Research into plant metabolism has a long history, and analytical approaches of ever-increasing breadth and sophistication have been brought to bear. We now have access to vast repositories of data concerning enzymology and regulatory features of enzymes, as well as large-scale datasets containing profiling information of transcripts, protein and metabolite levels. Nevertheless, despite this wealth of data, we remain some way off from being able to rationally engineer plant metabolism or even to predict metabolic responses. Within the past 18 months, rapid progress has been made, with several highly informative plant network interrogations being discussed in the literature. In the present review we will appraise the current state of the art regarding plant metabolic network analysis and attempt to outline what the necessary steps are in order to further our understanding of network regulation.

1. INTRODUCTION

Metabolism is perhaps the best characterized of all molecular-interaction networks in biology. From pioneering studies defining key metabolic pathways such as the Calvin cycle, through subsequent decades of enzymology characterizing the catalytic and regulatory properties of enzymes, through to more recent genetic studies of metabolism, there is an unprecedented density of both mechanistic and descriptive data relating to metabolic behaviour. Added to that list is the current accumulation of large molecular profiling datasets from transcriptomic, proteomic and metabolomic experiments. Nevertheless, despite this wealth of information, we still cannot accurately predict metabolic network behaviour. In the present review we wish to consider how we can take the next step forward in understanding the plant metabolic network. We will discuss what new information needs to be put into place, as well as detailing approaches through which the available experimental and theoretical information can be exploited and integrated.

One of the central goals of plant metabolic biology is to develop a sufficiently detailed view of the control hierarchy of the metabolic network to permit the rational design of genetic strategies to engineer metabolism for the overproduction of desirable end-products. This extends to placing heterologous enzymes in plants for the generation of novel (generally pharmaceutical) products [1], to altering the composition of crop plants for nutritional, health or flavour benefits [2] and to increasing the production of key secondary metabolites of commercial value [3,4]. Certainly plants have enormous untapped biosynthetic potential that could be exploited more fully by metabolic engineering [5], and there is considerable interest in the post-genomic analysis of metabolic networks [6,7]. However, it is fair to say that, to date, our efforts to engineer plant metabolism have largely met with failure. There are many more examples of unexpected or pleiotropic consequences of genetic intervention than there are of outcomes that matched prediction, and this is especially true in the more highly connected reactions of primary metabolism. A good example is the biosynthesis of storage starch in organs such as tubers and seeds. As an important dietary carbohydrate and with industrial uses as a food additive, there is much interest in increasing the yield of starch. Starch is synthesized in the plastid via a relatively simple linear pathway that is well characterized. Attempts to directly manipulate starch synthesis by targeting the enzymes catalysing these reactions have met with mixed results [8,9], but, surprisingly, manipulation of several enzymes that reside outside the direct pathway of starch synthesis have resulted in substantial increases in storage starch content [10] (Figure 1). These studies exemplify the two main reasons why we are currently so poor at predicting the outcome of metabolic-engineering experiments.

Effect of various transgenic manipulations on the starch content of potatoes

Figure 1
Effect of various transgenic manipulations on the starch content of potatoes

Reactions shown within the green box are those considered to be involved directly in the starch biosynthesis pathway, whereas those outside are only distally connected to starch metabolism. The blue boxes indicate the starch content of transgenic potato tubers (as a percentage of the wild-type value) for the indicated target enzymes. Abbreviations: ADK, adenylate kinase; ADPG, ADP-glucose; AGPase, ADP-glucose pyrophosphorylase; ANT, adenylate nucleotide translocase; ENZ, enzyme; G1P, glucose 1-phosphate; G6P, glucose 6-phosphate; REF, reference; UMPS, uridine monophosphate synthase. The photograph of the starch grain was kindly provided by Professor Alison Smith (John Innes Centre, Colney, Norwich, U.K.).

Figure 1
Effect of various transgenic manipulations on the starch content of potatoes

Reactions shown within the green box are those considered to be involved directly in the starch biosynthesis pathway, whereas those outside are only distally connected to starch metabolism. The blue boxes indicate the starch content of transgenic potato tubers (as a percentage of the wild-type value) for the indicated target enzymes. Abbreviations: ADK, adenylate kinase; ADPG, ADP-glucose; AGPase, ADP-glucose pyrophosphorylase; ANT, adenylate nucleotide translocase; ENZ, enzyme; G1P, glucose 1-phosphate; G6P, glucose 6-phosphate; REF, reference; UMPS, uridine monophosphate synthase. The photograph of the starch grain was kindly provided by Professor Alison Smith (John Innes Centre, Colney, Norwich, U.K.).

The first is a lack of a comprehensive knowledge of the contribution of individual enzymes to the control of metabolic flux. Despite the development of a formal mathematical framework for quantification of metabolic control [11,12] and an initial enthusiastic embracing of the ideas and methods of Metabolic Control Analysis by the plant community [13], flux control coefficients have been determined for only a small number of plant enzymes. In the starch example, analysis of metabolic control was only done belatedly after attempts to manipulate the pathway directly had failed. It was found that enzymes such as ADP-glucose pyrophosphorylase that were considered to be key control points actually had rather low flux control coefficients [14]. In fact, control of metabolic pathways is generally shared among all enzymes of the pathway, meaning that engineering of a single enzyme is unlikely to bring about significant flux change [15].

The second factor that contributes to our failure to predict the outcome of genetic intervention in metabolism is the predominance of the pathway paradigm. Metabolic pathways are convenient abstractions that allow metabolism to be broken down into accessible chunks and can justifiably claim to represent the main routes through the network in specific cases (e.g. glycolysis in yeast and the tricarboxylic acid cycle in pigeon muscle). However, in other organisms and tissues, the role of such pathways may be very different. For example, in developing oilseed embryos, the main flux through the tricarboxylic acid cycle is not cyclic at all, but rather follows a predominantly linear flux mode in which citrate is exported to support lipid biosynthesis [16]. Moreover, in photoautotrophic organisms, such as plants, in which all carbon-containing metabolites are ultimately derived from fixed CO2, it follows that all carbon-containing metabolites must in some way be linked together. Thus, in reality, metabolic pathways are just smaller subsets of a larger network, and there is potential for a change in one part of the network to have implications for any other part, no matter how distal [17]. This is clearly illustrated in the starch example, in which alteration of enzymes catalysing reactions in different subcellular compartments [NADME (NAD+-dependent malic enzyme) in the mitochondrion] and reactions only distantly connected to starch synthesis (uridine monophosphate synthase) nevertheless had major effects on the accumulation of starch. A final factor that complicates a predictive understanding of metabolism is the extent to which metabolism can be ‘rewired’ through altered expression of genes encoding metabolic enzymes in response to stress [18,19] or developmental progressions [20]. Ultimately one has to know which network one is dealing with, a fact that is often overlooked when dealing with genome-scale reconstructions of plant metabolism [21].

It is clear, therefore, that if we are to improve our ability to make rational manipulations of plant metabolism for engineering purposes, we have to generate a more complete view of metabolism as a network, rather than dealing with individual pathways, and that the new holistic view of the metabolic network must also include a quantification of the control exerted by each enzyme. Ultimately, it will be necessary to superimpose a gene-regulatory network on the metabolic network model so that the consequences of altered environmental or developmental circumstances on the structure and balance of the metabolic network can be accounted for. In the present review we consider several aspects of metabolic biology research that we believe will be central to attaining the goal of predictive metabolic engineering [22].

2. ESTABLISHING THE STRUCTURE OF THE PLANT METABOLIC NETWORK

Moving from a pathway to a network perspective immediately raises the question of how to obtain a description of the metabolic network of a specific plant, or more precisely, the metabolic network of a particular cell type in the plant at a given developmental stage in a defined environment. There is no ready-made source of such network descriptions! A traditional approach would be to assemble a network from consensus plant metabolic pathways in plant biochemistry textbooks and reviews and then to attempt to confirm the presence of each enzyme reaction in the target cell type by a literature search for experimental evidence. This could be assisted by using the metabolic maps from KEGG {Kyoto Encyclopedia of Genes and Genomes (http://www.genome.ad.jp/ [23]), though the only specific maps for higher plants are for Arabidopsis thaliana (thale cress) and Oryza sativa (rice)}, and searching for each enzyme in the database BRENDA {BRaunschweig ENzyme DAtabase (http://www.brenda.uni-koeln.de/ [24])} for organism-specific literature. The goal is to obtain a list of all the stoichiometrically balanced reactions. If the list also satisfies the other conditions discussed below, this gives a network description that can be a basis for the modelling and flux analysis methods described below in sections 3 and 5. However, this is clearly a large and labour-intensive approach that is far from guaranteed to result in a network where the presence of every reaction is fully validated.

In the post-genomic era, it might be imagined that this process can be automated for those plants whose genome has been sequenced. Indeed, since we understand exactly how the metabolic phenotype is coded in the genes, metabolism is a prime test case for whether we can use genome sequences to predict and understand phenotypes. Of course, the genome encodes the total metabolic potential of the organism, and it will still be necessary to consider what fraction of this is present in particular locations and circumstances. Genome-scale reconstructions of metabolism have already been undertaken for micro-organisms [2528], and show reasonably good agreement with experiment in terms of nutritional capabilities and susceptibility to loss of enzyme activities by mutation. However, except for the Saccharomyces cerevisiae (baker's yeast) metabolic model, the microbial metabolic networks do not need to make much allowance for compartmentation of metabolism, which will be a major issue for plant metabolism. Additionally, the microbial genomes are substantially smaller, so the task is somewhat easier than it will be for Arabidopsis.

The strategy for deriving the metabolic network from the sequence information is, in outline:

  • (i) Use the genome annotation to obtain a list of the EC numbers of all the genes identified as coding for enzymes, either by experiment or sequence homology.

  • (ii) Use the EC numbers and a database such as KEGG [23] or BioCyc™ (http://biocyc.org/ [29]) to obtain the reactions catalysed by each of the enzymes and add them to the reaction list. These databases are suggested because the metabolites participating in the reactions are identified by compound ID (identification) numbers that link to database entries giving synonyms, empirical formulae etc.

In principle, this could be automatic; however, networks obtained in this way exhibit various deficiencies [30], and the process currently requires manual intervention [25,31]. The problems arise from a number of different areas:

  • (i) The genome annotation is likely to be incomplete, since not all potential coding sequences can be assigned a function, and those that are may not be identified accurately or sufficiently precisely. Assignment by homology with known enzymes can be wrong, but in many cases it is incomplete, giving a partial identification in terms of a partial EC number that specifies enzyme class and subclass, but not a specific set of reactions. On the other hand, the annotation can implicate too many enzymes, since only a subset of the genome will be transcribed in given circumstances, and some sequences assigned to open reading frames and annotated may nevertheless be pseudogenes that are never transcribed. Other sources of incompleteness are that some metabolic reactions are spontaneous and not linked to genes, and that the transport reactions between intracellular compartments must also be included in the network, but the transporter proteins are often not identified at a molecular level. The latter is a particular problem with plants; for example, the Arabidopsis annotation contains numerous putative phosphate-linked translocators of undefined specificity and localization.

  • (ii) There are database errors, including incorrectly specified reactions, and chemically unbalanced equations.

  • (iii) There are systemic errors, where the reactions individually are described in a self-consistent manner, but do not combine to generate a consistent network. For example, when the same compound appears in the list with more than one ‘unique ID’ (see [30]), routes through the network are broken, since a product of one reaction is not recognized as the substrate of another.

These issues are discussed more fully by Poolman et al. [30]. The outcome is that models derived directly from the databases, and to some extent published genome-scale metabolic networks, exhibit a variety of unlikely and unsatisfactory characteristics, including separation of the network into a number of disconnected components, unbalanced reactions that violate conservation of mass, a significant fraction of ‘orphan’ metabolites that are either produced or consumed, but not both, and ‘dead’ reactions that can be shown to be unable to carry any metabolic flux at steady state.

The problems encountered in microbial genome-scale models are magnified in models of eukaryotic metabolism, and especially those of plants, by the additional uncertainty concerning to which compartment particular gene products are directed, although, for Arabidopsis at least, this is starting to be addressed (see section 3.1). If these problems can be resolved, the maximal metabolic network, obtained as described above, needs trimming back to the set of enzymes being expressed in a particular set of cells. The measurement of individual enzyme activities is one possibility, but is not currently automated for high-throughput analysis as are the other potential methods. Similarly, proteomics screens provide another set of clues concerning the presence of particular proteins in their active states, but lack the precision of comparisons of enzyme activities and could be misleading for the reasons given below (in section 4.2). At another level, microarrays can detect which genes are being transcribed, and, therefore, indicate which activities may be present in particular cells, as discussed below.

3. CONSTRUCTING MODELS OF PLANT METABOLISM

A system of the complexity of the plant metabolic network will only be fully encapsulated and understood through the construction of computer models. The ultimate goal of the modelling exercise is to fully describe the flux through all reactions in the network and to be able to predict how those fluxes change in response to different external environments or when the network is altered through genetic intervention. This is a considerable challenge. However, the convergence of a maturation of molecular profiling technologies and developments in the application of computational approaches means that we are now well poised to meet this challenge. In this section, the main approaches to modelling metabolism are explained and the relative advantages and limitations of each described. Examples of existing models of plant metabolism are used to illustrate the benefit of the approach in terms of understanding the control distribution within a set of connected reactions. Finally, we discuss some of the challenges that must be faced if we are to move towards a predictive model of the plant metabolic network.

3.1 Kinetic models

It has long been recognized that a knowledge of the kinetic properties of the enzymes that catalyse a network of reactions allows the flux through those reactions to be numerically simulated [32]. The principle can be illustrated by considering a reaction in which an enzyme catalyses the conversion of a single substrate into a single product via Michaelis–Menten kinetics (Figure 2). Given the concentration of substrate and the values for Km and Vmax, one can calculate the velocity (v) of the reaction according to the Michaelis–Menten equation:

 
formula

In reality, relatively few enzymes have such simple kinetics. Most enzymes catalyse reactions involving more than one substrate, and factors such as product inhibition and the presence of effector molecules lead to considerably more complicated rate equations. Nevertheless, the principle remains: given the reaction stoichiometry, rate equation and the values for the kinetic parameters of each enzyme in the metabolic system under consideration, then the system can be numerically simulated if one provides starting values for metabolite concentrations. This type of modelling has a long history, dating back to the work of Garfinkel and Hess [33], who modelled glycolysis in ascites-tumour cells, and several metabolic pathways in plants have been tackled using this approach. For a review of how this type of modelling has improved our understanding of several key pathways of primary carbon metabolism, see [34].

Methods for modelling metabolic networks

Figure 2
Methods for modelling metabolic networks

In constraints-based modelling, known properties of the network (network structure, reaction stoichiometry, thermodynamics and the requirement for flux balance), as well as experimentally measured parameters, are used to constrain the solution space (shown in grey). Within this bounded space, a unique solution is found (shown in green) that optimizes an objective function such as biomass production. In kinetic modelling, kinetic rate equations are used to describe the relationship between substrate concentration and reaction velocity. Sets of such equations are used as the basis for numerical simulation of network behaviour. Abbreviations: F1,6P2, fructose 1,6-diphosphate; F6P, fructose 6-phosphate; G6P, glucose 6-phosphate; HXK, hexokinase; PFK, phosphofructokinase; PGI, phosphoglucose isomerase; [t], time.

Figure 2
Methods for modelling metabolic networks

In constraints-based modelling, known properties of the network (network structure, reaction stoichiometry, thermodynamics and the requirement for flux balance), as well as experimentally measured parameters, are used to constrain the solution space (shown in grey). Within this bounded space, a unique solution is found (shown in green) that optimizes an objective function such as biomass production. In kinetic modelling, kinetic rate equations are used to describe the relationship between substrate concentration and reaction velocity. Sets of such equations are used as the basis for numerical simulation of network behaviour. Abbreviations: F1,6P2, fructose 1,6-diphosphate; F6P, fructose 6-phosphate; G6P, glucose 6-phosphate; HXK, hexokinase; PFK, phosphofructokinase; PGI, phosphoglucose isomerase; [t], time.

The great attraction of kinetic models is that they are mechanistic models based on differential equations that implicitly encapsulate the relationship between substrate/effector concentrations and reaction velocity. As such they are predictive. The effect of altering enzyme properties or amount can easily be simulated by altering the relevant parameters in the equations. Furthermore, the information from such in silico analyses can be used to provide a formal quantification of the extent to which each enzyme controls pathway flux within the framework of Metabolic Control Analysis [11,12,35]. This ability to rapidly interrogate the control distribution of a metabolic network and to examine in detail the consequences of alteration of specific components of that network means that kinetic models are extremely useful tools for the design of strategies to engineer plant metabolism. However, despite the undoubted power of kinetic models, their application to plant metabolism has been severely hampered by a lack of data [36]. The quality of a metabolic model is dependent on the completeness of the enzyme kinetic data used to parameterize the model. Metabolic systems in which the kinetics of all enzymes have been comprehensively characterized are the exception rather than the rule. Even where there is information in the literature for all enzymes of a system, invariably they are drawn from several different plant species. Ultimately, the more parameters that have to be assumed or ‘borrowed’ from other systems, the less reliable the model will be. Hence, kinetic models of plant metabolism have been confined to the consideration of specific subsets of metabolism such as the Calvin cycle or starch synthesis [34].

Traditionally, the values for the kinetic parameters of enzymes used in such models have been drawn from the enzymology literature. Enzymes are purified from plant tissues or recombinant enzymes are purified from bacteria, are and the kinetics of the purified enzyme are studied in vitro. The time-consuming and exacting nature of this work has inevitably contributed to the incomplete picture of enzyme kinetics that we are faced with today. The development of experimental approaches which allow enzyme kinetics to be estimated in a less labour-intensive fashion would clearly be of huge benefit and would allow an expansion of the scale and scope of kinetic models of plant metabolism. One promising approach in this regard is the derivation of enzyme kinetics from in vivo labelling experiments. An isotopically labelled metabolic precursor is supplied to plant cells, organs or tissues in culture and a time series of samples are taken. The kinetics of accumulation of the label in intermediates of the metabolic reactions under consideration can be determined by analysis of the samples. Generally 13C-labelled precursors are used, allowing analysis by either NMR or MS. The enzyme kinetics that lead to the observed labelling kinetics can then be modelled by data-fitting procedures [37,38].

This parameter-fitting approach has been used extensively in the modelling of microbial metabolism [39], but is still a relatively unexplored method for modelling plant metabolism. In fact, the development of the approach in plants has been undertaken almost single-handedly by David Rhodes and colleagues at Purdue University [40,41] and has only recently been taken up by other groups [42]. To date, the approach has been exclusively used to model plant secondary metabolism. This in part reflects the interest in engineering plant secondary metabolism, but also that secondary pathways are well suited to kinetic modelling. In particular, the pathways are often unbranched, making the recapitulation of enzyme kinetics from labelling curves computationally less demanding. A seminal example of the approach was the work by Andrew Hanson's group (in collaboration with David Rhodes and Yair Shachar-Hill) on glycine betaine (1-carboxy-N,N,N-trimethylmethanamimium inner salt) metabolism [41]. Glycine betaine is a compatible solute that is important in many plants in protecting against dehydration stress. Many plants lack the ability to synthesize glycine betaine and thus it is an established target for metabolic engineering [43]. By following the time-dependent redistribution of label from [14C]choline supplied to young tobacco (Nicotiana tabacum) leaves, it was possible to collect sufficient labelling information that the kinetics of the enzymes of glycine betaine synthesis could be fitted. The resultant enzyme kinetic parameters were used to numerically simulate glycine betaine metabolism, and a detailed analysis of metabolic capacities and fluxes was undertaken. The work revealed that the import of choline into the chloroplast is a major limiting constraint on glycine betaine synthesis. Transport of choline across the chloroplast inner membrane was thus revealed as a key target for engineering glycine betaine synthesis.

A similar approach was used more recently to investigate phenylpropanoid metabolism in potato (Solanum tuberosum) tubers in response to fungal elicitors [42]. In this technically and mathematically sophisticated work, values for initial metabolite concentrations, rate constants and exponents of power law equations were fitted to label distribution kinetics determined by liquid chromatography–MS. Parameter fitting was done by the least-squares method, and the output was analysed for statistical reliability. Not only was it possible to reliably estimate 18 fluxes in the phenylpropanoid pathway and demonstrate that fluxes were altered by the elicitor treatment, but also the analysis extended to a quantification of the control distribution within the phenylpropanoid pathway, in both the presence and absence of the elicitor.

Whether one is modelling flux from the kinetics of metabolite pool size or labelling change, the method has a number of advantages. In addition to the obvious benefit of less labour-intensive access to enzyme kinetics, the approach also has the advantage that the derived kinetic parameters are effectively in vivo kinetics, rather than in vitro, as is traditional. Numerous factors in the in vivo environment can influence the behaviour of enzymes. These include interaction with other proteins, modifications of post-translational state and the presence of effector metabolites (see section 5). Given that such factors are often absent or not faithfully replicated in vitro, it is doubtful as to whether kinetic parameters determined in vitro are an accurate reflection of the properties of the enzyme in vivo [44]. Therefore models based on in vivo kinetic parameters have a greater potential to successfully encapsulate the true behaviour of metabolic networks. Nevertheless, there are still some challenges to be overcome. Without doubt, the most significant of these is to decide on the form of the rate equation to be applied. In the absence of a priori knowledge of the kinetic behaviour of an enzyme, it is necessary to apply a general form of the rate equation. In some cases, simple first-order kinetics have been used, but saturation kinetics based on the Michaelis–Menten equation or power laws are more realistic. Of course, such a general kinetic form will not be appropriate for all enzymes in the system and it may be desirable to leave the form of the rate equation unconstrained. Given sufficient experimental data, a more global modelling approach can be shown to successfully recapitulate known kinetic forms and indeed identify putative novel interactions and effectors [45]. A second issue to deal with in the parameter-fitting approach is the extent to which there is a unique solution to the parameter set. In many cases there will be more than one solution, and one is then faced with the problem of which one predicts the correct values for enzyme kinetic parameters. This is a critical issue, because, although multiple solutions may faithfully model the behaviour of the experimental system, only the ‘correct’ solution will have genuine predictive power. To avoid this ‘non-unique-solution’ scenario, it is probably necessary to impose additional constraints on the values of the kinetic parameters using a priori knowledge.

3.2 Stoichiometric flux-balance models

In contrast with the kinetic modelling approach, in which extensive knowledge of enzyme kinetics is required, stoichiometric modelling is based primarily on a knowledge of the structure of the metabolic network under consideration [46]. In its simplest form, the structure of the network itself forms the central aspect of investigation. In a landmark paper, Stelling et al. [47] demonstrated, through analysis of elementary modes (the least decomposable metabolic units that support a steady state), that, when genes are systematically deleted, Escherichia coli maintains a maximally efficient metabolic network structure for a given carbon source. More powerful still, is a constraints-based approach in which known features of the network, such as reaction stoichiometry, thermodynamics (reversibility of reactions in vivo) and the demands of maintaining a steady state, impose constraints on the possible flux solution space [28,48] (Figure 2). Linear programming algorithms can be used to find a unique solution that satisfies an objective function, such as maximal growth rate within the constrained solution space. Although care has to be taken to ensure that sufficient constraints are used to overdetermine the network [49], the approach has been used with some success to model microbial metabolic networks [2527,50].

These are genuine network-scale models consisting of between 300 and 1100 reactions and give a depth of coverage of metabolic flux that has not been attained with other methods. In addition, such models do more than just providing a snapshot of fluxes throughout the system (which is in itself highly informative). The models can also be predictive within the limitations of certain assumptions about the behaviour of the network. For example, if one assumes that the network is structured to maintain maximal possible growth, then one can use the linear programming optimization to explore the possible flux distributions as the nature and amount of carbon source is varied. This has allowed the identification of a core set of reactions in different micro-organisms that carry a positive flux under all conditions [51] and that the network contains a high flux backbone (consisting mainly of primary metabolism), whereas branch fluxes are lower and more variant [52]. Similarly, if one assumes that the network reacts to the removal of a gene product by minimizing metabolic change, then one can predict the effect of gene mutation on metabolic fluxes [53]. Using this approach, the majority of growth phenotypes of gene knockouts in E. coli can be successfully predicted [54].

To date, the constraints-based approach has been mainly applied to microbial metabolism. However, all the necessary information is available to use the approach to analyse the metabolic network of model plants such as Arabidopsis. In particular, there are genome-scale curated reconstructions of the Arabidopsis metabolic network [21], which form a starting point for defining the network structure (see section 2). In addition, there is an extensive archive of transcriptomic datasets which can be mined for robust transcript-to-transcript correlations that may indicate the presence of functionally co-expressed enzyme subsets. Definition of functional subsets can considerably reduce the complexity of the network under consideration. Moreover, there are well-established Arabidopsis cell-suspension cultures that facilitate measurement of constraints such as metabolite inputs and outputs and rate of biomass accumulation. Although the development of a constraints-based genome-scale model of Arabidopsis metabolism would represent a major advance, the aim is an ambitious one and there are many potential hurdles to be overcome. Not the least of these hurdles is the complexity and degree of subcellular compartmentation of plant metabolism. However, on the basis of transcriptomic information [18], only 600–900 metabolic transcripts are expressed in dark-grown heterotrophic Arabidopsis cells in culture, giving a network of similar size and complexity to that of yeast. Reconstruction of a properly compartmented metabolic network requires accurate and comprehensive information about enzyme subcellular localization. Although there are clearly large gaps in the subcellular protein catalogue, organellar proteomic efforts are gathering pace [55,56], and information from a variety of sources is being gathered into single database points [57]. In addition, there has been steady progress in our characterization of integral membrane transporters, particularly at the mitochondrial, plastidic and vacuolar membranes [5860].

3.3 Future developments in metabolic modelling

Kinetic models and stoichiometric flux-balance models represent two extremes of the modelling spectrum. Kinetic models are detailed mechanistic descriptions of the system that provide quantitative prediction of network dynamics. However, they demand significant experimental parameterization and are difficult to scale up to cover the whole network. By contrast, stoichiometric flux-balance models work well at the network scale and generate quantitative predictions about network behaviour, but only within a single state of the system. Developments in both spheres are likely to expand the scope of each approach. The ability to engineer affinity-tagged enzymes at high throughput raises the possibility of large-scale analyses of enzyme kinetics, and analysis of isotope-labelling kinetics may provide an alternative and ready means of accessing enzyme kinetics in vivo. For stoichiometric modelling, an interesting development is the inclusion of thermodynamic constraints, thereby providing a link between metabolite concentration and flux [61], which may allow aspects of metabolic regulation to be inferred. Another important development is the addition of a gene-regulatory model to the stoichiometric model, allowing the response of the system to different conditions to be encapsulated in what is otherwise a static modelling approach [62]. The logical extension of this is to add proteomic and metabolomic data to the mix, and methods are being developed to allow the incorporation of such heterogenous and incomplete datasets [63,64].

Finally, an interesting new approach is structural-kinetic modelling [65,66]. This sits between kinetic modelling and constraints-based modelling. The approach allows a connection to be made between network structure and dynamic flux state without a priori knowledge of enzyme kinetic parameters. The key concept of the method is that many aspects of network behaviour (such as the dynamic response to perturbations and the stability of a metabolic state) do not require an explicit kinetic model, but instead can be accessed using a local linear approximation of the system at a given state. Essentially the approach allows an exploration of the dynamic capability of the system. Although this does not necessarily define the actual dynamic behaviour of the system, it does allow novel aspect of dynamic behaviour of metabolic networks to be uncovered. For example, an analysis of the tricarboxylic acid cycle revealed that feedback mechanisms present in the network can induce unstable flux behaviour, and the probability of a transition to an unstable state increases with increased strength of feedback [67].

4. MEASURING NETWORK BEHAVIOUR

Although the construction of mathematical models can be seen as of primary importance in improving our understanding of metabolic network behaviour, that is not to say that direct experimental measurements of metabolism are of diminished importance. In fact, the experimental information is more important than ever, and complementary experimental efforts will be vital to the progress of the computational effort. In particular, all types of modelling described require highly specific experimental inputs. In addition, a model is only useful if it successfully recapitulates the behaviour of the system, and one of the key future uses of molecular profiling datasets will be in model validation. One cannot assess the reliability of a mathematical description of a metabolic network unless one has an empirically determined analysis of network behaviour to compare it against! And even before this step, many important aspects of metabolic network behaviour and regulation can be inferred by analysis of molecular correlations in large-scale datasets.

Recent advances in analytical technologies have brought us to the point where a complete description of the metabolic network at the level of its molecular components can be contemplated. Although the 'omic analytical platforms individually probe different levels of the metabolic network, it is apparent that an integrated view spanning the multiple levels of the cellular control hierarchy will be most informative. However, to date, most studies have concentrated at the level of a single type of molecular entity with a vast number of transcript or gene networks described in the literature (see, for example, [6870]), as well as an increasing number of protein (see, for example, [7173]) and metabolite networks (see, for example, [7476]). In this section we describe how these global molecular profiling experiments have contributed to our understanding of metabolism. We outline the advantages and limitations of the different approaches and will make particular reference to studies that address metabolism at the network level.

4.1 Transcriptomic investigations of the metabolic network

Of all the single-entity networks, transcriptomic networks are probably the most comprehensive, reflecting the status of transcriptomics as a mature technology, in contrast with proteomics and metabolomics [77]. The comprehensive nature of transcriptomic studies means that information concerning all nodes of the network can be accessed. Another useful aspect of transcriptomic information is that it allows conserved upstream regulatory sequences to be uncovered that define the genetic response to environmental and developmental cues. Transcript networks are largely analysed on the basis of co-response of different transcripts. Such analyses allow the extent to which metabolic pathways are co-ordinately transcriptionally regulated to be investigated. They are also useful in defining the function of unknown genes using the ‘guilt by association’ principle. To date, novel gene annotations for enzymes of cell wall [78], brassinosteroid [79], flavonoid [80] and anthocyanin [81] metabolism have been achieved via this strategy. On a more global level, this approach can be used reveal the targets for transcription factors and other regulatory elements. For example, overexpression of the PAP1 (purple acid phosphatase 1) transcription factor in Arabidopsis combined with broad metabolite profiling allowed the recognition of novel targets of this transcription factor [82]. Similarly, an analysis of a cultivated tomato (Solanum lycopersicum L.) population containing introgressions from the wild species S. pennellii revealed an up-regulation in respiratory gene expression (which coincided with increases in the metabolic intermediates of this pathway [83,84]).

The examples discussed thus far concentrate on only small subsets within transcriptional networks and do not really explore the behaviour of the network as a whole. However, four recent articles do make explicit links between transcriptome datasets and metabolic-network behaviour. The first of these to be published, that by Wei et al. [70], analysed the co-expression network of 1330 genes from AraCyc (a biochemical-pathway database for Arabidopsis available at http://www.arabidopsis.org/biocyc/index.jsp) and revealed that genes associated with the same metabolic pathway are, on average, more highly expressed than those associated with different pathways. Perhaps more importantly, however, this study also revealed that the distribution of co-expression was highly skewed, with very few genes having numerous co-expression partners, but most having fewer than ten. Intriguingly, genes with multiple connections tend to be single-copy, whereas genes with multiple paralogues tend to exhibit co-expression with a relatively small number of genes. Wei et al. [70] argue that this suggests that the network expands through gene duplication, but that co-expression links involving duplicate nodes are subsequently weakened. Interestingly, a similar argument has been made for protein interaction networks (see subsequent discussion). Looking at gene network regulation from a slightly different perspective, Walther and co-workers [85] recently analysed transcriptional response diversity in Arabidopsis and quantified the influence that structural genomic properties had on this parameter. For this purpose they used the AtGenExpress transcriptomic compendium (http://www.uni-tuebingen.de/plantphys/AFGN/atgenex.htm) and known cis-element motifs mapped to upstream gene promoter regions and compared the breadth of response with the architectural properties of the genome. The most interesting finding to emerge from this analysis was that greater gene expression regulatory complexity appears to be encoded by an increased density of cis-regulatory elements and thus provide additional evidence for evolutionary adaptation of the regulatory code to that provided previously. Taking a similar approach, but one directly focused towards metabolism, Gutiérrez et al. [68] recently produced a list of what they define as C–N (carbon–nitrogen) responsive machines. They achieved this by exploring global gene expression responses in plants exposed transiently to a matrix of C and N treatments This work revealed that about half of the transcriptome is regulated by C, N or C–N interactions, with detailed evaluation of the 2620 genes interconnected across the experiment suggestive of protein complexes or highly connected signalling or metabolic network such as protein synthesis and degradation, chromatin assembly, RNA metabolism, transport, actin-cytoskeleton formation and many aspects of metabolism, including pathways of glycolysis and the oxidative pentose phosphate pathway. In addition, the authors hint at involvements of auxin signalling and, potentially, mitochondrial RNA interactions. Additional directed experiments are required to substantiate these claims. Finally, a proof-of-concept analysis of the differences in global gene expression within a recombinant inbred line population of A. thaliana allowed the identification of known key elements of the transition to flowering, as well as highlighting some previously unidentified genes that may be involved in this process [87]. It would seem likely that this procedure could be applied to metabolic regulation.

4.2 Protein–protein interaction networks

It has become increasingly apparent that many proteins do not exist independently in the cell, but rather interact specifically with other proteins. Comprehensive protein–protein interaction maps have been established for yeast and E. coli, revealing not only that is protein–protein interaction the norm rather than the exception, but also that the interactions are highly dynamic and condition-specific, implying a regulatory role [88]. In plants, far fewer such studies have been carried out, and analysis of plant protein–protein interactions is at a preliminary stage [89]. However, several large-scale yeast two-hybrid projects have been initiated (see, for example, [71]) and rapid progress can be anticipated. Thus far only ‘interologues’ have been mapped [72,73], that is to say, interactions predicted on the basis of sequence similarity to those proteins demonstrated to interact in other species. However, on a smaller scale, focusing on central regulators of plant meristematic function and leaf development, the study of a complex network of protein interactions led to the identification of a novel family of proteins involved in this process [90]. In addition, large-scale proteomic investigations are also likely to reveal novel protein–protein associations, as demonstrated by several small-scale analysis of several corners of plant metabolism. The developments both at the global level in microbial and mammalian fields [91,92], and at the level of specific developmental process in plants (described above) suggest that these techniques will be powerful tools in understanding the regulation of plant metabolism in the future.

A recent review article [93] raises several salient points concerning both the quality and utility of current protein–protein interaction data. They list the following as vital approaches: the two-hybrid system, affinity purification coupled with MS, the curation of literature reports and the effort to critically assess confidence levels for each technique [9397]. They state that, although both yeast two-hybrid and affinity-purification approaches clearly have their pitfalls, the yeast-hybrid system being prone to returning non-physiological artefacts and the affinity-purification method yielding little information on the exact details of binary interactions, useful data for incorporation into a whole-cell model of protein proximity can be gleaned from their application. This is illustrated both by the work of the Goodsell [94] and of Takamori and co-workers [96]. The former researcher has taken an integrative approach to produce a cellular model of E. coli, blood cells and HIV-infected blood cells [94], whereas the latter researchers recently published a structural model of a synaptic vessel [96]. Again, work of this type in plants would be highly instructive in terms of providing a better understanding of metabolism and processes of metabolite transport. However, to quote from Betts and Russell [91], “the true picture of the cell does not fit well into the blueprint or wiring diagram analogies that are so common in text book biology”. They suggest the adaptation of other analogies, such as the recently defined models of crowd behaviour of Helbing and co-workers [98]. These analogies necessitate the consideration of an important aspect of metabolism (and biology in general) that we have not yet tackled, namely that of cellular dynamism. Interestingly, analysis of the evolution of complex protein networks suggests that this appears to be under control of mechanisms similar to those described for gene networks [99,100]. A two-step model employing gene duplication and rewiring has been employed to explain the evolution of protein interaction networks and their topology (Figure 3; [99,101]). Recent studies suggest that duplication and divergence shape network architecture both at the level of single interactions as well as at the higher order of network motifs and modules [102,103], The high degree of gene duplication in Arabidopsis, and other plant species, as well as high rates of divergence of paralogue function (which has been demonstrated to be highly important in the evolution of plant secondary metabolism [104]), suggest that we will be able to address many interesting questions once such information is available for plants.

Evolution of networks

Figure 3
Evolution of networks

This schematic diagram is based on a protein interaction network, but could be equally representative of a metabolite network. Gene duplication followed by divergence resulting in gain and loss of interactions are key processes that shape interaction networks. This Figure is reproduced from [191] with kind permission from Springer Science and Business Media © 2006.

Figure 3
Evolution of networks

This schematic diagram is based on a protein interaction network, but could be equally representative of a metabolite network. Gene duplication followed by divergence resulting in gain and loss of interactions are key processes that shape interaction networks. This Figure is reproduced from [191] with kind permission from Springer Science and Business Media © 2006.

Another aspect of protein state that has relevance for metabolic regulation is their interaction with other proteins at the level of subunit associations, oligomerization, supercomplexes and metabolons. The association of proteins as a method of metabolic regulation has been long studied, with many of the best studied plant enzymes being active only as a complex of subunits [105107]. Protein purification and reconstitutions studies have, over the last few decades, shown that enzymes across a wide breadth of metabolic pathways exhibit this phenomena, as do several of the transport mechanisms of plants [108,109]. Recently an elegant study revealed that the soluble cytosolic C-terminus of an oligomeric ammonium transporter from Arabidopsis serves as an allosteric regulator that is essential for function with phosphorylation of sites in the C-terminus proposed as the cognate mechanism [108]. This feature of regulation also extends to well-known multienzyme complexes such as the pyruvate dehydrogenase complex [107] and the functional organization of the chloroplastic and mitochondrial electron-transport chains [110,111].

In addition to interactions of proteins within single-enzyme complexes, it is becoming increasingly apparent that the formation of multienzyme complexes is widespread and is likely to have regulatory consequences. For example, it is now well-established that the electron-transport-chain complexes in the mitochondrial inner membrane associate together in supercomplexes [111,112]. Nevertheless, the functional significance and regulation of such supercomplexes remains an open question. Although the close spatial proximity of proteins of associated function would have clear benefit in terms of efficiency, and may facilitate direct enzyme-to-enzyme substrate channelling, this is something that needs to be determined on a case-by-case basis. There is now a growing number of examples in which this phenomenon, often referred to as ‘metabolon formation’, has been demonstrated for proteins with metabolic function. These include the Calvin cycle [113], biosynthesis of the cyanogenic glycoside dhurrin [114], flavonoid pathways [115], and phenylpropanoid [116] and polyamine metabolism [117]. Since the presence and function of metabolons in plants have been described in detail previously [118,119], we will not dwell on this aspect of regulation here. Suffice it to say that although the majority of demonstrations of metabolons in plants occur in secondary metabolism, the recent findings that the enzymes of glycolysis functionally associate with Arabidopsis mitochondria [120] and structural aspects of the crystal structure of sucrose phosphatase [121] provide some support for their operation in plant primary metabolism as well. However, far more evidence is required to demonstrate substrate channelling is indeed occurring in these pathways. Given the technical challenge of measuring substrate channelling [122], it may be some time before we have a truly comprehensive picture of the extent and significance of enzyme–enzyme interactions in plant metabolism.

4.3 Metabolomic investigations of the metabolic network

Since the metabolic network is defined by metabolite–metabolite connections, analysis of the change in metabolite concentrations across the network is a potentially powerful method to gain insight into metabolic organization and regulation [75,76]. The vast majority of such metabolomic network studies are based on an analysis of linear-abundance correlations between metabolite pairs [123]. Given that current metabolomics methods do not provide a comprehensive analysis of plant metabolites [77,124,125], the networks obtained from these studies are illustrative at best. That said, some interesting findings have been made, and it seems likely that recent (and future) technological [126] and bioinformatic [127] advances will improve this situation. Furthermore, when used to study mixed populations exhibiting considerable natural variance, correlation analysis reveals which pathways appear to be co-ordinately regulated and which not. One recent example of this is that amino acid levels were found to be very highly co-ordinated in a tomato population in which the wild species S. pennellii was introgressed into the cultivated variety [84]. However, and interestingly, the content of phenylalanine-derived volatiles in this population showed little correlation with their precursor [129]. The first finding is in keeping with previous observations of concerted regulation of levels of amino acids [130133], although subtle differences are apparent between species and physiological circumstances. Indeed, as was detailed in previous reviews [134,135], the study of plant amino acid metabolism provides a clear illustration of the complexity of regulation exhibited within modules of the metabolic network.

To date, these networks have largely been analysed on the basis of the simple correlation of metabolite pairs. Perhaps the most striking feature of these studies was that a small number of metabolite pairs displayed a remarkably high correlation among biological replicates, even though the large majority of metabolite pairs showed little or no correlation [123]. This initial finding has been generalized by a number of subsequent studies [76,136,137]. Detailed evaluation of this phenomenon has suggested that metabolite correlations are not necessarily related to proximity in the biochemical network [74,138]. These analyses described four different regulatory configurations that are expected to be the origin of metabolite correlation. Simulations suggest that when the correlations are very strong, they are most likely due to chemical equilibrium. An interesting prediction, still to be confirmed, is that metabolites sharing conserved moieties (i.e. a common fragment of the metabolites that turns over more slowly than the metabolites themselves, such as the adenine of the adenine nucleotides), should have high correlations, and at least one of them being negatively correlated with the others [74]. Most high correlations in these networks are, however, thought to be due to stronger mutual control by a single enzyme or a much higher degree of variation of a single enzyme activity in respect to the others. Although analysis of the set of metabolites forming correlation cliques [139] can provide a good hint as to the identity of the enzyme in question, it seems likely that parallel analysis of protein or enzyme activities is likely to be required to fully exploit this approach in the identification of novel aspects of metabolic regulation.

Analysis of non-linear correlations, although challenging, is also potentially a rich source of such information. In the initial approach described above, Roessner et al. [123] documented hyperbolic relationships between the minority of metabolite pairs, including lysine and isoleucine, that had previously been demonstrated in another species to be interrelated by feedforward/feedback-control circuitry [140]. Perhaps surprisingly, this approach has not been much followed in the interim, perhaps largely due to a lack of similar behaviour between other known metabolites in GC–MS chromatograms. However, given the fact that technological developments will shortly allow us to measure a far greater number of metabolites, it seems likely that it may find greater utility in the future. Thus far we have concentrated this discussion on individual elements rather than the networks themselves. The study of Weckwerth et al. [76], although lacking a robust sensitivity analysis, suggests that analysing the networks of genetically modified plants may reveal differences that were not detected directly at the level of the metabolite. Comparison of network changes in various cellular circumstances is likely to be of increasing importance once a clear framework has been established that allows a high confidence of analysis. Ultimately this will also include the dynamic analysis of rewiring of the metabolic networks which to date has largely been carried out by describing changes in individual metabolite levels [141] although network changes have started to be reported [142].

4.4 Integration of 'omic data

The hierarchical nature of metabolic regulation highlights the fact that the previously described examples, although informative, are somewhat artificial, since transcripts, proteins or metabolites do not act independently of one another. With this in mind, several studies have attempted to integrate data from different analytical platforms with a view to establishing a more complete assessment of the metabolic network. Several of these studies are particularly worthy of mention. The studies of Hirai et al. [143] and Nikiforova and co-workers [144] both look at transcript–metabolite networks during sulphur stress, although the datasets they obtain were analysed from different perspectives. That of Hirai et al. [143] was evaluated with respect to assigning gene function on the basis of correlations between metabolite and transcript and hence of understanding metabolic regulation at the level of the constitutive pathway. By contrast, that of Nikiforova et al. [144] was more focused towards understanding the systemic response to sulfur stress and elucidating the informational processing networks by which this is achieved [145]. The work of Weckwerth and co-workers [146] extended this network approach by correlating both proteins and metabolites during diurnal cycles in mutant and wild-type Arabidopsis plants. Their study revealed that the combined approach resulted in a considerable improvement in discriminatory power over independent evaluation of the datasets. As a final example, in our own work, we followed tomato fruit development by monitoring changes in metabolites, transcripts and fluxes and evaluated the correlative behaviour between these factors [147]. That study revealed several interesting features of metabolic change across the ripening period, with much concerted regulation apparent at the pathway level and even between pathways. Generating insight from such datasets into how metabolic regulation is exerted remains rather challenging. For example, the fact that key ripening genes correlated with many sugars and organic acids implies an important role for transcriptional regulation of sugar and organic acid metabolism. But ultimately, as with any correlation study, follow-up experimentation is required to prove a causal link between the correlating elements.

Although not directly evaluating networks per se, the work of the Stitt group is also worth discussing here. In a series of multilevel profiling experiments in Arabidopsis, the effect of extended darkness, nutrient deficiency, and even global climate change, were studied, allowing both the confirmation of long-standing hypotheses and novel insight into the regulation of metabolism under these conditions [148]. These studies facilitated a detailed insight into hierarchical control in the regulation of metabolism of the illuminated leaf.

5. MEASURING METABOLIC FLUX

Although much can be gained from observation of changes in metabolic transcripts and proteins, ultimately the true behaviour of the metabolic network can only be gauged by direct measurement of flux. Even changes in concentration of metabolites are open to misinterpretation; it is entirely possible for metabolic pathway flux to change dramatically without perturbation of the concentration of pathway intermediates [15,149], and metabolites can change in the opposite direction to overall pathway flux. A precise knowledge of pathway flux is therefore essential to characterize the extent of metabolic change when plant cells face different environmental conditions or physiological demands. In addition, quantification of flux is an essential part of defining the control distribution within the metabolic network [150]. Finally, empirically determined flux values are vital data with which to compare and validate predicted flux values from mathematical models.

In its simplest form, pathway flux can be measured from the rate of accumulation of an end-product of the pathway. However, there are few true end-products and even relatively stable biopolymers, such as starch and cellulose, are still subject to a synthesis–degradation cycle. Therefore a more direct measure of movement of carbon or nitrogen through metabolic pathways is required. This is usually achieved by measuring the redistribution of label when an isotopically labelled precursor is fed to plant tissues or organs. Flux can be determined from such redistributions in two ways: either the rate of labelling of metabolite pools is considered or the pattern of labelling at isotopic steady state is analysed. The latter approach allows determination of relative fluxes on the basis that carbon arriving at a metabolite pool from different metabolic pathway branches will have a different fractional enrichment of label at each mapping atom of the precursor molecule [151]. Improvements in analytical methods [152,153] and computational developments (see, for example, [154]) have improved the scope of the steady-state approach, and increasingly complex metabolic flux networks are being analysed. There is an established literature on this subject, and several excellent reviews have recently been published that cover both the theoretical and practical aspects of flux measurement and review published flux maps [37,155158]. We do not intend to reiterate the content of these reviews here, but rather wish to focus on aspects of flux analysis that still require further development.

5.1 Subcellular compartmentation and flux

Perhaps the biggest challenge in the determination of fluxes in plant metabolic networks is the extensive subcellular compartmentation of metabolism and the existence of parallel metabolic reactions in more than one compartment [159]. In the steady-state approach, subcellular fluxes are unravelled from the labelling of end-products synthesized uniquely in a single subcellular compartment. Thus the labelling of the glucose moiety of starch and sucrose reflect the labelling of plastidic and cytosolic hexose phosphates respectively [155]. The problem is that the redundant nature of plant metabolism means that there are precious few such metabolites that accurately provide compartment-specific information. However, there are many unexplored possibilities, and progress on this front is still possible [160]. Moreover, in certain cases it is possible to simplify the compartmentation issue, because rapid exchange of some metabolites between compartments effectively renders the compartmentalized metabolite a single homogenous pool [158].

Failure to properly consider subcellular compartmentation of metabolites can lead to substantial misinterpretations of labelling patterns. A good example which highlights this issue is the analysis of sucrose cycling. In many heterotrophic tissues, sucrose is continuously degraded and resynthesized, creating an ATP-consuming futile cycle. This process has been accurately quantified by comparing the rate of sucrose synthesis (determined from short-term [14C]glucose experiments) with the rate of sucrose accumulation, and typically sucrose cycling accounts for approx. 5% of cellular ATP turnover [161]. However, various estimates of the same phenomenon using a steady-state 13C approach have reported much higher values of between 43 and 82% [162,163], leading to the proposal of a novel glucose phosphatase to account for the high rate of glucose resynthesis [164]. However, the validity of these high rates of sucrose cycling and glucose resynthesis have been recently questioned [165]. Instead, it is argued that the result is an artefact based on a failure to account for changes of labelling of glucose as a result of passage through the vacuole. Indeed the vacuolar compartment is rarely, if ever, considered in metabolic flux maps, but may introduce labelling variations to many primary metabolites.

Although this example highlights the importance of subcellular compartmentation in the interpretation of labelling patterns, it is not immediately obvious how we might make significant advances in our ability to account for it. Although it is possible to directly analyse metabolites from different subcellular compartments by non-aqueous fractionation [166] or by rapid fractionation of protoplasts through silicone oil [167], these methods remain rather specialist and have yet to be applied to analysis of stable isotope labelling. Currently there are practical limitations to both approaches. The non-aqueous method gives rather poor separation of different subcellular compartments, and some compartments, such as the mitochondrion, remain out of reach. By contrast, the silicone-oil method can give good separation of organelles such as plastids and mitochondria, but only works with protoplasts, entities that may have a metabolic state considerably different from that of intact cells. An alternative approach is to exploit the chemical shift generated by substantial pH differences between compartments such as the cytosol and vacuole to analyse subcellular metabolite labelling patterns by in vivo NMR. Typically this method is combined with 31P-NMR to give a measure of vacuolar and cytosolic pH [168], but 13C labelling patterns of primary metabolites can also be analysed [169]. It is likely that a combination of refinements to the methodologies of subcellular fractionation techniques and in vivo NMR approaches will be necessary to give a more complete picture of the extent to which compartmentation leads to multiple metabolite pools that are differentially labelled.

The problem of compartmentation has troubled plant biochemists for many years [170], and the techniques to analyse metabolic compartmentation that we have described are not new. The question therefore arises as to whether new molecular techniques have emerged that might assist with the compartmentation problem. A promising advance in this respect is the development of genetically encoded metabolite sensors – bacterial periplasmic binding proteins modified with cyan and yellow fluorescent proteins such that metabolite binding leads to FRET (fluorescence resonance energy transfer) [171,172]. In principle, such sensors can be targeted to different subcellular compartments and would permit in vivo quantification of subcellular metabolite concentration. Recently it has been suggested that such sensors might be useful in the context of flux analysis [173]. Because the sensors report only on metabolite concentration and not on labelling state, this use would be restricted to analysis of dynamic metabolic changes in which perturbations of metabolite pool sizes are used to reconstruct fluxes. Although it is tempting to marry the use of metabolite sensors to such an approach, there are many practical considerations to be overcome, not least of which is that the basis of such flux reconstructions is the related changes in linked metabolites. Because of the extent of biological variation in metabolite concentration, it is likely that accurate flux reconstruction will only be possible if dynamic changes in metabolites are measured in the same sample. Given that the FRET metabolite sensors are genetically encoded, it is unlikely that more than three or four sensors for different metabolites could be introduced into the same transgenic plant line, and therefore the ability of these sensors to report on multiple metabolites in the same sample is rather limited.

5.2 Measuring flux dynamics

The established method of measuring multiple metabolic fluxes depends on the system reaching isotopic and metabolic steady state. However, it is increasingly being recognized that the dynamics of metabolism are perhaps more informative than the steady state [67]. Not only are complex dynamic behaviours a feature of metabolism (e.g. the glycolytic oscillations of yeast [174] and the diurnal regulation of the plant Calvin cycle), but also the regulation of metabolic flux is based on the dynamic properties of the system. Transient regulatory interactions (protein–protein [88], protein–metabolite and protein–lipid) and feedback loops are essential features of a regulatory system that generates a homoeostatic state while at the same time allows rapid adaptation and response to altered physiological demands [134]. It therefore follows that an analysis of network flux dynamics is necessary to probe these regulatory features. In addition, an isotopic and metabolic steady state is often not experimentally achievable or physiologically relevant [149].

Attention is therefore turning to the analysis of dynamic flux patterns in biological systems. The problem is an inherently more complex one than the solution of flux at steady state, because change in metabolite pool sizes, as well as labelling patterns, need to be considered. Moreover, a system of non-linear differential equations of high dimension is needed to mathematically describe the system. Fitting the parameters of this high-dimension equation set to the experimental data is a computationally challenging task. Quantifying dynamic fluxes is also experimentally more intensive than at steady state, because a time series of samples must be analysed. Moreover, the estimation of dynamic fluxes generally requires information on the labelling and pool size of a greater number of metabolites than at steady state, because flux-balancing no longer applies. In general, the larger the network under consideration, the more challenging the computational and experimental problem, which is why most studies of dynamic flux changes have focused on relatively short linear pathways of plant secondary metabolism [40,176,177]. Experimental biochemists have sought pragmatic solutions to the matching of available labelling and pool-size data from standard GC–MS analyses with possible flux solutions [178,179]. Such GC–MS-based metabolite profiling focuses mainly on primary metabolism, where extensive branching and multiple compartmented pools can be confounding factors in analysing flux. Nevertheless, in certain areas of the primary metabolic network, it is possible to estimate fluxes [178].

Although we are clearly some way off being able to estimate non-steady state fluxes in large networks, recent computational developments have paved the way for analysing large-scale networks at metabolic – but not isotopic – steady state. This half-way state is particularly relevant for experimental systems where isotopic steady state is not likely to be reached [179] or where the labelling patterns at isotopic steady state hold no information about flux. A key example of the latter is when labelled CO2 is supplied to plants. Because only one carbon atom is labelled and all other carbon atoms in photoautotrophically synthesized molecules are ultimately derived from that carbon then at isotopic steady state, all carbon atoms in the system will be labelled to the same fractional enrichment as the supplied CO2, irrespective of metabolic flux. However, the rate of labelling of different compounds in the dynamic labelling phase is related to flux. There is much interest in using CO2 as a label source, because it allows labelling experiments to be done with intact plants rather than being restricted to cell-suspension cultures or tissues or organs in cultures [155], and recently experimental and technical assessments and preliminary modelling have been carried out [180,181]. Major steps forward in deriving network fluxes in such situations have been made by Wiechert and colleagues [182,183]. The experimental and data requirements have been assessed [182] and, importantly, freely available computational software has been devised that can estimate flux [183]. The software is essentially an ordinary differential-equation solver that can handle the very-high-dimension equation sets specified by the dynamic labelling state and estimate flux by fitting parameters in the equations to the experimental data by non-linear regression. The method has been demonstrated with an estimation of fluxes in central carbon metabolism of E. coli supplied with [13C]glucose and sampled over a subsequent 16 s time period [184].

6. PERSPECTIVES

The recent flood of post-genomic molecular profiling data has encouraged us to tackle previously untenable organizational structures, and we now find ourselves standing in the foothills of a systems-biology mountain. The challenges we face to bring together all the experimental data and the computational models are considerable. In particular, we have highlighted the considerable gaps in the experimental datasets in relation to providing model parameters. There are also, as ever, numerous technical challenges to overcome if we are to refine further our analytical ability to cope with issues such as subcellular compartmentation and the behaviour of different cell types. Although it may be the case that truly predictive and validated models of the Arabidopsis metabolic network remain some way off, it is clear that the systems-level analysis of molecular change is revealing new aspects of metabolic network behaviour and regulation. For example correlation studies have proven highly effective in uncovering novel gene function. The present review has highlighted the relative paucity of global network studies at the protein and metabolite levels, and this is an obvious area were technological developments are likely to facilitate improved understanding of network behaviour. Although technological advances can be anticipated that will improve further the quality of data that we can access, we should not play down the tremendous progress that has been made in the understanding of metabolic networks in the last few years. One field where this is particularly noticeable is that of flux analysis, and there are further exciting prospects for additional improvement. For example, recently developed tools for metabolic profiling [185] have the potential to dramatically enhance our understanding of metabolic dynamics. The field of metabolic network biology is very much at the crossroads of contemporary experimental and theoretical biology, with its development dependent on effective collaboration of practitioners in both fields. Reassuringly, despite the manifold technical obstacles and the sheer complexity of the problem at hand, it is apparent that there is much interest in unravelling plant metabolic networks, and we can look forward to continued rapid progress in years to come.

L. J. S. (grants E20443 and BBE0023231) and D. A. F. (grants CFB17702 and BBE00203X1) acknowledge support from the Biotechnology and Biological Sciences Research Council, and A. R. F. acknowledges support from the Deutsche Forschungsgemeinschaft, for their research in this area.

Abbreviations

     
  • FRET

    fluorescence resonance energy transfer

  •  
  • ID

    identification

  •  
  • KEGG

    Kyoto Encyclopedia of Genes and Genomes

  •  
  • NADME

    NAD+-dependent malic enzyme

References

References
1
Horn
M. E.
Woodard
S. L.
Howard
J. A.
Plant molecular farming: systems and products
Plant Cell Rep.
2004
, vol. 
22
 (pg. 
711
-
720
)
2
Kinney
A. J.
Metabolic engineering in plants for human health and nutrition
Curr. Opin. Biotechnol.
2006
, vol. 
17
 (pg. 
130
-
138
)
3
Aharoni
A.
Jongsma
M. A.
Bouwmeester
H. J.
Volatile science? Metabolic engineering of terpenoids in plants
Trends Plant Sci.
2005
, vol. 
10
 (pg. 
594
-
602
)
4
Broun
P.
Transcription factors as tools for metabolic engineering in plants
Curr. Opin. Plant Biol.
2004
, vol. 
7
 (pg. 
202
-
209
)
5
Hanson
A.
Shanks
J.
Plant metabolic engineering – entering the S curve
Metab. Eng.
2002
, vol. 
4
 (pg. 
1
-
2
)
6
Oksman-Caldentey
K.-M.
Saito
K.
Integrating genomics and metabolomics for engineering plant metabolic pathways
Curr. Opin. Biotechnol.
2005
, vol. 
16
 (pg. 
1
-
6
)
7
Papin
J. A.
Price
N. D.
Wiback
S. J.
Fell
D. A.
Palsson
B. O.
Metabolic pathways in the post-genome era
Trends Biochem. Sci.
2003
, vol. 
28
 (pg. 
250
-
258
)
8
Stark
D.
Timmerman
K.
Barry
G.
Preiss
J.
Kishore
G.
Regulation of the amount of starch in plant tissues by ADPglucose pyrophosphorylase
Science
1992
, vol. 
258
 (pg. 
287
-
292
)
9
Sweetlove
L. J.
Burrell
M. M.
apRees
T.
Starch metabolism in tubers of transgenic potato (Solanum tuberosum) with increased ADP-glucose pyrophosphorylase
Biochem. J.
1996
, vol. 
320
 (pg. 
493
-
498
)
10
Lytovchenko
A.
Sonnewald
U.
Fernie
A. R.
The complex network of non-cellulosic carbohydrate metabolism
Curr. Opin. Plant Biol.
2007
, vol. 
10
 (pg. 
227
-
235
)
11
Heinrich
R.
Rapoport
T. A.
A linear steady-state treatment of enzymatic chains. General properties, control and effector strength
Eur. J. Biochem.
1974
, vol. 
42
 (pg. 
89
-
95
)
12
Kacser
H.
Burns
J. A.
The control of flux
Symp. Soc. Exp. Biol.
1973
, vol. 
27
 (pg. 
65
-
104
)
13
Kruckeberg
A. L.
Neuhaus
H. E.
Feil
R.
Gottlieb
L. D.
Stitt
M.
Decreased-activity mutants of phosphoglucose isomerase in the cytosol and chloroplast of Clarkia xantiana. Impact on mass-action ratios and fluxes to sucrose and starch, and estimation of flux control coefficients and elasticity coefficients
Biochem J.
1989
, vol. 
261
 (pg. 
457
-
467
)
14
Sweetlove
L. J.
Muller-Rober
B.
Willmitzer
L.
Hill
S. A.
The contribution of adenosine 5′-diphosphoglucose pyrophosphorylase to the control of starch synthesis in potato tubers
Planta
1999
, vol. 
209
 (pg. 
330
-
337
)
15
Fell
D. A.
Enzymes, metabolites and fluxes
J. Exp. Bot.
2005
, vol. 
56
 (pg. 
267
-
272
)
16
Schwender
J.
Shachar-Hill
Y.
Ohlrogge
J. B.
Mitochondrial metabolism in developing embryos of Brassica napus
J. Biol. Chem.
2006
, vol. 
281
 (pg. 
34040
-
34047
)
17
Wagner
A.
Fell
D. A.
The small world inside large metabolic networks
Proc. R. Soc. London Ser. B
2001
, vol. 
268
 (pg. 
1803
-
1810
)
18
Baxter
C. J.
Redestig
H.
Schauer
N.
Repsilber
D.
Patil
K. R.
Nielsen
J.
Selbig
J.
Liu
J.
Fernie
A. R.
Sweetlove
L. J.
The metabolic response of heterotrophic Arabidopsis cells to oxidative stress
Plant Physiol.
2007
, vol. 
143
 (pg. 
312
-
325
)
19
Bohnert
H. J.
Sheveleva
E.
Plant stress adaptations – making metabolism move
Curr. Opin. Plant Biol.
1998
, vol. 
1
 (pg. 
267
-
274
)
20
Faurobert
M.
Mihr
C.
Bertin
N.
Pawlowski
T.
Negroni
L.
Sommerer
N.
Causse
M.
Major proteome variations associated with cherry tomato pericarp development and ripening
Plant Physiol.
2007
, vol. 
143
 (pg. 
1327
-
1346
)
21
Zhang
P.
Foerster
H.
Tissier
C. P.
Mueller
L.
Paley
S.
Karp
P. D.
Rhee
S. Y.
MetaCyc and AraCyc. Metabolic pathway databases for plant research
Plant Physiol.
2005
, vol. 
138
 (pg. 
27
-
37
)
22
Sweetlove
L. J.
Last
R. L.
Fernie
A. R.
Predictive metabolic engineering: a goal for systems biology
Plant Physiol.
2003
, vol. 
132
 (pg. 
420
-
425
)
23
Kanehisa
M.
Goto
S.
Hattori
M.
Aoki-Kinoshita
K. F.
Itoh
M.
Kawashima
S.
Katayama
T.
Araki
M.
Hirakawa
M.
From genomics to chemical genomics: new developments in KEGG
Nucleic Acids Res.
2006
, vol. 
34
 (pg. 
D354
-
D357
)
24
Schomburg
I.
Chang
A.
Ebeling
C.
Gremse
M.
Heldt
C.
Huhn
G.
Schomburg
D.
BRENDA, the enzyme database: updates and major new developments
Nucleic Acids Res.
2004
, vol. 
32
 (pg. 
D431
-
D433
)
25
Borodina
I.
Krabben
P.
Nielsen
J.
Genome-scale analysis of Streptomyces coelicolor A3(2) metabolism
Genome Res.
2005
, vol. 
15
 (pg. 
820
-
829
)
26
Edwards
J. S.
Ibarra
R. U.
Palsson
B. O.
In silico predictions of Escherichia coli metabolic capabilities are consistent with experimental data
Nat. Biotechnol.
2001
, vol. 
19
 (pg. 
125
-
130
)
27
Forster
J.
Famili
I.
Fu
P.
Palsson
B. B.
Nielsen
J.
Genome-scale reconstruction of the Saccharomyces cerevisiae metabolic network
Genome Res.
2003
, vol. 
13
 (pg. 
244
-
253
)
28
Price
N. D.
Papin
J. A.
Schilling
C. H.
Palsson
B. O.
Genome-scale microbial in silico models: the constraints-based approach
Trends Biotechnol.
2003
, vol. 
21
 (pg. 
162
-
169
)
29
Karp
P. D.
Ouzounis
C. A.
Moore-Kochlacs
C.
Goldovsky
L.
Kaipa
P.
Ahren
D.
Tsoka
S.
Darzentas
N.
Kunin
V.
Lopez-Bigas
N.
Expansion of the BioCyc collection of pathway/genome databases to 160 genomes
Nucleic Acids Res.
2005
, vol. 
33
 (pg. 
6083
-
6089
)
30
Poolman
M. G.
Bonde
B. K.
Gevorgyan
A.
Patel
H. H.
Fell
D. A.
Challenges to be faced in the reconstruction of metabolic networks from public databases
IEEC Proc. Syst. Biol.
2006
, vol. 
153
 (pg. 
379
-
384
)
31
Francke
C.
Siezen
R. J.
Teusink
B.
Reconstructing the metabolic network of a bacterium from its genome
Trends Microbiol.
2005
, vol. 
13
 (pg. 
550
-
558
)
32
Snoep
J. L.
The silicon cell initiative: working towards a detailed kinetic description at the cellular level
Curr. Opin. Biotechnol.
2005
, vol. 
16
 (pg. 
336
-
343
)
33
Garfinkel
D.
Hess
B.
Metabolic control mechanisms. vii. A detailed computer model of the glycolytic pathway in ascites cells
J. Biol. Chem.
1964
, vol. 
239
 (pg. 
971
-
983
)
34
Poolman
M. G.
Assmus
H. E.
Fell
D. A.
Applications of metabolic modelling to plant metabolism
J. Exp. Bot.
2004
, vol. 
55
 (pg. 
1177
-
1186
)
35
Kacser
H.
Burns
J. A.
The control of flux
Biochem. Soc. Trans.
1995
, vol. 
23
 (pg. 
341
-
366
)
36
Giersch
C.
Mathematical modelling of metabolism
Curr. Opin. Plant Biol.
2000
, vol. 
3
 (pg. 
249
-
253
)
37
Morgan
J. A.
Rhodes
D.
Mathematical modeling of plant metabolic pathways
Metab. Eng.
2002
, vol. 
4
 (pg. 
80
-
89
)
38
Sims
A. P.
Folkes
B. F.
A kinetic study of the assimilation of [15N]ammonia and the synthesis of amino acids in an exponentially growing culture of Candida utilis
Proc. R. Soc. London Ser. B
1964
, vol. 
159
 (pg. 
479
-
502
)
39
Rizzi
M.
Baltes
M.
Theobald
U.
Reuss
M.
In vivo analysis of metabolic dynamics in Saccharomyces cerevisiae. II. Mathematical model
Biotechnol. Bioeng.
1997
, vol. 
55
 (pg. 
592
-
608
)
40
Boatright
J.
Negre
F.
Chen
X.
Kish
C. M.
Wood
B.
Peel
G.
Orlova
I.
Gang
D.
Rhodes
D.
Dudareva
N.
Understanding in vivo benzenoid metabolism in petunia petal tissue
Plant Physiol.
2004
, vol. 
135
 (pg. 
1993
-
2011
)
41
McNeil
S. D.
Rhodes
D.
Russell
B. L.
Nuccio
M. L.
Shachar-Hill
Y.
Hanson
A. D.
Metabolic modeling identifies key constraints on an engineered glycine betaine synthesis pathway in tobacco
Plant Physiol.
2000
, vol. 
124
 (pg. 
153
-
162
)
42
Heinzle
E.
Matsuda
F.
Miyagawa
H.
Wakasa
K.
Nishioka
T.
Estimation of metabolic fluxes, expression levels and metabolite dynamics of a secondary metabolic pathway in potato using label pulse-feeding experiments combined with kinetic network modelling and simulation
Plant J.
2007
, vol. 
50
 (pg. 
176
-
187
)
43
McNeil
S. D.
Nuccio
M. L.
Hanson
A. D.
Betaines and related osmoprotectants. Targets for metabolic engineering of stress resistance
Plant Physiol.
1999
, vol. 
120
 (pg. 
945
-
950
)
44
Heijnen
J. J.
Approximative kinetic formats used in metabolic network modeling
Biotechnol. Bioeng.
2005
, vol. 
91
 (pg. 
534
-
545
)
45
Srividhya
J.
Crampin
E. J.
McSharry
P. E.
Schnell
S.
Reconstructing biochemical pathways from time course data
Proteomics
2007
, vol. 
7
 (pg. 
828
-
838
)
46
Varma
A.
Boesch
B. W.
Palsson
B. O.
Stoichiometric interpretation of Escherichia coli glucose catabolism under various oxygenation rates
Appl. Environ. Microbiol.
1993
, vol. 
59
 (pg. 
2465
-
2473
)
47
Stelling
J.
Klamt
S.
Bettenbrock
K.
Schuster
S.
Gilles
E. D.
Metabolic network structure determines key aspects of functionality and regulation
Nature
2002
, vol. 
420
 (pg. 
190
-
193
)
48
Borodina
I.
Nielsen
J.
From genomes to in silico cells via metabolic networks
Curr. Opin. Biotechnol.
2005
, vol. 
16
 (pg. 
350
-
355
)
49
Bonarius
H.
Schmid
G.
Tramper
J.
Flux analysis of underdetermined metabolic networks: the quest for the missing constraints
Trends Biotechnol.
1997
, vol. 
15
 (pg. 
308
-
314
)
50
Edwards
J. S.
Palsson
B. O.
The Escherichia coli MG1655 in silico metabolic genotype: its definition, characteristics, and capabilities
Proc. Natl. Acad. Sci. U.S.A.
2000
, vol. 
97
 (pg. 
5528
-
5533
)
51
Almaas
E.
Oltvai
Z. N.
Barabasi
A. L.
The activity reaction core and plasticity of metabolic networks
PLoS Comput. Biol.
2005
, vol. 
1
 pg. 
e68
 
52
Almaas
E.
Kovacs
B.
Vicsek
T.
Oltvai
Z. N.
Barabasi
A. L.
Global organization of metabolic fluxes in the bacterium Escherichia coli
Nature
2004
, vol. 
427
 (pg. 
839
-
843
)
53
Segre
D.
Vitkup
D.
Church
G. M.
Analysis of optimality in natural and perturbed metabolic networks
Proc. Natl. Acad. Sci. U.S.A.
2002
, vol. 
99
 (pg. 
15112
-
15117
)
54
Covert
M. W.
Knight
E. M.
Reed
J. L.
Herrgard
M. J.
Palsson
B. O.
Integrating high-throughput and computational data elucidates bacterial networks
Nature
2004
, vol. 
429
 (pg. 
92
-
96
)
55
Millar
A. H.
Location, location, location: surveying the intracellular real estate through proteomics in plants
Funct. Plant Biol.
2004
, vol. 
31
 (pg. 
563
-
571
)
56
van Wijk
K. J.
Plastid proteomics
Plant Physiol. Biochem.
2004
, vol. 
42
 (pg. 
963
-
977
)
57
Heazlewood
J. L.
Tonti-Filippini
J.
Verboom
R. E.
Millar
A. H.
Combining experimental and predicted datasets for determination of the subcellular location of proteins in Arabidopsis
Plant Physiol.
2005
, vol. 
139
 (pg. 
598
-
609
)
58
Martinoia
E.
Maeshima
M.
Neuhaus
H. E.
Vacuolar transporters and their essential role in plant metabolism
J. Exp. Bot.
2007
, vol. 
58
 (pg. 
83
-
102
)
59
Picault
N.
Hodges
M.
Palmieri
L.
Palmieri
F.
The growing family of mitochondrial carriers in Arabidopsis
Trends Plant Sci.
2004
, vol. 
9
 (pg. 
139
-
146
)
60
Weber
A. P.
Fischer
K.
Making the connections – the crucial role of metabolite transporters at the interface between chloroplast and cytosol
FEBS Lett.
2007
, vol. 
581
 (pg. 
2215
-
2222
)
61
Kummel
A.
Panke
S.
Heinemann
M.
Systematic assignment of thermodynamic constraints in metabolic network models
BMC Bioinformatics
2006
, vol. 
7
 pg. 
512
 
62
Covert
M. W.
Palsson
B. O.
Transcriptional regulation in constraints-based metabolic models of Escherichia coli.
J. Biol. Chem.
2002
, vol. 
277
 (pg. 
28058
-
28064
)
63
Liebermeister
W.
Klipp
E.
Bringing metabolic networks to life: integration of kinetic, metabolic, and proteomic data
Theor. Biol. Med. Model.
2006
, vol. 
3
 pg. 
42
 
64
Mendes
P.
Camacho
D.
de la Fuente
A.
Modelling and simulation for metabolomics data analysis
Biochem. Soc. Trans.
2005
, vol. 
33
 (pg. 
1427
-
1429
)
65
Steuer
R.
Computational approaches to the topology, stability and dynamics of metabolic networks
Phytochemistry
2007
, vol. 
68
 (pg. 
2139
-
2151
)
66
Steuer
R.
Gross
T.
Selbig
J.
Blasius
B.
Structural kinetic modeling of metabolic networks
Proc. Natl. Acad. Sci. U.S.A.
2006
, vol. 
103
 (pg. 
11868
-
11873
)
67
Steuer
R.
Nunes Nesi
A.
Fernie
A. R.
Gross
T.
Blasius
B.
Selbig
J.
From structure to dynamics of metabolic pathways: application to the plant mitochondrial TCA cycle
Bioinformatics
2007
, vol. 
23
 (pg. 
1378
-
1385
)
68
Gutiérrez
R. A.
Lejay
L. V.
Dean
A.
Chiaromonte
F.
Shasha
D. E.
Coruzzi
G. M.
Qualitative network models and genome-wide expression data define carbon/nitrogen-responsive molecular machines in Arabidopsis
Genome Biol.
2007
, vol. 
8
 pg. 
R7
 
69
Walther
D.
Brunnemann
R.
Selbig
J.
The regulatory code for transcriptional response diversity and its relation to genome structural properties in A. thaliana
Plos Genet.
2007
, vol. 
3
 (pg. 
216
-
229
)
70
Wei
H. R.
Persson
S.
Mehta
T.
Srinivasasainagendra
V.
Chen
L.
Page
G. P.
Somerville
C.
Loraine
A.
Transcriptional coordination of the metabolic network in Arabidopsis
Plant Physiol.
2006
, vol. 
142
 (pg. 
762
-
774
)
71
Kersten
B.
Burkle
L.
Kuhn
E. J.
Giavalisco
P.
Konthur
Z.
Lueking
A.
Walter
G.
Eickhoff
H.
Schneider
U.
Large-scale plant proteomics
Plant Mol. Biol.
2002
, vol. 
48
 (pg. 
133
-
141
)
72
Walhout
A. J. M.
Sordella
R.
Lu
X. W.
Hartley
J. L.
Temple
G. F.
Brasch
M. A.
Thierry-Mieg
N.
Vidal
M.
Protein interaction mapping in C. elegans using proteins involved in vulval development
Science
2000
, vol. 
287
 (pg. 
116
-
122
)
73
Yu
H. Y.
Luscombe
N. M.
Lu
H. X.
Zhu
X. W.
Xia
Y.
Han
J. D. J.
Bertin
N.
Chung
S.
Vidal
M.
Gerstein
M.
Annotation transfer between genomes: protein–protein interologs and protein–DNA regulogs
Genome Res.
2004
, vol. 
14
 (pg. 
1107
-
1118
)
74
Camacho
D.
de la Fuente
A.
Mendes
P.
The origin of correlations in metabolomics data
Metabolomics
2005
, vol. 
1
 (pg. 
53
-
63
)
75
Steuer
R.
Kurths
J.
Fiehn
O.
Weckwerth
W.
Observing and interpreting correlations in metabolomic networks
Bioinformatics
2003
, vol. 
19
 (pg. 
1019
-
1026
)
76
Weckwerth
W.
Loureiro
M. E.
Wenzel
K.
Fiehn
O.
Differential metabolic networks unravel the effects of silent plant phenotypes
Proc. Natl. Acad. Sci. U.S.A.
2004
, vol. 
101
 (pg. 
7809
-
7814
)
77
Fernie
A. R.
Trethewey
R. N.
Krotzky
A. J.
Willmitzer
L.
Metabolite profiling: from diagnostics to systems biology
Nat. Rev. Mol. Cell. Biol.
2004
, vol. 
5
 (pg. 
763
-
769
)
78
Usadel
B.
Kuschinsky
A. M.
Steinhauser
D.
Pauly
M.
Transcriptional co-response analysis as a tool to identify new components of the wall biosynthetic machinery
Plant Biosyst.
2005
, vol. 
139
 (pg. 
69
-
73
)
79
Rautengarten
C.
Steinhauser
D.
Bussis
D.
Stintzi
A.
Schaller
A.
Kopka
J.
Altmann
T.
Inferring hypotheses on functional relationships of genes: analysis of the Arabidopsis thaliana subtilase gene family
Plos Comput. Biol.
2005
, vol. 
1
 (pg. 
297
-
312
)
80
Yonekura-Sakakibara
K.
Tohge
T.
Niida
R.
Saito
K.
Identification of a flavonol 7-O-rhamnosyltransferase gene determining flavonoid pattern in Arabidopsis by transcriptome coexpression analysis and reverse genetics
J. Biol. Chem.
2007
, vol. 
282
 (pg. 
14932
-
14941
)
81
Vanderauwera
S.
Zimmermann
P.
Rombauts
S.
Vandenabeele
S.
Langebartels
C.
Gruissem
W.
Inze
D.
Van Breusegem
F.
Genome-wide analysis of hydrogen peroxide-regulated gene expression in Arabidopsis reveals a high light-induced transcriptional cluster involved in anthocyanin biosynthesis
Plant Physiol.
2005
, vol. 
139
 (pg. 
806
-
821
)
82
Tohge
T.
Nishiyama
Y.
Hirai
M. Y.
Yano
M.
Nakajima
J.
Awazuhara
M.
Inoue
E.
Takahashi
H.
Goodenowe
D. B.
Kitayama
M.
, et al. 
Functional genomics by integrated analysis of metabolome and transcriptome of Arabidopsis plants over-expressing an MYB transcription factor
Plant J.
2005
, vol. 
42
 (pg. 
218
-
235
)
83
Baxter
C. J.
Sabar
M.
Quick
W. P.
Sweetlove
L. J.
Comparison of changes in fruit gene expression in tomato introgression lines provides evidence of genome-wide transcriptional changes and reveals links to mapped QTLs and described traits
J. Exp. Bot.
2005
, vol. 
56
 (pg. 
1591
-
1604
)
84
Schauer
N.
Semel
Y.
Roessner
U.
Gur
A.
Balbo
I.
Carrari
F.
Pleban
T.
Perez-Melis
A.
Bruedigam
C.
Kopka
J.
, et al. 
Comprehensive metabolic profiling and phenotyping of interspecific introgression lines for tomato improvement
Nat. Biotechnol.
2006
, vol. 
24
 (pg. 
447
-
454
)
85
Walther
D.
Brunnemann
R.
Selbig
J.
The regulatory code for transcriptional response diversity and its relation to genome structural properties in A. thaliana
PLoS Genet.
2007
, vol. 
3
 pg. 
e11
 
86
Reference deleted
87
Keurentjes
J. J. B.
Fu
J. Y.
Terpstra
I. R.
Garcia
J. M.
van den Ackerveken
G.
Snoek
L. B.
Peeters
A. J. M.
Vreugdenhil
D.
Koornneef
M.
Jansen
R. C.
Regulatory network construction in Arabidopsis by using genome-wide gene expression quantitative trait loci
Proc. Natl. Acad. Sci. U.S.A.
2007
, vol. 
104
 (pg. 
1708
-
1713
)
88
Han
J. D.
Bertin
N.
Hao
T.
Goldberg
D. S.
Berriz
G. F.
Zhang
L. V.
Dupuy
D.
Walhout
A. J.
Cusick
M. E.
Roth
F. P.
Vidal
M.
Evidence for dynamically organized modularity in the yeast protein-protein interaction network
Nature
2004
, vol. 
430
 (pg. 
88
-
93
)
89
Uhrig
J. F.
Protein interaction networks in plants
Planta
2006
, vol. 
224
 (pg. 
771
-
781
)
90
Hackbusch
J.
Richter
K.
Muller
J.
Salamini
F.
Uhrig
J. F.
A central role of Arabidopsis thaliana ovate family proteins in networking and subcellular localization of 3-aa loop extension homeodomain proteins
Proc. Natl. Acad. Sci. U.S.A.
2005
, vol. 
102
 (pg. 
4908
-
4912
)
91
Kerrien
S.
Alam-Faruque
Y.
Aranda
B.
Bancarz
I.
Bridge
A.
Derow
C.
Dimmer
E.
Feuermann
M.
Friedrichsen
A.
Huntley
R.
, et al. 
IntAct – open source resource for molecular interaction data
Nucleic Acids Res.
2007
, vol. 
35
 (pg. 
D561
-
D565
)
92
von Mering
C.
Krause
R.
Snel
B.
Cornell
M.
Oliver
S. G.
Fields
S.
Bork
P.
Comparative assessment of large-scale data sets of protein–protein interactions
Nature
2002
, vol. 
417
 (pg. 
399
-
403
)
93
Betts
M. J.
Russell
R. B.
The hard cell: from proteomics to a whole cell model
FEBS Lett.
2007
, vol. 
581
 (pg. 
2870
-
2876
)
94
Goodsell
D. S.
The molecular perspective: major histocompatibility complex
Stem Cells
2005
, vol. 
23
 (pg. 
454
-
455
)
95
Jansen
R.
Yu
H. Y.
Greenbaum
D.
Kluger
Y.
Krogan
N. J.
Chung
S. B.
Emili
A.
Snyder
M.
Greenblatt
J. F.
Gerstein
M.
A Bayesian networks approach for predicting protein–protein interactions from genomic data
Science
2003
, vol. 
302
 (pg. 
449
-
453
)
96
Takamori
S.
Holt
M.
Stenius
K.
Lemke
E. A.
Gronborg
M.
Riedel
D.
Urlaub
H.
Schenck
S.
Brugger
B.
Ringler
P.
, et al. 
Molecular anatomy of a trafficking organelle
Cell
2006
, vol. 
127
 (pg. 
831
-
846
)
97
Uetz
P.
Giot
L.
Cagney
G.
Mansfield
T. A.
Judson
R. S.
Knight
J. R.
Lockshon
D.
Narayan
V.
Srinivasan
M.
Pochart
P.
, et al. 
A comprehensive analysis of protein–protein interactions in Saccharomyces cerevisiae
Nature
2000
, vol. 
403
 (pg. 
623
-
627
)
98
Helbing
D.
Farkas
I.
Vicsek
T.
Simulating dynamical features of escape panic
Nature
2000
, vol. 
407
 (pg. 
487
-
490
)
99
Teichmann
S. A.
Babu
M. M.
Gene regulatory network growth by duplication
Nat. Genet.
2004
, vol. 
36
 (pg. 
492
-
496
)
100
Wagner
A.
How the global structure of protein interaction networks evolves
Proc. R. Soc. London Ser. B
2003
, vol. 
270
 (pg. 
457
-
466
)
101
Pastor-Satorras
R.
Smith
E.
Sole
R. V.
Evolving protein interaction networks through gene duplication
J. Theor. Biol.
2003
, vol. 
222
 (pg. 
199
-
210
)
102
Kashtan
N.
Alon
U.
Spontaneous evolution of modularity and network motifs
Proc. Natl. Acad. Sci. U.S.A.
2005
, vol. 
102
 (pg. 
13773
-
13778
)
103
Pereira-Leal
J. B.
Teichmann
S. A.
Novel specificities emerge by stepwise duplication of functional modules
Genome Res.
2005
, vol. 
15
 (pg. 
552
-
559
)
104
Ober
D.
Seeing double: gene duplication and diversification in plant secondary metabolism
Trends Plant Sci.
2005
, vol. 
10
 (pg. 
444
-
449
)
105
Carlisle
S. M.
Blakeley
S. D.
Hemmingsen
S. M.
Trevanion
S. J.
Hiyoshi
T.
Kruger
N. J.
Dennis
D. T.
Pyrophosphate-dependent phosphofructokinase – conservation of protein sequence between the α-subunit and β-subunit and with the ATP-dependent phosphofructokinase
J. Biol. Chem.
1990
, vol. 
265
 (pg. 
18366
-
18371
)
106
Iglesias
A. A.
Barry
G. F.
Meyer
C.
Bloksberg
L.
Nakata
P. A.
Greene
T.
Laughlin
M. J.
Okita
T. W.
Kishore
G. M.
Preiss
J.
Expression of the potato-tuber ADP-glucose pyrophosphorylase in Escherichia coli.
J. Biol. Chem.
1993
, vol. 
268
 (pg. 
1081
-
1086
)
107
Tovar-Mendez
A.
Miernyk
J. A.
Randall
D. D.
Regulation of pyruvate dehydrogenase complex activity in plant cells
Eur. J. Biochem.
2003
, vol. 
270
 (pg. 
1043
-
1049
)
108
Loque
D.
Lalonde
S.
Looger
L. L.
von Wiren
N.
Frommer
W. B.
A cytosolic trans-activation domain essential for ammonium uptake
Nature
2007
, vol. 
446
 (pg. 
195
-
198
)
109
Reinders
A.
Schulze
W.
Kuhn
C.
Barker
L.
Schulz
A.
Ward
J. M.
Frommer
W. B.
Protein–protein interactions between sucrose transporters of different affinities colocalized in the same enucleate sieve element
Plant Cell.
2002
, vol. 
14
 (pg. 
1567
-
1577
)
110
Bartoli
C. G.
Pastori
G. M.
Foyer
C. H.
Ascorbate biosynthesis in mitochondria is linked to the electron transport chain between complexes III and IV
Plant Physiol.
2000
, vol. 
123
 (pg. 
335
-
344
)
111
Heinemeyer
J.
Eubel
H.
Wehmhoner
D.
Jansch
L.
Braun
H. P.
Proteomic approach to characterize the supramolecular organization of photosystems in higher plants
Phytochemistry
2004
, vol. 
65
 (pg. 
1683
-
1692
)
112
Dudkina
N.
Heinemeyer
J.
Keegstra
W.
Boekema
E.
Braun
H. P.
Structure of dimeric ATP synthase from mitochondria
Biochim. Biophys. Acta Bioenerg.
2006
(pg. 
301
-
301
)
113
Suss
K. H.
Prokhorenko
I.
Adler
K.
In situ association of Calvin cycle enzymes, ribulose-1,5-bisphosphate carboxylase/oxygenase activase, ferredoxin–NADP+ reductase, and nitrite reductase with thylakoid and pyrenoid membranes of Chlamydomonas reinhardtii chloroplasts as revealed by immunoelectron microscopy
Plant Physiol.
1995
, vol. 
107
 (pg. 
1387
-
1397
)
114
Moller
B. L.
Conn
E. E.
The biosynthesis of cyanogenic glucosides in higher plants. Channeling of intermediates in dhurrin biosynthesis by a microsomal system from Sorghum bicolor
(Linn.) Moench. J. Biol. Chem.
1980
, vol. 
255
 (pg. 
3049
-
3056
)
115
Winkel-Shirley
B.
Biosynthesis of flavonoids and effects of stress
Curr. Opin. Plant Biol.
2002
, vol. 
5
 (pg. 
218
-
223
)
116
Achnine
L.
Blancaflor
E. B.
Rasmussen
S.
Dixon
R. A.
Colocalization of L-phenylalanine ammonia-lyase and cinnamate 4-hydroxylase for metabolic channeling in phenylpropanoid biosynthesis
Plant Cell.
2004
, vol. 
16
 (pg. 
3098
-
3109
)
117
Panicot
M.
Minguet
E. G.
Ferrando
A.
Alcazar
R.
Blazquez
M. A.
Carbonell
J.
Altabella
T.
Koncz
C.
Tiburcio
A. F.
A polyamine metabolon involving aminopropyl transferase complexes in Arabidopsis
Plant Cell.
2002
, vol. 
14
 (pg. 
2539
-
2551
)
118
Jorgensen
K.
Rasmussen
A. V.
Morant
M.
Nielsen
A. H.
Bjarnholt
N.
Zagrobelny
M.
Bak
S.
Moller
B. L.
Metabolon formation and metabolic channeling in the biosynthesis of plant natural products
Curr. Opin. Plant Biol.
2005
, vol. 
8
 (pg. 
280
-
291
)
119
Winkel
B. S. J.
Metabolic channeling in plants
Annu. Rev. Plant Biol.
2004
, vol. 
55
 (pg. 
85
-
107
)
120
Giege
P.
Heazlewood
J. L.
Roessner-Tunali
U.
Millar
A. H.
Fernie
A. R.
Leaver
C. J.
Sweetlove
L. J.
Enzymes of glycolysis are functionally associated with the mitochondrion in Arabidopsis cells
Plant Cell
2003
, vol. 
15
 (pg. 
2140
-
2151
)
121
Fieulaine
S.
Lunn
J. E.
Borel
F.
Ferrer
J. L.
The structure of a cyanobacterial sucrose-phosphatase reveals the sugar tongs that release free sucrose in the cell
Plant Cell
2005
, vol. 
17
 (pg. 
2049
-
2058
)
122
Shearer
G.
Lee
J. C.
Koo
J. A.
Kohl
D. H.
Quantitative estimation of channeling from early glycolytic intermediates to CO in intact Escherichia coli.
FEBS J.
2005
, vol. 
272
 (pg. 
3260
-
3269
)
123
Roessner
U.
Luedemann
A.
Brust
D.
Fiehn
O.
Linke
T.
Willmitzer
L.
Fernie
A. R.
Metabolic profiling allows comprehensive phenotyping of genetically or environmentally modified plant systems
Plant Cell
2001
, vol. 
13
 (pg. 
11
-
29
)
124
Kopka
J.
Fernie
A.
Weckwerth
W.
Gibon
Y.
Stitt
M.
Metabolite profiling in plant biology: platforms and destinations
Genome Biol.
2004
, vol. 
5
 pg. 
109
 
125
Sumner
L. W.
Mendes
P.
Dixon
R. A.
Plant metabolomics: large-scale phytochemistry in the functional genomics era
Phytochemistry
2003
, vol. 
62
 (pg. 
817
-
836
)
126
Breitling
R.
Pitt
A. R.
Barrett
M. P.
Precision mapping of the metabolome
Trends Biotechnol.
2006
, vol. 
24
 (pg. 
543
-
548
)
127
Nobeli
I.
Ponstingl
H.
Krissinel
E. B.
Thornton
J. M.
A structure-based anatomy of the E. coli metabolome
J. Mol. Biol.
2003
, vol. 
334
 (pg. 
697
-
719
)
128
Reference deleted
129
Tieman
D. M.
Zeigler
M.
Schmelz
E. A.
Taylor
M. G.
Bliss
P.
Kirst
M.
Klee
H. J.
Identification of loci affecting flavour volatile emissions in tomato fruits
J. Exp. Bot.
2006
, vol. 
57
 (pg. 
887
-
896
)
130
Fait
A.
Angelovici
R.
Less
H.
Ohad
I.
Urbanczyk-Wochniak
E.
Fernie
A. R.
Galili
G.
Arabidopsis seed development and germination is associated with temporally distinct metabolic switches
Plant Physiol.
2006
, vol. 
142
 (pg. 
839
-
854
)
131
Fritz
C.
Mueller
C.
Matt
P.
Feil
R.
Stitt
M.
Impact of the C–N status on the amino acid profile in tobacco source leaves
Plant Cell Environ.
2006
, vol. 
29
 (pg. 
2055
-
2076
)
132
Noctor
G.
Novitskaya
L.
Lea
P. J.
Foyer
C. H.
Co-ordination of leaf minor amino acid contents in crop species: significance and interpretation
J. Exp. Bot.
2002
, vol. 
53
 (pg. 
939
-
945
)
133
Roessner-Tunali
U.
Hegemann
B.
Lytovchenko
A.
Carrari
F.
Bruedigam
C.
Granot
D.
Fernie
A.
Metabolic profiling of transgenic tomato plants overexpressing hexokinase reveals that the influence of hexose phosphorylation diminishes during fruit development
Plant Physiol.
2003
, vol. 
133
 (pg. 
84
-
99
)
134
Sweetlove
L. J.
Fernie
A. R.
Regulation of metabolic networks: understanding metabolic complexity in the systems biology era
New Phytol.
2005
, vol. 
168
 (pg. 
9
-
24
)
135
Zhu
X. H.
Galili
G.
Increased lysine synthesis coupled with a knockout of its catabolism synergistically boosts lysine content and also transregulates the metabolism of other amino acids in Arabidopsis seeds
Plant Cell
2003
, vol. 
15
 (pg. 
845
-
853
)
136
Broeckling
C. D.
Huhman
D. V.
Farag
M. A.
Smith
J. T.
May
G. D.
Mendes
P.
Dixon
R. A.
Sumner
L. W.
Metabolic profiling of Medicago truncatula cell cultures reveals the effects of biotic and abiotic elicitors on metabolism
J. Exp. Bot.
2005
, vol. 
56
 (pg. 
323
-
336
)
137
Fiehn
O.
Weckwerth
W.
Deciphering metabolic networks
Eur. J. Biochem.
2003
, vol. 
270
 (pg. 
579
-
588
)
138
Steuer
R.
Kurths
J.
Fiehn
O.
Weckwerth
W.
Interpreting correlations in metabolomic networks
Biochem. Soc. Trans.
2003
, vol. 
31
 (pg. 
1476
-
1478
)
139
Kose
F.
Weckwerth
W.
Linke
T.
Fiehn
O.
Visualizing plant metabolomic correlation networks using clique-metabolite matrices
Bioinformatics
2001
, vol. 
17
 (pg. 
1198
-
1208
)
140
Bartlem
D.
Lambein
I.
Okamoto
T.
Itaya
A.
Uda
Y.
Kijima
F.
Tamaki
Y.
Nambara
E.
Naito
S.
Mutation in the threonine synthase gene results in an over-accumulation of soluble methionine in Arabidopsis
Plant Physiol.
2000
, vol. 
123
 (pg. 
101
-
110
)
141
Schauer
N.
Fernie
A. R.
Plant metabolomics: towards biological function and mechanism
Trends Plant Sci.
2006
, vol. 
11
 (pg. 
508
-
516
)
142
Gibon
Y.
Usadel
B.
Blaesing
O. E.
Kamlage
B.
Hoehne
M.
Trethewey
R.
Stitt
M.
Integration of metabolite with transcript and enzyme activity profiling during diurnal cycles in Arabidopsis rosettes
Genome Biol.
2006
, vol. 
7
 pg. 
R76
 
143
Hirai
M.
Fujikawa
Y.
Yano
M.
Dayan
G.
Kanaya
S.
Saito
K.
Transcriptome and metabolome analyses reveal a whole adaptive process of plant to sulfur deficiency
Plant Cell Physiol.
2004
, vol. 
45
 (pg. 
S122
-
S122
)
144
Nikiforova
V. J.
Kopka
J.
Tolstikov
V.
Fiehn
O.
Hopkins
L.
Hawkesford
M. J.
Hesse
H.
Hoefgen
R.
Systems rebalancing of metabolism in response to sulfur deprivation, as revealed by metabolome analysis of Arabidopsis plants
Plant Physiol.
2005
, vol. 
138
 (pg. 
304
-
318
)
145
Nikiforova
V. J.
Daub
C. O.
Hesse
H.
Willmitzer
L.
Hoefgen
R.
Integrative gene-metabolite network with implemented causality deciphers informational fluxes of sulphur stress response
J. Exp. Bot.
2005
, vol. 
56
 (pg. 
1887
-
1896
)
146
Morgenthal
K.
Wienkoop
S.
Scholz
M.
Selbig
J.
Weckwerth
W.
Correlative GC–TOF–MS-based metabolite profiling and LC–MS-based protein profiling reveal time-related systemic regulation of metabolite-protein networks and improve pattern recognition for multiple biomarker selection Metabolomics
2005
, vol. 
1
 (pg. 
109
-
120
)
147
Carrari
F.
Baxter
C.
Usadel
B.
Urbanczyk-Wochniak
E.
Zanor
M. I.
Nunes-Nesi
A.
Nikiforova
V.
Centero
D.
Ratzka
A.
Pauly
M.
, et al. 
Integrated analysis of metabolite and transcript levels reveals the metabolic shifts that underlie tomato fruit development and highlight regulatory aspects of metabolic network behavior
Plant Physiol.
2006
, vol. 
142
 (pg. 
1380
-
1396
)
148
Stitt
M.
Gibon
Y.
Lunn
J. E.
Piques
M.
Multilevel genomics analysis of carbon signalling during low carbon availability: coordinating the supply and utilisation of carbon in a fluctuating environment
Funct. Plant Biol.
2007
, vol. 
34
 (pg. 
526
-
549
)
149
Fernie
A. R.
Geigenberger
P.
Stitt
M.
Flux: an important, but neglected component of functional genomics
Curr. Opin. Plant Biol.
2005
, vol. 
8
 (pg. 
174
-
182
)
150
Rossell
S.
van der Weijden
C. C.
Lindenbergh
A.
van Tuijl
A.
Francke
C.
Bakker
B. M.
Westerhoff
H. V.
Unraveling the complexity of flux regulation: a new method demonstrated for nutrient starvation in Saccharomyces cerevisiae
Proc. Natl. Acad. Sci. U.S.A.
2006
, vol. 
103
 (pg. 
2166
-
2171
)
151
Roscher
A.
Kruger
N. J.
Ratcliffe
R. G.
Strategies for metabolic flux analysis in plants using isotope labelling
J. Biotechnol.
2000
, vol. 
77
 (pg. 
81
-
102
)
152
Massou
S.
Nicolas
C.
Letisse
F.
Portais
J. C.
NMR-based fluxomics: Quantitative 2D NMR methods for isotopomers analysis
Phytochemistry
2007
, vol. 
68
 (pg. 
2330
-
2340
)
153
Sekiyama
Y.
Kikuchi
J.
Towards dynamic metabolic network measurements by multi-dimensional NMR-based fluxomics
Phytochemistry
2007
, vol. 
68
 (pg. 
2320
-
2329
)
154
Wiechert
W.
Mollney
M.
Petersen
S.
de Graaf
A. A.
A universal framework for 13C metabolic flux analysis
Metab. Eng.
2001
, vol. 
3
 (pg. 
265
-
283
)
155
Ratcliffe
R. G.
Shachar-Hill
Y.
Measuring multiple fluxes through plant metabolic networks
Plant J.
2006
, vol. 
45
 (pg. 
490
-
511
)
156
Rios-Estepa
R.
Lange
B. M.
Experimental and mathematical approaches to modeling plant metabolic networks
Phytochemistry
2007
, vol. 
68
 (pg. 
2351
-
2374
)
157
Sauer
U.
Metabolic networks in motion: 13C-based flux analysis
Mol. Syst. Biol.
2006
, vol. 
2
 pg. 
62
 
158
Schwender
J.
Ohlrogge
J.
Shachar-Hill
Y.
Understanding flux in plant metabolic networks
Curr. Opin. Plant Biol.
2004
, vol. 
7
 (pg. 
309
-
317
)
159
Lunn
J. E.
Compartmentation in plant metabolism
J. Exp. Bot.
2007
, vol. 
58
 (pg. 
35
-
47
)
160
Allen
D. K.
Shachar-Hill
Y.
Ohlrogge
J. B.
Compartment-specific labeling information in 13C metabolic flux analysis of plants
Phytochemistry
2007
, vol. 
68
 (pg. 
2197
-
2210
)
161
Dancer
J.
Hatzfeld
W. D.
Stitt
M.
Cytosolic cycles regulate the turnover of sucrose in heterotrophic cell-suspension cultures of Chenopodium rubrum
L. Planta
1990
, vol. 
182
 (pg. 
223
-
231
)
162
Dieuaide-Noubhani
M.
Raffard
G.
Canioni
P.
Pradet
A.
Raymond
P.
Quantification of compartmented metabolic fluxes in maize root tips using isotope distribution from 13C- or 14C-labeled glucose
J. Biol. Chem.
1995
, vol. 
270
 (pg. 
13147
-
13159
)
163
Rontein
D.
Dieuaide-Noubhani
M.
Dufourc
E. J.
Raymond
P.
Rolin
D.
The metabolic architecture of plant cells. Stability of central metabolism and flexibility of anabolic pathways during the growth cycle of tomato cells
J. Biol. Chem.
2002
, vol. 
277
 (pg. 
43948
-
43960
)
164
Alonso
A. P.
Vigeolas
H.
Raymond
P.
Rolin
D.
Dieuaide-Noubhani
M.
A new substrate cycle in plants. Evidence for a high glucose-phosphate-to-glucose turnover from in vivo steady-state and pulse-labeling experiments with [13C]glucose and [14C]glucose
Plant Physiol.
2005
, vol. 
138
 (pg. 
2220
-
2232
)
165
Kruger
N. J.
Le Lay
P.
Ratcliffe
R. G.
Vacuolar compartmentation complicates the steady-state analysis of glucose metabolism and forces reappraisal of sucrose cycling in plants
Phytochemistry
2007
, vol. 
68
 (pg. 
2189
-
2196
)
166
Farre
E. M.
Tiessen
A.
Roessner
U.
Geigenberger
P.
Trethewey
R. N.
Willmitzer
L.
Analysis of the compartmentation of glycolytic intermediates, nucleotides, sugars, organic acids, amino acids, and sugar alcohols in potato tubers using a nonaqueous fractionation method
Plant Physiol.
2001
, vol. 
127
 (pg. 
685
-
700
)
167
Wirtz
W.
Stitt
M.
Heldt
H. W.
Enzymic determination of metabolites in the subcellular compartments of spinach protoplasts
Plant Physiol.
1980
, vol. 
66
 (pg. 
187
-
193
)
168
Libourel
I. G.
van Bodegom
P. M.
Fricker
M. D.
Ratcliffe
R. G.
Nitrite reduces cytoplasmic acidosis under anoxia
Plant Physiol.
2006
, vol. 
142
 (pg. 
1710
-
1717
)
169
Gout
E.
Bligny
R.
Pascal
N.
Douce
R.
13C nuclear magnetic resonance studies of malate and citrate synthesis and compartmentation in higher plant cells
J. Biol. Chem.
1993
, vol. 
268
 (pg. 
3986
-
3992
)
170
Ap Rees
T.
Davies
D. D.
Compartmentation of plant metabolism
Biochemistry of Plants
1983
London
Academic Press
(pg. 
87
-
115
)
171
Deuschle
K.
Chaudhuri
B.
Okumoto
S.
Lager
I.
Lalonde
S.
Frommer
W. B.
Rapid metabolism of glucose detected with FRET glucose nanosensors in epidermal cells and intact roots of Arabidopsis RNA-silencing mutants
Plant Cell.
2006
, vol. 
18
 (pg. 
2314
-
2325
)
172
Lalonde
S.
Ehrhardt
D. W.
Frommer
W. B.
Shining light on signaling and metabolic networks by genetically encoded biosensors
Curr. Opin. Plant Biol.
2005
, vol. 
8
 (pg. 
574
-
581
)
173
Wiechert
W.
Schweissgut
O.
Takanaga
H.
Frommer
W. B.
Fluxomics: mass spectrometry versus quantitative imaging
Curr. Opin. Plant. Biol.
2007
, vol. 
10
 (pg. 
323
-
330
)
174
Dano
S.
Sorensen
P. G.
Hynne
F.
Sustained oscillations in living cells
Nature
1999
, vol. 
402
 (pg. 
320
-
322
)
175
Reference deleted
176
Matic
S.
Geisler
D. A.
Moller
I. M.
Widell
S.
Rasmusson
A. G.
Alamethicin permeabilizes the plasma membrane and mitochondria but not the tonoplast in tobacco (Nicotiana tabacum L. cv Bright Yellow) suspension cells
Biochem. J.
2005
, vol. 
389
 (pg. 
695
-
704
)
177
Matsuda
F.
Morino
K.
Ano
R.
Kuzawa
M.
Wakasa
K.
Miyagawa
H.
Metabolic flux analysis of the phenylpropanoid pathway in elicitor-treated potato tuber tissue
Plant Cell Physiol.
2005
, vol. 
46
 (pg. 
454
-
466
)
178
Baxter
C. J.
Liu
J.
Fernie
A. R.
Sweetlove
L. J.
Determination of metabolic fluxes in a non-steady-state system
Phytochemistry
2007
, vol. 
68
 (pg. 
2313
-
2319
)
179
Roessner-Tunali
U.
Liu
J.
Leisse
A.
Balbo
I.
Perez-Melis
A.
Willmitzer
L.
Fernie
A. R.
Kinetics of labelling of organic and amino acids in potato tubers by gas chromatography–mass spectrometry following incubation in C labelled isotopes
Plant J.
2004
, vol. 
39
 (pg. 
668
-
679
)
180
Huege
J.
Sulpice
R.
Gibon
Y.
Lisec
J.
Koehl
K.
Kopka
J.
GC–EI–TOF–MS analysis of in vivo carbon-partitioning into soluble metabolite pools of higher plants by monitoring isotope dilution after 13CO2 labelling
Phytochemistry
2007
, vol. 
68
 (pg. 
2258
-
2272
)
181
Romisch-Margl
W.
Schramek
N.
Radykewicz
T.
Ettenhuber
C.
Eylert
E.
Huber
C.
Romisch-Margl
L.
Schwarz
C.
Dobner
M.
Demmel
N.
, et al. 
13CO2 as a universal metabolic tracer in isotopologue perturbation experiments
Phytochemistry
2007
, vol. 
68
 (pg. 
2273
-
2289
)
182
Noh
K.
Wiechert
W.
Experimental design principles for isotopically instationary 13C labeling experiments
Biotechnol. Bioeng.
2006
, vol. 
94
 (pg. 
234
-
251
)
183
Noh
K.
Wahl
A.
Wiechert
W.
Computational tools for isotopically instationary 13C labeling experiments under metabolic steady state conditions
Metab. Eng.
2006
, vol. 
8
 (pg. 
554
-
577
)
184
Noh
K.
Gronke
K.
Luo
B.
Takors
R.
Oldiges
M.
Wiechert
W.
Metabolic flux analysis at ultra short time scale: isotopically non-stationary 13C labeling experiments
J. Biotechnol.
2007
, vol. 
129
 (pg. 
249
-
267
)
185
Schmidt
J.
Boettcher
C.
Kuhnt
C.
Kutchan
T. M.
Zenk
M. H.
Poppy alkaloid profiling by electrospray tandem mass spectrometry and electrospray FT–ICR mass spectrometry after [ring-13-C-6]tyramine feeding
Phytochemistry
2007
, vol. 
68
 (pg. 
189
-
202
)
186
Tjaden
J.
Möhlmann
T.
Kampfenkel
K.
Henrichs
G.
Neuhaus
H. E.
Altered plastidic ATP/ADP transporter activity influences potato (Solanum tuberosum L.) tuber morphology, yield and composition of tuber starch
Plant J.
1998
, vol. 
16
 (pg. 
531
-
540
)
187
Regierer
B.
Fernie
A. R.
Springer
F.
Perez-Melis
A.
Leisse
A.
Koehl
K.
Willmitzer
L.
Geigenberger
P.
Kossman
J.
Starch content and yield increase as a result of altering adenylate pools in transgenic plants
Nat. Biotechnol.
2002
, vol. 
20
 (pg. 
1256
-
1260
)
188
Geigenberger
P.
Regierer
B.
Nunes-Nesi
A.
Leisse
A.
Urbanczyk-Wochniak
E.
Springer
F.
van Dongen
J. T.
Kossmann
J.
Fernie
A. R.
Inhibition of de novo pyrimidine synthesis in growing potato tubers leads to a compensatory stimulation of the pyrimidine salvage pathway and a subsequent increase in biosynthetic performance
Plant Cell
2005
, vol. 
17
 (pg. 
2077
-
2088
)
189
Jenner
H. L.
Winning
B. M.
Millar
A. H.
Tomlinson
K. L.
Leaver
C. J.
Hill
S. A.
NAD malic enzyme and the control of carbohydrate metabolism in potato tubers. Plant. Physiol
2001
, vol. 
126
 (pg. 
1139
-
1149
)
190
McKibbin
R. S.
Muttucumaru
N.
Paul
M. J.
Powers
S. J.
Burrell
M. M.
Coates
S.
Purcell
P. C.
Tiessen
A.
Geigenberger
P.
Halford
N. G.
Production of high-starch, low-glucose potatoes through over-expression of the metabolic regulator SnRK1
Plant Biotechnol. J.
2006
, vol. 
4
 (pg. 
409
-
418
)
191
Uhrig
J. F.
Protein interaction networks in plants
Planta
2006
, vol. 
224
 (pg. 
771
-
781
)