M) are shown between BS and M during C4 photosynthesis. Green and red lines represent increased and decreased flux in BS, respectively. Gray lines represent flux values that do not differ ...

For 50 of the 66 proteins or protein complexes annotated in the proteome assigned to enzymatic reactions, differential expression (or absence thereof) was consistent with predicted flux differences (Table V). Of these 50, two predictions were a direct consequence of the model assumptions for Rubisco and starch synthesis (Table IV). Twenty of the reactions were assigned to fatty acid biosynthesis, and the average and median BS-to-M ratios for proteins involved in fatty acids synthesis were 0.91 and 0.94, respectively, supporting the assumption that the demand for fatty acids is similar in M and BS.

Table V.
Regulation of key enzymes/transporters during C4 photosynthesis in M and BS predicted by C4GEM and correlated/not correlated to proteomics studies

The simulations show that flux through enzymes involved in carbon fixation, pentose phosphate pathway, and starch synthesis is up-regulated in BS, while flux through enzymes involved in purine, pyruvate, and nitrogen metabolism and nitrogen assimilation is up-regulated in M. The model also highlights contrasts in flux distribution through isoenzymes distributed in multiple organelles, such as higher flux through malate dehydrogenase in M plastids and higher flux through ME in BS plastids. Interestingly, in silico analysis shows no difference in flux through chloroplastic pyruvate dehydrogenase during photosynthesis, indicating an essential need to maintain malate and pyruvate pools required for operation of the C4 cycle. These flux predictions agree with the proteome studies.

Of the 16 reactions not accurately predicted, six are involved in chlorophyll and isoprenoid synthesis. While C4GEM has been validated for its ability to synthesize chlorophyll, the biomass content of chlorophyll was not specified in the model. Since BS chloroplasts have lower chlorophyll b levels, most of the predictions would agree with the proteome data if the content was accurately specified in the biomass equation. The reason for higher protoporphyrinogen oxidase activity in BS is unclear, given that the enzyme catalyzes an early step in chlorophyll biosynthesis. Another seven of the reactions not predicted in the simulated fluxes relate to the expected overexpression of oxidative stress response activities in M tissues. The model currently does not describe the generation of reactive oxygen species or their downstream products, and hence it does not predict the need for these activities. Similarly, the overexpression of three enzymes involved in degradation pathways was not predicted. The photon-optimal solution for making biomass clearly will not predict degradation, which is a wasteful process. The flux through maintenance and turnover processes needs to be expressed explicitly to be accounted for in the model.

The Model Predicts Cyclic Electron Flow in Bundle Chloroplast for NADPH-ME Subtype

Whereas linear electron transport in photosynthesis produces both ATP and NADPH, the cyclic electron transport (CET) around PSI has been shown to produce only ATP (Shikanai, 2007; Endo et al., 2008). Two alternative routes have been shown for CET: NAD(P)H dehydrogenase-dependent and ferredoxin:plastoquinone oxidoreductase-dependent flows, but their physiological relevance has not been elucidated in detail. Meanwhile, because C4 photosynthesis requires more ATP than does C3 photosynthesis to concentrate CO2, it has not been clear how the extra ATP is produced. Linear electron transport alone cannot produce enough ATP to satisfy the stoichiometry required for CO2 fixation. The shortage of ATP may be compensated for by the function of CET. Interestingly, two subtypes of C4 photosynthesis, NAD-ME and NADP-ME, have been shown to have different cell-specific ATP requirements. The ATP/NADPH ratio required in NAD-ME species is higher in M cells than in BS cells, but the opposite is true for NADP-ME species (Edwards and Walker, 1983; Takabayashi et al., 2005). This cell type-specific ATP requirement suggests that the activity of CET should be higher in M cells of NAD-ME species and in BS cells of NADP-ME species, if CET plays a critical role in supplying the additional ATP needed in C4 photosynthesis.

C4GEM predicts conventional linear electron transfer in M chloroplasts and CET in BS chloroplast for NADP-ME species (Fig. 4). One explanation for preferential CET in BS could be that it requires less photons to get the extra ATP in BS chloroplast than in M chloroplast. We tested this hypothesis by comparing the minimum photon requirement predicted with CET in M chloroplast with the minimum photon requirement for the model with CET in BS chloroplast. C4GEM highlights that CET in BS is energetically more efficient (about 25% less photons required) in NADP-ME-type metabolism than if CET is active in M. Our simulations are consistent with the proteome data showing high NAD(P)H dehydrogenase activity in BS (Table V), which is backed by previous studies showing that ndh gene expression is up-regulated in BS for NADP-ME subtypes (Dariel et al., 2006; Shikanai, 2007).

Figure 4.
Flux simulation for each C4 subtype with no leakage of CO2 and considering 50% of CO2 leakage out of BS. LET, Linear electron transport.

The Model Predicts Little Difference in Theoretical Quanta Requirement for the C4 Subtypes

The costs of concentrating CO2 in BS in NADP-ME- and NAD-ME-type plants consist of two ATP per CO2 fixed for regeneration of PEP, plus the amount of ATP required to pump extra CO2 (overcycle CO2) to compensate for the CO2 that leaks out of the BS. In PCK type, one extra ATP and 0.5 extra NADPH are required in addition to the amount required for overcycling.

Our flux simulations show little difference in the theoretical quantum yield of photosynthesis among the three C4 decarboxylation types (Fig. 4), also discussed in the literature (Shikanai, 2007). Such results indicate that these distinct biochemical subtypes among C4 plants represent different, equally optimal biochemical solutions to the same problem.

Although the theoretical yield is similar for all the C4 subtypes, in practice, differences in quantum yield are observed (Hatch et al., 1995). The reasons for the differences in quantum yield between decarboxylation types have proved elusive to identify. Our results suggest that these differences are not due the distinct C4 decarboxylation pathways. It has been suggested that if CO2 leakage (from BS back into M) occurs, this would change quantum yield between C4 plants (Furbank et al., 1989). However, estimates of leak rates show little correlation between C4 subtype and quantum yield. Our flux simulations show little change in total quanta requirement under the assumption that leakage occurs (Fig. 4), but the simulation shows considerable differences in activity of linear electron transport and CET around PSI and PSII for each of the C4 species. These results are consistent with the differences in photochemical activity in C4 subtypes observed in the literature (Edwards and Walker, 1983; Takabayashi et al., 2005; Romanowska and Drozak, 2006). As discussed previously, photosynthetic electron transport can involve either a linear flow from water to NADP, via PSII and PSI, or a cyclic flow just involving PSI. Because C4 subtypes have different cell-specific ATP requirements (Edwards and Walker, 1983; Dariel et al., 2006), there are differences in activity of PSI and PSII (Romanowska and Drozak, 2006). It was found, for example, that NAD-ME and PCK species have substantial PSII activity in the BS tissues, while NADP-ME species have little PSII activity.


C4GEM, to our knowledge, is the first genome-scale metabolic reconstruction of C4 plants. The two-tissue model based on C4GEM, to our knowledge, is the first attempt at constructing and integrating context-specific metabolic networks in multicellular interactions. The use of this model for in silico flux analysis illustrates the potential of using genome-scale models to explore complex and compartmentalized networks to derive nontrivial hypotheses. The in silico-predicted differences in the metabolic activity in BS and M during C4 photosynthesis agreed with observed differences in proteome data for many plastidic enzymatic reactions. The fact that in silico fluxes predicted assuming optimal photon use agree with observed proteome differences and C4 metabolic features reported in the literature lends support to the assumption that gene expression in BS and M has evolved to realize photon optimality.


Metabolic Reconstruction

The reconstruction process employed here is the same as the ones used for mouse (Quek and Nielsen, 2008) and Arabidopsis (Arabidopsis thaliana) GEMs (de Oliveira Dal’Molin et al., 2010). The workflow for the development and refinement of the C4 genome-scale model (C4GEM) is illustrated in Figure 5. The process of the metabolic reconstruction naturally lends itself to an iterative approach, and subsequent rounds of model refinement facilitate the testing of hypotheses.

Figure 5.
C4 genome-scale model reconstruction process.

The network reconstruction followed these steps. (1) Extraction of the metabolic information. A gene-centric organization of metabolic information was adopted, in which each known metabolic gene is mapped to one or several reactions. The metabolic elements contained in C4GEM (genomic metabolic reactions) were collected from KEGG (release 49.0, August 3, 2009; Kanehisa et al., 2008) for three C4 plant models: maize (Zea mays), sorghum (Sorghum bicolor), and sugarcane (Saccharum officinarum; stored in an Excel spreadsheet).

(2) Manual curation. The metabolic information was merged into a single generic plant cell model and subsequently curated using literature information. AraGEM was used as a core model to cover gaps found in primary metabolism of the C4 plant models. Curation was performed in Excel, accounting for cellular enzyme localization.

(3) Enzyme localization. Enzymes were assigned to different compartments according to literature evidence or enzyme localization databases (e.g. PPDB, a plant proteome database for Arabidopsis and maize).

(4) Gene association. In many cases, there are many genes for a particular enzyme (i.e. isoenzymes) found in multiple compartments. Attempts to assign localization based on in silico predictions failed (see “Results and Discussion”), and generally, all genes for a particular reaction had to be assigned to all organelles to which the reaction had been assigned in (3).

(5) Stoichiometric matrix. The set of unique reaction identifiers were extracted and stored as a stoichiometric matrix (Java application). In this step, multiple entries for a reaction in a particular compartment appearing in the Excel gene-enzyme-reaction table are collapsed to a single reaction entry. The stoichiometric matrix is used for flux balance analysis.

Intercellular and Extracellular Transport Elements

A multitissue model capturing the interactions between BS and M was formulated by combining two instances of the C4GEM and adding the movement of photosynthetic intermediates between M and BS cells (Fig. 1), which are restricted largely or entirely to the plasmodesmata (Evert et al., 1977). Transport through the bridge channel plasmodesmata was restricted to the main metabolites that give the characteristics of the C4 photosynthetic pathway inferred from the literature (Furbank et al., 1989, 1990b): that is, malate, pyruvate, phosphoglycerate, triose phosphates, and phosphates (pyrophosphate and orthophosphate; Fig. 1). The extracellular transporters were included for water and gases exchange flux through the cell wall (Evert et al., 1977), biomass drains for M and BS, and Suc drain from BS to vascular parenchyma to represent the translocation of carbon for the sustenance of the sink tissues (Braun and Slewinski, 2009). We have also considered gases efflux (CO2 and oxygen) to be exchanged through plasmodesmata to test the hypothesis, for example, that some percentage of CO2 that is released from the C4 acids in BS leaks back to M, as reported in the literature (Farguhar and Hatch, 1983; Furbank et al., 1990a; Hatch et al., 1995; see “Results and Discussion”).

Optimality Criterion

For the C4 metabolic model, the corresponding photon minimization solution can be expressed as

An external file that holds a picture, illustration, etc.
Object name is pp166488E06_LW.jpg

subject to

An external file that holds a picture, illustration, etc.
Object name is pp166488E07_LW.jpg

The last constraint can be used to define the previously mentioned irreversibility constraints as well as tissue specificity (e.g. zero flux for Rubisco in M). It is also used to define the fixed output of the system, in the form of biomass drains. To simplify comparisons, biomass synthesis rates were assumed to be the same for the two tissue types (except for starch; see below) and estimated based on literature values (Poorter and Bergkotte, 1992). Inorganic carbon was assumed to be limited in BS (no CO2 uptake) and free in M, where CO2 is initially fixed (by PEPC) to form C4 dicarboxylic acids. It was further assumed that the two tissues can freely exchange other inorganic compounds with the environment, while the exchange rate for organic compounds is assumed to be zero except for exchange via plasmodesmata.

Flux Simulations

Following manual curation, flux balance analysis was applied to confirm that each biomass component could be synthesized and in the subsequent model validation. C4GEM was compiled and curated as described for the compartmentalized C3 plant model (de Oliveira Dal’Molin et al., 2010). From the data collection, a two-dimensional reaction-centric SBML (www.sbml.org) database was generated. The stoichiometric matrix, S, and reversibility constraints (defining Vmin) were extracted from the SBML database in MATLAB (version 9; The MathWorks), and the linear programming problems were solved using the MOSEK Optimization Toolbox for MATLAB (version 6). Constraints were applied in MATLAB (C4Flux tool box) to represent cell type interactions (Fig. 2B). Reactions were activated or deactivated based on gene and enzyme activity studies (Westhoff et al., 1991; Bassi et al., 1995) for the specified cells (e.g. Rubisco is not present in M but present in BS chloroplasts). Biomass drains (see the following section for more details) were also specified based on literature values (Poorter and Bergkotte, 1992). Optimum flux distribution was simulated by linear programming and visualized on a metabolic flux map (which represents the central metabolism of a compartmentalized plant cell) drawn in Excel (Fig. 3).

Biomass Synthesis and Network Functionality

In the final step of the manual curation process, network gaps were identified based on the model’s ability to produce biomass components from substrates. Biomass drain reactions are incorporated into C4GEM as the accumulation terms of the biomass precursors (e.g. “Starch = Starch_biomass”) in order to simplify the task of uncovering the pathway gaps in the each of the biosynthetic routes separately. The list of biomass components considered is shown in Table II and includes major structural and storage components as well as trace elements such as vitamins.

For each biomass component, the following linear programming problem was formulated and solved

An external file that holds a picture, illustration, etc.
Object name is pp166488E08_LW.jpg

where ni is the corresponding biomass drain reaction. In this study, the problem was solved for leaf tissue (photons as energy source, CO2 as carbon source, and nitrate as nitrogen source) to represent biomass synthesis during C4 photosynthesis. The model can also be used for nonphotosynthetic tissues, when the minimization criterion is applied for carbon source usage (e.g. Suc).

Flux Variability Analysis

Due to the lack of experimental constraints, flux solutions generated by linear programming for large-scale models are often not unique. Despite having identical network topology and maintaining the same set of constraints and objective function, it is possible to have different flux distributions that give rise to the same objective function value, with each flux distribution representing a different metabolic state (e.g. different reaction utilization and/or directionality; Lee et al., 2000; Papin et al., 2002). Instead of a single flux distribution, it is therefore important to characterize the flux space encompassing all alternative optimal solutions before making reasonable interpretation of whether a metabolic phenotype is sufficiently different between two systems (e.g. BS versus M). An efficient linear programming-based strategy was used to quantify the range and variability of flux values, which involves minimizing and subsequently maximizing the flux value of all reactions in the network while keeping the same optimum objective function value (Mahadevan and Schilling, 2003). This method identifies the range of fluxes that is possible within all possible alternative optimal solutions. The following linear programming problem was formulated and solved

An external file that holds a picture, illustration, etc.
Object name is pp166488E09_LW.jpg


An external file that holds a picture, illustration, etc.
Object name is pp166488E10_LW.jpg

where vi represents each reaction in the network, c is the objective function vector, and Zobj is the objective function value calculated from the first optimization instance.

This method was used to identify and interpret the main contrasts between the two systems, M and BS, as shown in Table V (see “Results and Discussion”).

Supplemental Data

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

Supplementary Material

[Supplemental Data]


  • Agarie S, Miura A, Sumikura R, Tsukamoto S, Nose A, Arima S, Matsuoka M, Miyao-Tokutomi M. (2002) Overexpression of C4PEPC caused O2-insensitive photosynthesis in transgenic rice plants. Plant Sci 162: 257–265
  • Bassi R, Marquardt J, Lavergne J. (1995) Biochemical and functional properties of photosystem II in agranal membranes from maize mesophyll and bundle sheath chloroplasts. Eur J Biochem 233: 709–719 [PubMed]
  • Black CC, Mollenha H. (1971) Structure and distribution of chloroplasts and other organelles in leaves with various rates of photosynthesis. Plant Physiol 47: 15–23 [PMC free article] [PubMed]
  • Blazeck J, Alpert H. (2010) Systems metabolic engineering: genome-scale models and beyond. Biotechnol J 5: 647–659 [PMC free article] [PubMed]
  • Braun DM, Slewinski TL. (2009) Genetic control of carbon partitioning in grasses: roles of sucrose transporters and tie-dyed loci in phloem loading. Plant Physiol 149: 71–81 [PMC free article] [PubMed]
  • Bull TA. (1969) Photosynthetic efficiencies and photorespiration in Calvin cycle and C4-dicarboxylic acid plants. Crop Sci 9: 726–729
  • Dai SJ, Chen TT, Chong K, Xue YB, Liu SQ, Wang T. (2007) Proteomics identification of differentially expressed proteins associated with pollen germination and tube growth reveals characteristics of germinated Oryza sativa pollen. Mol Cell Proteomics 6: 207–230 [PubMed]
  • Dariel CC, De Pascallis L, Mutschler B, Haehnel W. (2006) Studies of the Ndh complex and photosystem II from mesophyll and bundle sheath chloroplasts of the C4-type plant Zea mays. Plant Physiol 163: 800–808 [PubMed]
  • de Oliveira Dal’Molin CG, Quek LE, Palfreyman RW, Brumbley SM, Nielsen LK. (2010) AraGEM, a genome-scale reconstruction of the primary metabolic network in Arabidopsis. Plant Physiol 152: 579–589 [PMC free article] [PubMed]
  • Dietz S, Panke S. (2010) Microbial systems engineering: first successes and the way ahead. Bioessays 32: 356–362 [PubMed]
  • Dong LY, Masuda T, Hata S, Izui K. (1998) Molecular evolution of C4-form PEPC in maize: comparison of primary sequences and kinetic properties with a newly cloned recombinant root-form PEPC. In Garab G, editor. , ed, Photosynthesis: Mechanisms and Effects, Vol V. Kluwer Academic Publishers, Dordrecht, The Netherlands, pp 3411–3414
  • Downton WJS, Hawker JS. (1973) Enzymes of starch and sucrose metabolism in Zea mays leaves. Phytochemistry 12: 1551–1556
  • Dufour P, Deu M, Grivet L, D’Hont A, Paulet F, Bouet A, Lanaud C, Glaszmann JC, Hamon P. (1997) Construction of a composite sorghum genome map and comparison with sugarcane, a related complex polyploid. Theor Appl Genet 94: 409–418
  • Edwards G, Walker DA. (1983) C3, C4: Mechanisms, and Cellular and Environmental Regulation, of Photosynthesis. University of California Press, Berkeley, CA
  • Edwards JS, Palsson BO. (2000) The Escherichia coli MG1655 in silico metabolic genotype: its definition, characteristics, and capabilities. Proc Natl Acad Sci USA 97: 5528–5533 [PMC free article] [PubMed]
  • Endo T, Ishida S, Ishikawa N, Sato F. (2008) Chloroplastic NAD(P)H dehydrogenase complex and cyclic electron transport around photosystem I. Mol Cells 25: 158–162 [PubMed]
  • Evert RF, Eschrich W, Heyser W. (1977) Distribution and structure of plasmodesmata in mesophyll and bundle-sheath cells of Zea mays L. Planta 136: 77–89 [PubMed]
  • Famili I, Forster J, Nielsen J, Palsson BO. (2003) Saccharomyces cerevisiae phenotypes can be predicted by using constraint-based analysis of a genome-scale reconstructed metabolic network. Proc Natl Acad Sci USA 100: 13134–13139 [PMC free article] [PubMed]
  • Farguhar GD, Hatch MD. (1983) On the nature of carbon isotope discrimination in C4 species. Aust J Plant Physiol 10: 205–226
  • Friso G, Majeran W, Huang MS, Sun Q, van Wijk KJ. (2010) Reconstruction of metabolic pathways, protein expression, and homeostasis machineries across maize bundle sheath and mesophyll chloroplasts: large-scale quantitative proteomics using the first maize genome assembly. Plant Physiol 152: 1219–1250 [PMC free article] [PubMed]
  • Furbank RT, Jenkins CLD, Hatch MD. (1989) CO2 concentrating mechanism of C4 photosynthesis: permeability of isolated bundle sheath cells to inorganic carbon. Plant Physiol 91: 1364–1371 [PMC free article] [PubMed]
  • Furbank RT, Jenkins CLD, Hatch MD. (1990a) C4 photosynthesis: quantum requirement, C4 acid overcycling and Q-cycle involvement. Aust J Plant Physiol 17: 1–7
  • Furbank RT, Jenkins CLD, Hatch MD. (1990b) The CO2 concentrating function of C4 photosynthesis. Current Research in Photosynthesis 1–4: D541–D544
  • Ghosh S, Glick BR, Heikkila JJ, Dumbroff EB. (1994) Phosphoenolpyruvate carboxylase and ribulose-1,5-bisphosphate carboxylase gene-expression in response to elevated-temperatures in C3 and C4 plants native to hot and temperate climates. Plant Physiol Biochem 32: 45–54 [PMC free article] [PubMed]
  • Guimarães CT, Sills GR, Sobral BWS. (1997) Comparative mapping of Andropogoneae: Saccharum L. (sugarcane) and its relation to sorghum and maize. Proc Natl Acad Sci USA 94: 14261–14266 [PMC free article] [PubMed]
  • Guinn G. (1966) Extraction of nucleic acids from lyophilized plant material. Plant Physiol 41: 689–695 [PMC free article] [PubMed]
  • Gutteridge S, Gatenby AA. (1995) Rubisco synthesis, assembly, mechanism, and regulation. Plant Cell 7: 809–819 [PMC free article] [PubMed]
  • Hatch MD. (1974) Mechanism and function of C4 pathway of photosynthesis. Proc Aust Biochem Soc 7: 1–P1
  • Hatch MD. (1992) C(4) photosynthesis: an unlikely process full of surprises. Plant Cell Physiol 33: 333–342
  • Hatch MD, Agostino A, Jenkins CLD. (1995) Measurement of the leakage of CO2 from bundle-sheath cells of leaves during C4 photosynthesis. Plant Physiol 108: 173–181 [PMC free article] [PubMed]
  • Hatch MD, Kagawa T. (1973) Enzymes and functional capacities of mesophyll chloroplasts from plants with C4-pathway photosynthesis. Arch Biochem Biophys 159: 842–853
  • Hatch MD, Kagawa T. (1974) NAD malic enzyme in leaves with C-pathway photosynthesis and its role in C4 acid decarboxylation. Arch Biochem Biophys 160: 346–349 [PubMed]
  • Hatch MD, Kagawa T. (1976) Photosynthetic activities of isolated bundle sheath cells in relation to differing mechanisms of C-4 pathway photosynthesis. Arch Biochem Biophys 175: 39–53 [PubMed]
  • Horton P, Park KJ, Obayashi T, Fujita N, Harada H, Adams-Collier CJ, Nakai K. (2007) WoLF PSORT: protein localization predictor. Nucleic Acids Res 35: W585–W587 [PMC free article] [PubMed]
  • Kanehisa M, Araki M, Goto S, Hattori M, Hirakawa M, Itoh M, Katayama T, Kawashima S, Okuda S, Tokimatsu T, et al. (2008) KEGG for linking genomes to life and the environment. Nucleic Acids Res 36: D480–D484 [PMC free article] [PubMed]
  • Kennedy RA, Barnes JE, Laetsch WM. (1977) Photosynthesis in C4 plant tissue cultures: significance of Kranz anatomy to C4 acid metabolism in C4 plants. Plant Physiol 59: 600–603 [PMC free article] [PubMed]
  • Ku MSB, Cho DH, Li X, Jiao DM, Pinto M, Miyao M, Matsuoka M. (2001) Introduction of genes encoding C4 photosynthesis enzymes into rice plants: physiological consequences. Rice Biotechnol 236: 100–116 [PubMed]
  • Laetsch WM. (1974) C4 syndrome: structural analysis. Annu Rev Plant Physiol Plant Mol Biol 25: 27–52
  • Lee S, Phalakornkule C, Domach MM, Grossmann IE. (2000) Recursive MILP model for finding all the alternate optima in LP models for metabolic networks. Comput Chem Eng 24: 711–716
  • Mahadevan R, Schilling CH. (2003) The effects of alternate optimal solutions in constraint-based genome-scale metabolic models. Metab Eng 5: 264–276 [PubMed]
  • Majeran W, Cai Y, Sun Q, van Wijk KJ. (2005) Functional differentiation of bundle sheath and mesophyll maize chloroplasts determined by comparative proteomics. Plant Cell 17: 3111–3140 [PMC free article] [PubMed]
  • Majeran W, Zybailov B, Ytterberg AJ, Dunsmore J, Sun Q, van Wijk KJ. (2008) Consequences of C4 differentiation for chloroplast membrane proteomes in maize mesophyll and bundle sheath cells. Mol Cell Proteomics 7: 1609–1638 [PMC free article] [PubMed]
  • Minic Z, Jamet E, San-Clemente H, Pelletier S, Renou JP, Rihouey C, Okinyo DPO, Proux C, Lerouge P, Jouanin L. (2009) Transcriptomic analysis of Arabidopsis developing stems: a close-up on cell wall genes. BMC Plant Biol 9: 6. [PMC free article] [PubMed]
  • Mo ML, Jamshidi N, Palsson BO. (2007) A genome-scale, constraint-based approach to systems biology of human metabolism. Mol Biosyst 3: 598–603 [PubMed]
  • Morgenthal K, Weckwerth W, Steuer R. (2006) Metabolomic networks in plants: transitions from pattern recognition to biological interpretation. Biosystems 83: 108–117 [PubMed]
  • Oberhardt MA, Palsson BO, Papin JA. (2009) Applications of genome-scale metabolic reconstructions. Mol Syst Biol 5: 320. [PMC free article] [PubMed]
  • Papin JA, Price ND, Edwards JS, Palsson BO. (2002) The genome-scale metabolic extreme pathway structure in Haemophilus influenzae shows significant network redundancy. J Theor Biol 215: 67–82 [PubMed]
  • Papini-Terzi FS, Rocha FR, Vêncio RZN, Oliveira KC, Felix JdeM, Vicentini R, Rocha CdeS, Simões ACQ, Ulian EC, di Mauro SM, et al. (2005) Transcription profiling of signal transduction-related genes in sugarcane tissues. DNA Res 12: 27–38 [PubMed]
  • Patel M, Berry JO. (2008) Rubisco gene expression in C4 plants. J Exp Bot 59: 1625–1634 [PubMed]
  • Paterson AH, Bowers JE, Bruggmann R, Dubchak I, Grimwood J, Gundlach H, Haberer G, Hellsten U, Mitros T, Poliakov A, et al. (2009) The Sorghum bicolor genome and the diversification of grasses. Nature 457: 551–556 [PubMed]
  • Poolman M, 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]
  • Poorter H, Bergkotte M. (1992) Chemical composition of 24 wild species differing in relative growth rate. Plant Cell Environ 15: 221–229
  • Price ND, Papin JA, Schilling CH, Palsson BO. (2003) Genome-scale microbial in silico models: the constraints-based approach. Trends Biotechnol 21: 162–169 [PubMed]
  • Quek LE, Nielsen LK. (2008) On the reconstruction of the Mus musculus genome-scale metabolic network model. Genome Inform 21: 89–100 [PubMed]
  • Romanowska E, Drozak A. (2006) Comparative analysis of biochemical properties of mesophyll and bundle sheath chloroplasts from various subtypes of C4 plants grown at moderate irradiance. Acta Biochim Pol 53: 709–719 [PubMed]
  • Sage RF. (2004) The evolution of C-4 photosynthesis. New Phytol 161: 341–370
  • Sawada Y, Akiyama K, Sakata A, Kuwahara A, Otsuki H, Sakurai T, Saito K, Hirai MY. (2009) Widely targeted metabolomics based on large-scale MS/MS data for elucidating metabolite accumulation patterns in plants. Plant Cell Physiol 50: 37–47 [PMC free article] [PubMed]
  • Schnable PS, Ware D, Fulton RS, Stein JC, Wei FS, Pasternak S, Liang CZ, Zhang JW, Fulton L, Graves TA, et al. (2009) The B73 maize genome: complexity, diversity, and dynamics. Science 326: 1112–1115 [PubMed]
  • Sheikh K, Förster J, Nielsen LK. (2005) Modeling hybridoma cell metabolism using a generic genome-scale metabolic model of Mus musculus. Biotechnol Prog 21: 112–121 [PubMed]
  • Shikanai T. (2007) Cyclic electron transport around photosystem I: genetic approaches. Annu Rev Plant Biol 58: 199–217 [PubMed]
  • Sigurdsson MI, Jamshidi N, Jonsson JJ, Palsson BO. (2009) Genome-scale network analysis of imprinted human metabolic genes. Epigenetics 4: 43–46 [PubMed]
  • Smith BN. (1976) Evolution of C4 photosynthesis in response to changes in carbon and oxygen concentrations in the atmosphere through time. Biosystems 8: 24–32 [PubMed]
  • Spilatro SR, Preiss J. (1987) Regulation of starch synthesis in the bundle sheath and mesophyll of Zea mays L.: intercellular compartmentalization of enzymes of starch metabolism and the properties of the ADPglucose pyrophosphorylases. Plant Physiol 83: 621–627 [PMC free article] [PubMed]
  • Suzuki S, Murai N, Burnell JN, Arai M. (2000) Changes in photosynthetic carbon flow in transgenic rice plants that express C4-type phosphoenolpyruvate carboxykinase from Urochloa panicoides. Plant Physiol 124: 163–172 [PMC free article] [PubMed]
  • Suzuki S, Murai N, Kasaoka K, Hiyoshi T, Imaseki H, Burnell JN, Arai M. (2006) Carbon metabolism in transgenic rice plants that express phosphoenolpyruvate carboxylase and/or phosphoenolpyruvate carboxykinase. Plant Sci 170: 1010–1019
  • Takabayashi A, Kishine M, Asada K, Endo T, Sato F. (2005) Differential use of two cyclic electron flows around photosystem I for driving CO2-concentration mechanism in C4 photosynthesis. Proc Natl Acad Sci USA 102: 16898–16903 [PMC free article] [PubMed]
  • Thiele I, Palsson BO. (2010) A protocol for generating a high-quality genome-scale metabolic reconstruction. Nat Protoc 5: 93–121 [PMC free article] [PubMed]
  • Thompson GA, Goggin FL. (2006) Transcriptomics and functional genomics of plant defence induction by phloem-feeding insects. J Exp Bot 57: 755–766 [PubMed]
  • Tyo KEJ, Kocharin K, Nielsen J. (2010) Toward design-based engineering of industrial microbes. Curr Opin Microbiol 13: 255–262 [PMC free article] [PubMed]
  • Vicentini R, Menossi M. (2009) The predicted subcellular localisation of the sugarcane proteome. Funct Plant Biol 36: 242–250
  • Wang XY, Gowik U, Tang HB, Bowers JE, Westhoff P, Paterson AH. (2009) Comparative genomic analysis of C4 photosynthetic pathway evolution in grasses. Genome Biol 10: R68. [PMC free article] [PubMed]
  • Westhoff P, Offermannsteinhard K, Hofer M, Eskins K, Oswald A, Streubel M. (1991) Differential accumulation of plastid transcripts encoding photosystem II components in the mesophyll and bundle-sheath cells of monocotyledonous NADP-malic enzyme-type C4 plants. Planta 184: 377–388 [PubMed]
  • Yilmaz A, Nishiyama MY, Jr, Fuentes BG, Souza GM, Janies D, Gray J, Grotewold E. (2009) GRASSIUS: a platform for comparative regulatory genomics across the grasses. Plant Physiol 149: 171–180 [PMC free article] [PubMed]
  • Yonekura-Sakakibara K, Tohge T, Matsuda F, Nakabayashi R, Takayama H, Niida R, Watanabe-Takahashi A, Inoue E, Saito K. (2008) Comprehensive flavonol profiling and transcriptome coexpression analysis leading to decoding gene-metabolite correlations in Arabidopsis. Plant Cell 20: 2160–2176 [PMC free article] [PubMed]

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


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...