subcellular flux analysis of central metabolism in a ... · subcellular flux analysis of central...

Subcellular Flux Analysis of Central Metabolism in aHeterotrophic Arabidopsis Cell Suspension UsingSteady-State Stable Isotope Labeling1[W][OA]

Shyam K. Masakapalli, Pascaline Le Lay, Joanna E. Huddleston, Naomi L. Pollock,Nicholas J. Kruger2, and R. George Ratcliffe2*

Department of Plant Sciences, University of Oxford, Oxford OX1 3RB, United Kingdom

The presence of cytosolic and plastidic pathways of carbohydrate oxidation is a characteristic feature of plant cell metabolism.Ideally, steady-state metabolic flux analysis, an emerging tool for creating flux maps of heterotrophic plant metabolism, wouldcapture this feature of the metabolic phenotype, but the extent to which this can be achieved is uncertain. To address thisquestion, fluxes through the pathways of central metabolism in a heterotrophic Arabidopsis (Arabidopsis thaliana) cell sus-pension culture were deduced from the redistribution of label in steady-state 13C-labeling experiments using [1-13C]-,[2-13C]-, and [U-13C6]glucose. Focusing on the pentose phosphate pathway (PPP), multiple data sets were fitted simultaneouslyto models in which the subcellular compartmentation of the PPP was altered. The observed redistribution of the label could beexplained by any one of three models of the subcellular compartmentation of the oxidative PPP, but other biochemical evidencefavored the model in which the oxidative steps of the PPP were duplicated in the cytosol and plastids, with flux through thesereactions occurring largely in the cytosol. The analysis emphasizes the inherent difficulty of analyzing the PPP withoutpredefining the extent of its compartmentation and the importance of obtaining high-quality data that report directly onspecific subcellular processes. The Arabidopsis flux map also shows that the potential ATP yield of respiration in heterotrophicplant cells can greatly exceed the direct metabolic requirements for biosynthesis, highlighting the need for caution whenpredicting flux through metabolic networks using assumptions based on the energetics of resource utilization.

Extensive subcellular compartmentation, withunique locations for many steps and pathways, aswell as the duplication of other steps and pathways indifferent compartments, adds greatly to the structuralcomplexity of the plant metabolic network (Lunn,2007; Kruger and Ratcliffe, 2008). The need to considerdiscrete pools of metabolites in specific compartments,and the transporters that link them, complicates thequest for a detailed, predictive understanding of theregulation of plant metabolism; as a result, it remainsdifficult to manipulate flows of material throughthe central metabolic network in a predictableway (Carrari et al., 2003; Kruger and Ratcliffe, 2008;Sweetlove et al., 2008).

The emergence of steady-state metabolic flux anal-ysis (MFA) as a practicable systems biology tool forgenerating flux maps of the central metabolic path-ways in plants offers new opportunities for analyzing

plant metabolic phenotypes (Ratcliffe and Shachar-Hill, 2006; Libourel and Shachar-Hill, 2008;Schwender, 2008; Kruger and Ratcliffe, 2009). In thisapproach, substrates labeled with stable isotopes areintroduced into the network, and fluxes are deter-mined by measuring the labeling of the system after ithas reached an isotopic and metabolic steady state.Subcellular compartmentation of steps and pathwayscan be incorporated into the model that describes theredistribution of the label, and flux maps of centralcarbon metabolism in plant cells typically aim todistinguish between fluxes in the cytosol, mitochon-dria, and plastids. In principle, compartmented fluxescan be deduced from the labeling patterns of metab-olites that are synthesized in a single compartment(Ratcliffe and Shachar-Hill, 2005, 2006; Allen et al.,2007), and this property has been exploited withvarying degrees of success in different studies.

Most recent flux maps, such as those for developingembryos of oilseed rape (Brassica napus; Schwenderet al., 2006; Junker et al., 2007) and sunflower (Heli-anthus annuus; Alonso et al., 2007a) or the one for aheterotrophic cell suspension of Arabidopsis (Arabi-dopsis thaliana; Williams et al., 2008), have well-definedmitochondrial fluxes. This reflects the high sensitivityof the available label measurements to these fluxes, butit also reflects the arbitrary way in which the modelsassume that some fluxes, for example from citrate to2-oxoglutarate, are exclusively mitochondrial andothers, for example between malate and oxaloacetate,

1 This work was supported by the Biotechnology and BiologicalSciences Research Council (grant no. 43/B17210), United Kingdom.

2 These authors contributed equally to the article.* Corresponding author; e-mail [email protected] author responsible for distribution of materials integral to the

findings presented in this article in accordance with the policydescribed in the Instructions for Authors (www.plantphysiol.org) is:R. George Ratcliffe ([email protected]).

[W] The online version of this article contains Web-only data.[OA] Open Access articles can be viewed online without a sub-

scription.www.plantphysiol.org/cgi/doi/10.1104/pp.109.151316

602 Plant Physiology�, February 2010, Vol. 152, pp. 602–619, www.plantphysiol.org � 2009 American Society of Plant Biologists www.plantphysiol.orgon May 11, 2020 - Published by Downloaded from

Copyright © 2010 American Society of Plant Biologists. All rights reserved.

http://www.plantphysiol.org

occur in an unspecified compartment. There is lessconsensus in the description of the pathways of car-bohydrate oxidation, with the subcellular compart-mentation of glycolysis and the oxidative pentosephosphate pathway (PPP) being described at variouslevels of detail. For example, a recent model of soy-bean (Glycine max) cotyledons included completepathways for glycolysis and the PPP in both cytosoland plastids (Iyer et al., 2008), whereas models fordeveloping sunflower embryos (Alonso et al., 2007a),maize (Zea mays) root tips (Alonso et al., 2007b), and anArabidopsis cell culture (Williams et al., 2008) con-fined the oxidative and nonoxidative steps of the PPPto the plastid. The limited compartmentation of thepathways of carbohydrate oxidation in many fluxmaps can be traced to an analysis of developingoilseed rape embryos (Schwender et al., 2003), whichindicated that the labeling data could be satisfactorilyexplained without duplication of glycolysis and thePPP. Since some of the enzymes of these pathways areknown to be present in both compartments, the modelimplied fast exchange of intermediates across theplastid envelope, giving rise to a situation in whichthe physically compartmented pathways are func-tionally uncompartmented (Schwender et al., 2003;Ratcliffe and Shachar-Hill, 2006).Although the extent to which the steps of the PPP

are duplicated in the cytosol and plastids has not beenfully established, most of the biochemical evidencesuggests that the enzymes for the oxidative steps atleast are present in both compartments (Kruger andvon Schaewen, 2003). So modeling the PPP as fullyduplicated in both cytosol and plastids, or alterna-tively as exclusively plastidic, ignores the most likelystructure of the network. Models based on all threenetwork structures have now been tested in an Arabi-dopsis cell culture. Building on a steady-state MFA ofthe same culture (Williams et al., 2008), similar exper-iments were conducted with multiple substrates andthe labeling data sets were simultaneously fitted tomodels in which the subcellular compartmentation ofthe PPP was varied. Acceptable fits to the data werefound for all three models, emphasizing the impor-tance of building the most appropriate level of sub-cellular compartmentation into the model at theoutset. Moreover, the flux maps showed that the datawere consistent with a major contribution from thecytosol to the oxidative steps of the PPP.

RESULTS

Biosynthetic Outputs

The Arabidopsis cell culture was maintained in thedark with Glc as the sole respiratory substrate, and theprincipal activities of the metabolic network, togetherwith the associated biosynthetic outputs, were as-sessed by determining the redistribution of radiolabelfollowing metabolism of [U-14C]Glc (Table I). Overall,

the pattern of labeling was comparable to that ob-served previously for heterotrophic cell cultures ofArabidopsis and is similar to that reported for a widerange of nonphotosynthetic plant material (Krugeret al., 2007a). However, extending the analysis toinclude components that are frequently ignored dur-ing similar fractionation procedures, such as the or-ganic phase obtained during chloroform/methanolfractionation and the medium following removal ofcells, allowed a more comprehensive analysis of thefate of metabolized Glc and measurement of the la-beling of lipids and ethanol. The uptake of label waslinear over 24 h, and the proportion of radioactivityrecovered in each of the major chemical fractions wasthe same after 12 and 24 h of incubation (SupplementalFig. S1). This implies that the pools of metabolicintermediates had achieved isotopic steady statewithin this time and that the accumulation of radiola-bel in the products reported in Table I reflects the rateof synthesis of these compounds rather than justisotopic equilibration within existing pools. The datashow that almost 40% of the metabolized Glc wasreleased as CO2, while less than 25% of the carbon was

Table I. Metabolism of [U-14C]Glc by cell suspension culturesof Arabidopsis

Samples (5 mL, 0.618 6 0.017 g fresh weight) from 6-d-old culturesgrown in Glc at natural abundance 13C were supplemented with 185kBq [U-14C]Glc (yielding a specific activity of 308 6 15 MBq mol21)and incubated for 24 h prior to extraction in methanol and subsequentmetabolic fractionation. Total 14C uptake is the sum of radioactivity inCO2, ethanol, and the chloroform-soluble, aqueous methanol-soluble,and methanol-insoluble fractions in each cell culture. Each value is themean 6 SE of the distribution of radioactivity in four cell cultures.

Metabolic Fraction Radioactivity in Specified Fraction

dpm

CO2 1,076,468 6 35,772Chloroform soluble 63,439 6 5,420Neutral lipids 41,242 6 2,217Free fatty acids 1,083 6 225Neutral P-lipids 5,046 6 722Acidic P-lipids 21,597 6 2,921

Methanol soluble 1,021,858 6 26,185Organic acids/P-esters 109,533 6 553Amino acids 386,305 6 13,201Asp/Glu 31,938 6 2,563Others 334,737 6 20,786

Sugars 389,986 6 17,377Suc 251,596 6 16,562Glc 58,680 6 5,384Fru 37,060 6 2,298Others 3,943 6 623

Methanol insoluble 591,272 6 53,215Starch 72,424 6 11,971Protein 116,647 6 10,699Cell wall 103,763 6 13,417Residue 77,010 6 21,924

Ethanol 18,500 6 2,115Total 14C uptake 2,771,536 6 67,703Total 14C recovered 2,361,265 6 25,419Recovery (%) 84.2 6 2.1

Steady-State Metabolic Flux Analysis in Arabidopsis

Plant Physiol. Vol. 152, 2010 603 www.plantphysiol.orgon May 11, 2020 - Published by Downloaded from



converted into biopolymers (comprising starch, cellwall, protein, and lipid), with the remainder accumu-lating mainly as soluble sugars, organic acids, andamino acids. While this analysis defines the majoroutput fluxes associated with growth, further infor-mation is required to determine the fluxes that gener-ate the precursors needed for biomass production, andthis can be provided by steady-state MFA.

Metabolic Network

In MFA, the metabolic network is reduced to amodel that can account for the observed redistributionof the label. Typically, this leads to a model in which (1)multiple steps between branch points in the networkare represented as a single reversible or irreversiblestep and (2) subcellular pools of metabolites aretreated as single pools in the absence of compart-ment-specific information about their labeling. Thescope of the model is strongly influenced by the choiceof label precursor and by the extent and quality of themeasurements that define the redistribution of thelabel after the system has reached an isotopic andmetabolic steady state. It has already been shown thatstatistically robust flux maps of central carbon metab-olism can be obtained for heterotrophic Arabidopsiscell suspensions using [1-13C]Glc as the label source(Williams et al., 2008), and here the aim was to estab-lish whether improved definition of the subcellularcompartmention of carbohydrate oxidation could beobtained by combining the results of experiments with[1-13C]Glc, [2-13C]Glc, and [U-13C6]Glc as label precur-sors.

The existing model, in which the oxidative andnonoxidative reactions of the PPP were confined to theplastids (Fig. 1A), was compared with two alterna-tives: (1) a model in which the oxidative steps of thePPP were allowed to occur in both the cytosol andplastids, with the nonoxidative steps confined to theplastids (Fig. 1B); and (2) a model in which thecomplete PPP was present in both compartments(Fig. 1C). Models were constructed in the formatused by the steady-state MFA software 13C-FLUX,and full details of the model in which the oxidativesteps of the PPP were duplicated in the cytosol andplastids (Figs. 1B and 2) are given in Supplemental FileS1. In essence, the model defines the relationshipbetween the carbon atoms in substrates and productsfor each step in the network so that the underlyingfluxes can be deduced from the redistribution of labelin the steady-state labeling experiment. The modelsused here are directly related in most respects to theone used earlier (Williams et al., 2008), and they wereconstructed in the same way. Several features of themodels, and the biochemical networks they represent,should be noted.

First, analysis of the subcellular distribution of fluxdepends on obtaining information about the steady-state labeling of metabolic intermediates in specificcompartments. Here, the labeling of the plastidic pools

Figure 1. Alternative metabolic models for the subcellular compart-mentation of the PPP. A, Oxidative and nonoxidative reactions of thePPP confined to the plastid. B, Oxidative steps of the PPP in both thecytosol and plastid, with the nonoxidative steps in the plastid. C,The complete pathway in both compartments. Net and exchange fluxesare shown with unidirectional and bidirectional arrows, respectively,and the direction of a unidirectional arrow corresponds to a positiveflux in the output of the model. Flux names are given in italics, and thepentose phosphates are grouped into a single indistinguishable pool inall three models. Ery-4-P, Erythrose 4-phosphate; Rbu-5-P, ribulose5-phosphate; Sed-7-P, sedoheptulose 7-phosphate; TA-C3, the three-carbon substituted enzyme intermediate in the reaction catalyzed byTA; TK-C2, the two-carbon substituted enzyme intermediate in thereaction catalyzed by TK; Xlu-5-P, xylulose 5-phosphate.

Masakapalli et al.

604 Plant Physiol. Vol. 152, 2010 www.plantphysiol.orgon May 11, 2020 - Published by Downloaded from



was deduced from the labeling of starch and fromamino acids that are known to be synthesized exclu-sively in the plastid (Cys, Gly, His, Ile, Leu, Lys, Phe,Ser, Trp, Tyr, and Val). Note that although Ser, follow-ing synthesis in the plastid (Ho and Saito, 2001), can bereversibly converted to Gly via Ser hydroxymethyl-transferase, the fact that this can potentially occur inseveral compartments has no effect on the interpreta-tion of the Ser labeling pattern. Similarly, the labelingof cytosolic intermediates was deduced from the la-beling of Suc and Ala (Miyashita et al., 2007), andArg, g-aminobutyrate (GABA), Glu, Gln, and Prowere assumed to be derived from mitochondrial2-oxoglutarate. In contrast, itwasnot possible todeduceinformation about the subcellular labeling of oxaloace-tate and malate, and the synthesis of these metaboliteswas not assigned to a specific compartment.Second, because the model describes the redistribu-

tion of label through the network, expected biochem-

ical features of the network may be concealed oromitted. For example, it is possible to synthesize Glyfrom Ser using Thr aldolase (Joshi et al., 2006b), butusing this as the sole route to Gly leads to a substan-tially worse fit to the data, and adding the Thr aldolasereaction as an extra component of the model in Figure2 leads to no change in the flux estimates (data notshown). Thus, there is no evidence from the labelingdata to justify the inclusion of the Thr aldolase reactionin the model. Theoretical analysis also has a bearingon the construction of the model. For example, fluxthrough the GABA shunt is combined with theflux through the tricarboxylic acid cycle between2-oxoglutarate and fumarate because these pathwayshave identical labeling outcomes in the model. Simi-larly, Suc cycling is omitted because this can only becorrectly analyzed in steady-state MFA with informa-tion on the labeling of the cytosolic and vacuolar Glcpools (Kruger et al., 2007b). The flux supported by the

Figure 2. Metabolic model of central carbon metabolism with the oxidative steps of the plastidic oxidative PPP duplicated in thecytosol. Output fluxes used to constrain the model are shown with dotted black arrows; net and exchange fluxes deduced fromthe 13C label redistribution are shown with solid black unidirectional and bidirectional arrows, respectively. The direction of aunidirectional arrow corresponds to a positive flux in the output of the model, and the dotted gray boxes indicate that theenclosed metabolites were considered as a single pool. Standard abbreviations are used for the amino acids. C3-P, Three-carbonphosphate ester pool; cit, citrate; CO2in, internal carbon dioxide; fum, fumarate; aKG, 2-oxoglutarate; OAA, oxaloacetate; PEP,phosphoenolpyruvate; TAG, triacylglycerol.





putative Glc-6-phosphatase (Alonso et al., 2005) isomitted for the same reason, because there is no easyway to measure the labeling of the cytosolic andvacuolar Glc pools, and ignoring the compartmenta-tion of Glc is likely to produce very misleading fluxvalues (Kruger et al., 2007b). In contrast, discretecytosolic and plastidic pools of Glc-6-P were retainedin the model because the glucosyl units of Suc andstarch differed appreciably in their patterns of label-ing, establishing that they were synthesized fromisotopically distinct precursors (Supplemental Fig.S2, A–C; Supplemental Table S7). Subsequent in silicoanalysis established that the calculated flux distribu-tion would produce differential labeling in products

derived directly from cytosolic and plastidic hexosephosphate pools (Supplemental Fig. S2, D–F).

Third, the steps catalyzed by transketolase (TK) andtransaldolase (TA) in the nonoxidative steps of the PPPare represented in terms of half-reactions to allow forthe underlying mechanism of the enzymes and theirlack of specificity (Kruger and von Schaewen, 2003;Kleijn et al., 2005; Selivanov et al., 2005). Conventionalstoichiometric models of the nonoxidative reactions(Fig. 3A) assume that the label is directed through asubset of the possible reactions and thus misrepresentthe underlying fluxes. Adding the extra reactions foran equivalent formulation of the pathway (Fig. 3B)leads to ambiguity in the flux solution (Fig. 3D), but

Figure 3. Representation of the oxidative PPP for steady-state MFA. A and B, Conventional (A) and alternative (B) stoichiometriesfor the nonoxidative reactions of the PPP reflecting the substrate specificity of TK. C, Formulation of the nonoxidative reactions ofthe PPP in terms of the two- and three-carbon substituted enzyme intermediates formed during the ping-pong mechanisms of TKand TA. D, Steady-state analysis of the fluxes through the reactions catalyzed by TK for a simulated set of 13C-labeling data.Isotopomer abundances were first deduced for a flux distribution corresponding to the conventional network in A, and then theseabundances were used to predict the flux distribution through a metabolic network that permitted flux through both theconventional (A) and alternative (B) pathways. The 100 solutions each correspond to a different proportion of the three exchangereactions catalyzed by TK: triangles, Xlu-5-P + Rib-5-P 4 Tri-3-P + Sed-7-P; circles, Xlu-5-P + Ery-4-P 4 Tri-3-P + Fru-6-P;squares, Sed-7-P + Ery-4-P4 Rib-5-P + Fru-6-P. E, The result of the same exercise performed with the network (C) defined usinghalf reactions for TK: circles, Xlu-5-P4 Tri-3-P + Enz-C2; squares, Fru-6-P4 Ery-4-P + Enz-C2; triangles, Sed-7-P4 Rib-5-P +Enz-C2. Using half-reactions removes the ambiguity apparent in D and leads to a unique solution for the flux distribution (E). Allflux solutions were obtained using the metabolic models defined in Supplemental File S3.

Masakapalli et al.




this can be avoided by using the half-reactions basedon the ping-pong mechanisms of TK and TA (Fig. 3, Cand E). Theoretical analysis has shown that this for-mulation is essential for a correct description of theredistribution of label and that it allows the investiga-tion of parallel pathways supported by different TKisozymes (Kleijn et al., 2005). Accordingly, the half-reaction scheme is used here to maximize the possi-bility of resolving cytosolic and plastidic contributionsto the PPP fluxes.

Data Handling and Mathematical Modeling

Heterotrophic Arabidopsis cell cultures were incu-bated with [1-13C]Glc, [2- 13C]Glc, or 10% [U-13C6]Glc,the label redistribution at isotopic and meta-bolic steady state was quantified by 13C-NMR, andflux solutions, constrained by measurements of Glcconsumption and biosynthetic outputs (SupplementalTables S1–S4), were calculated in 13C-FLUX using theprotocol summarized in Figures 4 and 5. Label wasroutinely measured in soluble metabolites, includingamino acids, citrate, fumarate, GABA, malate, and Suc,in the amino acids derived from protein hydrolysateand in the Glc derived from starch. The best fit fluxsolutions presented here were based largely on Suc,starch, and protein hydrolysate labeling data, and fulldetails of the measurements used in the fitting processare given in Supplemental File S2. Measurement errorswere generally set at 5% for amino acid data and 15%for other metabolites, but a few measurements wereassigned larger deviations as specified in Supplemen-tal File S2.Label measurements for a particular metabolite were

initially grouped together for the fitting process, butsubsequently it was established that better fits (lower

residua) could be obtained by treating 13C peaks withmultiplet structures as separate subgroups. Simulationsin 13C-FLUX showed that the reduction in the resid-uum could be attributed to the inclusion of additionalscaling factors, compensating for discrepancies in therelative intensities of the signals from different carbonatoms within a metabolite, and that the subgroupingprocedure did not compromise the accuracy of fluxestimation. Subgrouping multiplet signals avoids theneed to establish the relative intensities of multipletsfrom the samemetabolite, and it is used routinely whentwo-dimensional NMR methods, which often mask thetrue relative intensities, are used to analyze [U-13C6]Glclabeling experiments (Sriram et al., 2004).

Each of the three labeling experiments was per-formed twice, and various combinations of experi-ments were analyzed simultaneously in 13C-FLUX bysetting up models in which the reaction network wasreplicated two, three, or six times. Software limita-tions, arising from the large size of the network and thequantity of data, placed some restriction on thisapproach, but a three-substrate model, comprising asubnetwork for each labeled precursor (i.e. [1-13C]Glc,[2-13C]Glc, or 10% [U-13C6]Glc) and two data sets foreach subnetwork, worked well, and a single fluxsolution corresponding to the entire data set could begenerated by constraining equivalent fluxes in thethree subnetworks to be equal in the model definitionfile. Ultimately, this procedure allowed a more reliabledefinition of the fluxes in the network than could beobtained by analyzing different labeling experimentsseparately (see below).

Preliminary fits of the data to the model were usedto refine the model and to identify errors in thelabeling data. In this iterative process, 13C-FLUX

Figure 4. Iterative procedure for determining metabolic fluxes fromsteady-state stable isotope labeling experiments.

Figure 5. Procedure for determining the global best fit for the modeland the optimal flux values.





was used to generate 50 to 100 solutions for a givenmodel and data set. Changes were then made to eitherthe model or the data set, for example, the addition of apathway or the correction of an incorrectly annotateddata point, and the fitting process was repeated. Com-paring the residua for the five best fits provided thebasis for accepting or rejecting a change to the modelor the data set. For example, the reaction catalyzed byisocitrate dehydrogenase, and the mitochondrial up-take of pyruvate, were both defined as unidirectionalbecause making them reversible had no effect on thefit. Similarly, CO2 uptake at natural abundance wasremoved from the model because it had no discernibleeffect on the fit. In contrast, adding unidirectionalplastidic pyruvate uptake improved the fit, as did theinclusion of separate cytosolic and plastidic pools forphosphoenolpyruvate and triose phosphate. However,the latter metabolites were combined to create a singlepool of three-carbon phosphate esters in each com-partment, since this removed indeterminable fluxesfrom the network and had no effect on the fit. Theiteration process led to the model in Figure 2 and thevalidated data set in Supplemental File S2, thus pro-viding a secure foundation for testing the extent towhich the method could reveal the subcellular com-partmentation of carbohydrate oxidation (Fig. 1).

Best Fit Fluxes and Statistical Analysis

The six data sets, comprising over 700 positionalisotopomer measurements and 29 biomass fluxes,were initially analyzed in a three-substrate modelbased on the flux network in Figure 2. Estimates forthe free fluxes were obtained by running the optimi-zation routine 1,000 times, generating a correspond-ingly large set of flux solutions, each with a set ofpredicted labeling measurements. Some of these solu-tions failed to satisfy constraints in the model andwere identified as infeasible solutions by the software;but the majority were feasible solutions, and thesewere extracted from the 13C-FLUX output file with acustomized software tool (available from the authorson request). The residua for the feasible solutionsvaried over a wide range (Fig. 6A), and a subset of thesolutions corresponding to residua below 3,000 wasselected for further analysis. This cutoff was justifiedboth by visual inspection of the distribution of residua(Fig. 6A) and by principal component analysis of themean-centered, unit variance-scaled flux solutions(Fig. 6B). The one-dimensional scores plot correspond-ing to the first principal component confirmed thesimilarity between the selected feasible solutions andprovided further justification for rejecting the solu-tions with residua above the cutoff.

Monte Carlo simulation with bootstrap sampling ofisotopomer measurements provides an effectivemethod for characterizing the size and shape of theflux space (Wiback et al., 2004; Joshi et al., 2006a), andthe selected solutions showed considerable variabilityin the definition of the free fluxes (Supplemental Fig.

S3). Some fluxes were tightly defined, for example,many of those in the tricarboxylic acid cycle, whileothers were normally distributed over a wider rangeof values. To reduce the hundreds of selected feasiblesolutions to a global best fit, the mean values from theMonte Carlo flux space were used as starting free fluxvalues for a final run of the optimizer with the vali-dated data set. This run was performed without boot-strapMonte Carlo sampling, and it resulted in a globalbest fit solution with the lowest residuum (Table II).These procedures were repeated for the models inwhich the full PPP was either confined to the plastids

Figure 6. Establishing the threshold for the best fit flux solutions used todefine the subset of Monte Carlo flux space for calculating the globalbest fit (Fig. 5). A, Histogram showing the frequency distribution for theresidua of the feasible solutions. The inset shows an expansion of theregion covering the lowest residua, with an arrow to indicatethe threshold for inclusion in the subsequent analysis. B, Distributionof scores for the first component following principal componentanalysis of the flux solutions from 1,000 Monte Carlo simulations.Feasible flux solutions with less than 3,000 residua (black circles) forma distinct cluster that is well resolved from those with more than 3,000residua (white circles). The latter have a more scattered distribution thatis similar to that obtained from random combinations of fluxes thatproduce infeasible flux maps (crosses).

Masakapalli et al.




Table II. Influence of metabolic network structure on determination of flux estimates

Fluxes were determined by fitting isotopomer abundances and biosynthetic output estimates to a familyof models differing in the extent of PPP compartmentation (Figs. 1 and 2). Each model contained acomplete plastidic PPP pathway. Free fluxes that varied during the fitting procedure are indicated inboldface. Net fluxes and normalized hyperbolic transformed exchange fluxes are the global best fitestimate 6 SD as determined by EstimateStat and are calculated relative to the rate of Glc uptake (set tounity). A normalized exchange flux for which SD takes the estimated value outside the permissible [0,1]range is considered indeterminate (Ind).

Flux NameFlux (Relative to Glc Uptake)

No Cytosolic PPP Cytosolic Oxidative PPP Steps Only Complete Cytosolic PPP

Net fluxchex1 0.317 6 0.062 0.191 6 0.083 0.665 6 0.074chex2 0.241 6 0.062 0.115 6 0.083 0.365 6 0.187chex3 0.778 6 0.016 0.786 6 0.021 0.894 6 0.040phex1 0.049 6 0.067 0.209 6 0.101 20.160 6 0.090phex2 0.305 6 0.063 0.442 6 0.088 0.228 6 0.187phex3 0.072 6 0.016 0.075 6 0.020 0.000 6 0.040cppp1 2 0.372 6 0.028 0.266 6 0.019cppp2a 2 2 20.224 6 0.200cppp2b 2 2 20.112 6 0.100cppp2c 2 2 20.112 6 0.100cppp3a 2 2 20.112 6 0.100cppp3b 2 2 20.112 6 0.100pppp1 0.405 6 0.018 0.000 6 0.039 0.000 6 0.027pppp2a 0.138 6 0.006 0.127 6 0.009 0.204 6 0.102pppp2b 0.118 6 0.006 0.107 6 0.009 0.184 6 0.102pppp2c 0.256 6 0.012 0.234 6 0.018 0.388 6 0.204pppp3a 0.138 6 0.006 0.127 6 0.009 0.204 6 0.102pppp3b 0.138 6 0.006 0.127 6 0.009 0.204 6 0.102tca1 0.650 6 0.006 0.661 6 0.009 0.696 6 0.009tca2 0.650 6 0.006 0.661 6 0.009 0.696 6 0.009tca3 0.624 6 0.006 0.635 6 0.009 0.670 6 0.009tca4 0.572 6 0.006 0.583 6 0.009 0.618 6 0.009tca5 0.573 6 0.006 0.584 6 0.009 0.619 6 0.009ana1 0.241 6 0.004 0.241 6 0.004 0.243 6 0.005ana2 0.039 6 0.003 0.040 6 0.004 0.039 6 0.004ana3 0.026 6 0.001 0.026 6 0.001 0.029 6 0.003cmex 0.610 6 0.007 0.621 6 0.009 0.657 6 0.009cpex 0.112 6 0.015 0.109 6 0.019 0.181 6 0.037gpt 0.496 6 0.062 0.251 6 0.088 20.118 6 0.082tpt/ppt 0.540 6 0.124 0.801 6 0.171 0.523 6 0.471xpt 2 2 0.603 6 0.309

Normalized exchange fluxchex1 0.990 6 0.102 0.990 6 0.108 0.423 6 0.066chex2 0.682 6 0.055 0.667 6 0.053 0.803 6 0.073phex1 0.476 6 0.040 0.475 6 0.054 0.148 6 0.103cppp2a 2 2 0.990 6 0.276cppp2b 2 2 0.320 6 0.082cppp2c 2 2 Indcppp3a 2 2 Indcppp3b 2 2 0.000 6 0.194pppp2a Ind Ind Indpppp2b 0.990 6 0.141 0.990 6 0.166 0.990 6 0.588pppp2c 0.990 6 0.120 0.990 6 0.115 0.990 6 0.502pppp3a Ind Ind Indpppp3b 0.268 6 0.039 0.299 6 0.059 0.252 6 0.123tca5 0.719 6 0.043 0.719 6 0.043 0.730 6 0.044gpt 0.970 6 0.051 0.695 6 0.067 0.462 6 0.096tpt/ppt 0.227 6 0.140 0.441 6 0.115 0.000 6 1.424xpt 2 2 0.309 6 0.362





(Fig. 1A) or present in both compartments (Fig. 1C),and the final best fit fluxes for the three reactionnetworks are given in Table II.

Statistically reliable flux values were generated forall three models of the PPP (Table II) with only minordifferences in residuum. The flux distribution for thesteps involving sugar phosphates differed substan-tially for the three models (Fig. 7), and it was notice-able that the bulk of the flux through the oxidativesteps of the PPP switched to the cytosolic pathwaywhen this option was included in the network struc-ture. Thus, while the flux through the oxidative stepswas necessarily confined to the plastids in the absenceof a cytosolic pathway (Figs. 1A and 7A), the plastidicpathway made only a minor contribution when thecytosolic pathway was included in the model (Figs. 1,B and C, and 7, B and C). The degrees of freedom in themodels increased with the number of fluxes, and asexpected, this led to a small improvement in theresiduum achieved in the fitting procedure and anincrease in the SD values for the flux estimates. Forexample, including the full PPP in both cytosol andplastids (Fig. 1C) produced a marginal improvementin residuum, but it produced a flux map in which 28 ofthe 33 fluxes shared by all three models had the largestSD values.

The validity of the flux maps was tested bycomparing the predicted contribution of the carbonatoms of metabolized Glc with the relative specificactivity of 14CO2 released from specifically labeled Glc(Fig. 8). Ratios of 14CO2 labeling from [1-14C]-, [2-14C]-,[3,4-14C]-, and [6-14C]Glc are typically used to charac-terize carbohydrate oxidation (Harrison and Kruger,2008), and the pattern of 14CO2 release predicted by theglobal best fit flux estimates for each model was notsignificantly different from the experimentally deter-mined pattern (Fig. 8), implying that each of the fluxmaps provides an adequate explanation of the ob-served data for the release of 14CO2 from positionallylabeled Glc.

Since the models are indistinguishable, it can beconcluded that the subcellular compartmentation ofthe PPP cannot be determined unambiguously fromthe available labeling data alone. However, based oncurrent biochemical evidence, the most likely model isthe one in which the oxidative steps of the PPP areduplicated in the cytosol and the plastids (Figs. 1B and7B; Kruger and von Schaewen, 2003). The global bestfit solution for this network is presented in detail inSupplemental Table S5, and it provides good agree-ment between the observed labeling data and theisotopomer abundances predicted from the global bestfit flux map (Fig. 9A). There is also excellent corre-spondence between the measured, scaled, and pre-dicted relative fractional abundance of positionalisotopomers for individual metabolites, as exemplifiedby the data on protein-derived Asp/Asn (Fig. 9B).More than 80% of the total residuum was confined to176 measurements (25% of total), and the flux mapobtained after removal of these values was essentially

identical to that obtained from the complete data set,suggesting that the poorly fitting measurements didnot have a major influence on the flux solution (Sup-plemental Fig. S4). Furthermore, there was no con-sistency in the identities of the isotopomers that

Figure 7. Flux maps corresponding to three models for the subcellularcompartmentation of the oxidative PPP. A, Oxidative and nonoxidativereactions of the PPP confined to the plastid. B, Oxidative steps of thePPP in both the cytosol and plastid, with the nonoxidative steps in theplastid. C, The complete pathway in both compartments. The width ofthe arrows is proportional to the net molar flux for each step as listed inTable II.

Masakapalli et al.




contributed most to the residua between duplicatefeeding studies involving identically labeled sub-strates, implying that the discrepancies between theobserved and predicted values were due to poorestimates of measurement error rather than to a failureof the model to account adequately for the pattern oflabel redistribution in a particular section of the net-work.Finally, in silico analysis of the chosen network

revealed that the reliability of the flux estimates im-proved by combining data from labeling studies using[1-13C]-, [2-13C]-, and [U-13C6]Glc. This was particu-larly noticeable for the oxidative steps of the PPP in thecytosol and the plastids, and it also emphasized theimportance of using well-defined labeling measure-ments (Fig. 10).

DISCUSSION

Choice of Experimental System and Application ofSteady-State MFA

A plant cell suspension was chosen for the investi-gation to simplify the task of testing different models

of the subcellular compartmentation of the PPP. TheArabidopsis cell suspension is experimentally tracta-ble, it has been used successfully for MFA (Williamset al., 2008), and it avoids the difficulties that mightarise from multiple cell types in differentiated tissues.Moreover, many of the general metabolic featuresof the chosen Arabidopsis culture are similar to

Figure 8. Validation of Arabidopsis flux map by radiorespirometricanalysis. The ratios of the specific activity of 14CO2 released from cellcultures supplied with [14C]Glc specifically labeled in the carbonpositions indicated were compared with those predicted from the fluxmaps. The latter were derived from the specific abundance of label inCO2 determined by simulations in CumoNet using different Glcisotopomers as the input substrate. Experimental data are shown asmeans 6 SD of ratios from four replicate measurements for each of twoindependent cell cultures (white bars). The predicted ratios weredetermined from the global best fit solutions (Table II; SupplementalTable S5) for the networks in which the PPP is confined to the plastid(model A; light gray bars), the oxidative steps of the PPP are located inboth the cytosol and plastids (model B; dark gray bars), or the entirepathway is present in both compartments (model C; black bars). Thereis no significant difference in the experimentally determined ratios forthe two separately grown cultures as assessed by repeated-measuresANOVA, and all the predicted ratios are within the 95% predictionintervals of the ratios determined experimentally.

Figure 9. Assessment of goodness of fit of isotopomer data to the globalbest fit solution of the metabolic network defined in Figure 2. Datafrom cell cultures labeled separately with [1-13C]Glc, [2-13C]Glc, and[U-13C6]Glc were used in these evaluations. A, Comparison of pre-dicted and measured isotopomer abundances for all metabolites ana-lyzed. B, Comparison of the predicted relative fractional abundance ofisotopomers of Asp with measured values, before and after scaling ofsubgroups of 13C peaks with multiplet structure. Shading is conservedfor individual isotopomers in the pie charts for a given labeledsubstrate.





the heterotrophic metabolism that occurs in roots(Lehmann et al., 2009).

The steady-state MFA protocol used here wasbroadly similar to existing best practice, but the appli-cation of this tool in plant biology is still developing,and several new features were introduced with theaim of further improving the precision and reliabilityof the flux estimates. First, half-reactions were used torepresent the underlying carbon transitions that occurduring the steps catalyzed by TA and TK. This ap-proach, which provides a more accurate description ofthe redistribution of the label by the PPP, has not, toour knowledge, been used previously in steady-stateMFA of plants and has only rarely been used inmicrobial work. Second, Monte Carlo simulationswith bootstrap sampling of the isotopomer abun-dances provided an effective method for exploringthe flux space that was consistent with the observedredistribution of label. Stochastic sampling of thestarting fluxes is routine in the modeling process(Schwender et al., 2006), and combining this withstatistical sampling of the data points improves thefitting procedure (Kelly et al., 1990; Namba, 2004).Surprisingly, the availability of a subroutine to exploitthis method in the commonly used 13C-FLUX soft-ware appears to have been overlooked. Third, theoutput from the Monte Carlo simulations was sub-jected to principal component analysis to identify theset of best fit feasible solutions. This novel approachprovided an objective method for analyzing the out-puts of the modeling process and an effective route tothe global best fit. Finally, redistribution of radioactiv-ity from U-14C-substrate was used as a convenient

method for ensuring that all major metabolic productshad been identified and as a sensitive method fordetermining the output fluxes.

Quantifying the Compartmented Fluxes ofCentral Metabolism

The subcellular compartmentation of metabolism ineukaryotes complicates steady-state MFA (Roscheret al., 2000), and the problem is acute in plants becauseof the more extensive duplication of steps and path-ways in the metabolic network (Kruger and Ratcliffe,2008). Capturing the characteristic compartmentationof the pathways of carbohydrate oxidation in a real-istic model of heterotrophic plant metabolism exem-plifies the problem, and it would be particularly usefulif this could be done by comparing models for net-works compartmented to different extents, since thiswould allow an in vivo assessment of the functionalcompartmentation of the network. Since most steady-state MFA models are based on overdetermined datasets, reflecting the wealth of stable isotope labelinginformation that can be obtained from NMR or massspectrometry (Ratcliffe and Shachar-Hill, 2006), it iseasy to add extra steps to the model to allow it tomatch biochemical reality more closely. However, therelationship between fluxes and measurements is notalways obvious, and even if the extra fluxes arepotentially resolvable, they may not be determinablewith any great confidence.

In practice, many steady-state MFA investigations ofthe PPP in plants use models in which the compart-mentation of the pathway is predefined. For example,

Figure 10. Statistical analysis of flux estimates obtained with different steady-state labeling strategies. The SD values of optimumflux estimates were derived from the global best fit solution of the metabolic network defined in Figure 2 using the EstimateStatroutine in 13C-FLUX. Isotopomer measurements obtained from analysis of extracts of cell cultures labeled separately with[1-13C]Glc (1-13C), [2-13C]Glc (2-13C), or [U-13C6]Glc (U-13C) were used independently or in combination. Fluxes wereexpressed relative to the rate of Glc consumption (set to unity), and SD values are coded from black (low; flux well defined) towhite (high; flux poorly defined), as indicated by the logarithmic gray scale at the top. The bottom row contains an analysis ofdata combined from all three feeding strategies (1/2/U-13C), in which the SD of all isotopomer measurements was reduced to 5%.

Masakapalli et al.




a single plastidic pathway has been assumed in maizeroot tips (Dieuaide-Noubhani et al., 1995; Alonso et al.,2007b), tomato (Solanum lycopersicum) cell cultures(Rontein et al., 2002), and Arabidopsis cell cultures(Williams et al., 2008), whereas complete duplicationof the pathway in the cytosol and plastid has beenassumed in the analysis of cultured soybean embryos(Sriram et al., 2004; Iyer et al., 2008). In some tissues,such as oilseed rape embryos (Schwender et al., 2003)and soybean embryos (Allen et al., 2009), the modelingtask is made easier because the observed labelingpatterns imply that intermediates in glycolysis and thePPP are in fast exchange between the cytosol and theplastids, ensuring that duplicate pathways in differentphysical locations function indistinguishably from asingle uncompartmented pathway. However, in mostof the tissues that have been examined so far, includ-ing soybean embryos in some instances (Sriram et al.,2004), the hexose phosphate pools in the cytosol andthe plastids are found to be labeled to different extents.Typically, this evidence might be based on differencesin the labeling patterns of the glucosyl moieties fromSuc and starch (Supplemental Fig. S2), and it is thennecessary to consider the subcellular compartmenta-tion of glycolysis and the oxidative PPP explicitly.Inspection of the labeling patterns in products de-

rived specifically from either the cytosolic or plastidichexose phosphate pool may provide circumstantialevidence for the operation of the full PPP in bothcompartments (Krook et al., 1999), justifying the con-struction of a model with duplicate pathways. How-ever, intuitive interpretation of the data is complicatedby the multiple metabolic processes that contribute tothe labeling of these metabolites (Kruger et al., 2003),and a more robust quantitative approach is needed tocompare the results of fitting all the labeling datato models in which the pathways are duplicated todifferent extents. Here, steady-state MFA was used tocompare three different models of the oxidative PPP(Fig. 1), and the analysis indicates that several aspectsof the protocol contribute to its effective implementa-tion. First, it is necessary to work with overdeterminedmodels so that the statistical significance of the fluxsolutions can be properly evaluated. Second, for thePPP, it is important to define the carbon transitionnetwork in terms of the half-reactions for TA and TK tomaximize the chance of being able to discriminatebetween parallel events in the cytosol and plastid.Third, it is essential to use high-quality data fromexperiments with several differently labeled sub-strates in order to deduce fluxes with sufficient con-fidence to allow a comparison of the models (Fig. 10).Finally, it is necessary to establish a procedure (Figs. 4and 5) that leads to an optimal best fit based on clearlydefined criteria so that the results for different modelscan be compared.Analysis of three models for the oxidative PPP (Fig.

1) emphasized the difficulty of determining the fluxdistribution with sufficient confidence to be able todiscriminate between the models. There was little to

distinguish a plastid-only model (Fig. 1A) from themodel in which the oxidative steps were duplicated inthe cytosol (Fig. 1B), even though the latter is moreplausible on the basis of molecular evidence. Theincrease in the degrees of freedom in the second model(Fig. 1B) produced only a marginal improvement inthe residuum obtained for the global best fit, and themain observations that support the model are thenonrandom allocation of flux between the oxidativesteps in the two compartments (Supplemental Fig. S3)and the reliability of these flux estimates, as reflectedin their confidence intervals (Table II). Adding furtherdegrees of freedom by duplicating the complete oxi-dative PPP in the two compartments (Fig. 1C) led to afurther, modest improvement in the fit and a radicallydifferent flux distribution through the pathways ofcarbohydrate oxidation (Table II), but it also reducedthe confidence in the flux values, with most of thefluxes showing an increase in SD. Thus, this modelprovides a less reliable flux map and can be rejected onthe basis that the fit to the available data is lessconvincing. Such a conclusion is always subject tothe proviso that data might become available thatwould put further constraints on the assessment ofthe compartmentation of the pathway (Kruger andRatcliffe, 2009), but currently, the available data implyonly a plastidic location for the nonoxidative steps ofthe PPP in Arabidopsis cells.

In silico analysis of the sensitivity of cumomer abun-dance to changes in flux through the irreversible oxi-dative section of the PPP can identify the isotopomermeasurements that are best able to define the fluxesthrough this step in the cytosol and plastids (Supple-mental File S4). Although the sensitivities are influ-enced by the label distribution in the supplied substrateand by the flux distribution through the network,estimates based on the biochemically preferred fluxmap (Fig. 7B) indicate that isotopomers of cytosolic andplastidic hexose phosphates and compounds deriveddirectly from these pools are among the most diagnos-tic measures of these PPP fluxes. This discriminationarises in part because, in the absence of plastidicFru-1,6-bisphosphatase, cytosolic Glc-6-P is the onlysubstrate replenishing the plastidic Glc-6-P pool, andany difference in labeling must therefore stem fromplastidic PPP activity. However, intermediates withinthe reversible nonoxidative section of the PPP, and theirderivatives such as His and aromatic amino acids, arealso relatively sensitive to perturbations in PPP flux.Thus, measurements of label distribution in these com-pounds would be expected to further improve the fluxestimates and could lead to greater discriminationbetween the different models.

A similar, but less comprehensive, analysis of thecompartmentation of metabolism in sunflower em-bryos found that models of the PPP that includedcytosolic steps were no improvement over a model inwhich the PPP was confined to the plastids (Alonsoet al., 2007a). However, the flux distributions in thealternative models were not discussed, and it is per-





haps surprising that the additional degrees of freedomin the models with cytosolic steps led to no improve-ment in the fit. Irrespective of these points, since all themodels gave effectively the same fit to the data, it isunclear on what basis one of them was preferred.

The difficulties encountered in analyzing the com-partmentation of the oxidative PPP by steady-stateMFA highlight some of the practical limitations of theapproach. The objective is to capture a realistic met-abolic phenotype, but in practice the flux map is arepresentation of the movement of the chosen labelthrough the system. Resolving the contribution ofchemically distinct pathways between two metabo-lites in the same compartment, or of duplicated stepsand pathways in distinct compartments, is not alwayspossible or even attempted. For example, the stepsfrom citrate to 2-oxoglutarate are routinely assignedto the mitochondrion in steady-state MFA, eventhough there may be parallel fluxes in the cytosol.Thus, if a function of MFA is to guide metabolicengineering (Kruger and Ratcliffe, 2008; Libourel andShachar-Hill, 2008), then current flux maps areunlikely to be useful in predicting the result ofmanipulating a compartment-specific enzyme activ-ity. Similarly, even if MFA provides evidence that aphysically compartmented pathway is functionallyuncompartmented as a result of fast exchange ofintermediates (Schwender et al., 2003; Allen et al.,2009), the resulting metabolic phenotype still con-ceals the associated compartmental demands forreducing/oxidizing power and energy. It can be con-cluded that it is essential to push steady-state MFAto its limits to obtain the most informative modelfor the flux distribution.

Metabolism in a Heterotrophic ArabidopsisCell Suspension

The network structure used to determine the bio-chemically preferred flux map (Fig. 7B) differs fromthe one used in the previous analysis of Arabidopsiscell suspension cultures (Williams et al., 2008) in twomain ways. First, the models differ in their treatmentof the lower section of glycolysis. In steady-state MFAof plant metabolism, there is usually insufficient la-beling information to support a model with separatecytosolic and plastidic pools of both glyceraldehyde3-phosphate and phosphoenolpyruvate. In the earliermodel (Williams et al., 2008), each metabolite wasassumed to be in a common uncompartmented pool,implying fast exchange across the plastid envelope,whereas here it was assumed that the two metaboliteswere at isotopic equilibrium in a three-carbon phos-phate ester pool (Fig. 2) within each compartment.This assumption allowed the degree of compartmen-tation of the pools to be tested by not arbitrarilyassuming fast exchange between the cytosol and plas-tid, and it led to a marked alteration in the fluxdistribution, since all three flux maps showed a sig-nificant net flux from the plastid to the cytosol (Fig. 7)

at the level of the three-carbon phosphate ester pooland only a limited exchange flux (Table II). Second, thebiochemically preferred model that emerged from thecomparison of three different models for the oxidativePPP (Fig. 1) allowed the oxidative steps to operate inboth the cytosol and plastids. This compartmentationis supported by estimates of transcript abundancebased on publicly available microarray data for genesencoding isozymes of the PPP (Supplemental Fig. S5A)and associated plastid envelope translocators (Supple-mental Fig. S5B).

A striking feature of the flux map generated fromthe biochemically preferred model is that the oxida-tive steps of the PPP are almost completely confinedto the cytosol. Since most biosynthetic processesrequiring NADPH are plastidic, this flux distributionimplies extensive exchange of reducing power equiv-alents across the plastid envelope. However, this isunlikely to be problematic, since previous studies onmaize root tips suggest that the oxidative PPP in thecytosol and plastids can cooperate in providingNADPH for biosynthesis throughout the cell (Averillet al., 1998). Nevertheless, the dominance of thecytosolic pathway in catalyzing flux through theoxidative section of the PPP is unexpected, and it isunclear whether this is a general characteristic of theorganization of carbohydrate metabolism in higherplants, since compartmentation of the PPP is notincluded in most flux maps. The only other planttissues for which information is available are rosyperiwinkle (Catharanthus roseus) hairy root culturesand developing soybean embryos, and in both ofthese only 25% to 50% of flux through the oxidativesection of the PPP is cytosolic (Sriram et al., 2004,2007; Iyer et al., 2008). However, in these systems,total flux through the oxidative PPP is considerablyhigher than that observed in Arabidopsis cells, andwhen expressed as a proportion of the rate of sugarconsumption, the cytosolic fluxes in all three systemsare similar (rosy periwinkle, 42.6%; soybean, 47.4%;Arabidopsis, 37.2%), implying that variation in thecompartmentation of PPP flux is due to fluctuationsin plastidic activities rather than to a decrease in thesignificance of the cytosolic flux. Moreover, irrespec-tive of the exact proportion of the oxidative PPP fluxoccurring in the cytosol, two recent studies demon-strate the importance of the cytosolic activity. First,disrupting the expression of the two genes encodingcytosolic isozymes of Glc-6-P dehydrogenase (G6PD5and G6PD6) in Arabidopsis through insertional inac-tivation alters seed lipid content (Wakao et al., 2008).Second, supplementation or replacement of cytosolicGlc-6-P dehydrogenase with an alternative isoformimproves the resistance of tobacco (Nicotiana tabacum)leaves to Phytophthora infestans, and this is attributedto changes in the provision of NADPH and stimula-tion of the hypersensitive defense response (Scharteet al., 2009). Both observations highlight a require-ment for the oxidative section of the PPP in thecytosol to support normal metabolic activity.

Masakapalli et al.




Energy and Redox Balance in Plant Metabolic Networks

The metabolic activity of the Arabidopsis cell cul-ture appears to be broadly representative of a widerange of heterotrophic plant systems, and one measureof this can be found in its efficiency in convertingsugar to biomass components. The proportion of the[U-14C]Glc metabolized to macromolecules and solu-ble compounds (Supplemental Table S1) indicates acarbon use efficiency of 60%, in good agreement withthe 65% calculated from a comparison of substrateuptake and estimated CO2 output in the previousstudy of Arabidopsis cells grown under equivalentconditions (Williams et al., 2008). These estimates ofcarbon use efficiency are within the range obtained forother heterotrophic plant systems, including a rosyperiwinkle hairy root culture (24%; Sriram et al., 2007),maize root tips (42%–47%; Alonso et al., 2007b), devel-oping sunflower embryos (50%; Alonso et al., 2007a),and a tomato cell culture (52%–68%; Rontein et al.,2002). Although considerably higher values have beenreported for developing embryos of oilseed rape (85%–95%) and soybean (82%–83%), in these systems there isappreciable reassimilation of respired CO2 throughRubisco, and light makes significant contributions tothe provision of ATP and/or reductant required forbiosynthesis (Goffman et al., 2005; Allen et al., 2009). Incontrast, the energy and reducing power required forbiosynthesis to support growth in heterotrophic Arabi-dopsis cells must be generated exclusively by oxidationof Glc, the sole respiratory substrate, through the cen-tral pathways of carbon metabolism.

Comparison of the oxidation state of the substratesin the culture medium and the products generated bythe Arabidopsis cells confirms that the metabolic net-work is able to operate without an input of reducingpower. The primary substrates for cell growth wereGlc, nitrate plus ammonium, and sulfate, in which theaverage oxidation states of carbon, nitrogen, andsulfur are 0, +1, and +6, respectively. The analysis inSupplemental File S5 shows average carbon oxidationstates of 0 for carbohydrates, 21.56 for lipid, +0.92 fororganic acids, and21.02 for protein and soluble aminoacids in which nitrogen and sulfur have oxidationstates of 23 and 22, respectively. The correspondingvalues for carbon in CO2 and ethanol, the principalrespiratory end products, are +4 and 22, respectively.It follows from the relative partitioning of substrateduring growth (Table I) that the average change inoxidation state in metabolic end products is +1.33 percarbon (Supplemental File S5). The analysis alsoshows that assimilation of nitrate used in amino acidproduction represents the greatest demand for reduc-ing power in the Arabidopsis cell culture under thegrowth conditions used in this study.

However, this simple redox analysis ignores thecoenzyme specificity of different processes and doesnot consider the preferential involvement of NADPHin many reductive biosynthetic steps. This limitationcan be overcome by using flux maps to evaluate thecoenzyme balance in the network. In contrast to thesituation in developing soybean embryos (Iyer et al.,2008), inspection of the fluxes determined for theArabidopsis cells establishes that the oxidative PPP

Table III. Coenzyme requirements for the biosynthetic outputs from the central metabolic network in heterotrophic Arabidopsis cells

The coenzyme requirements for biosynthesis were determined from the fluxes in Supplemental Table S4 using the reaction stoichiometries inSupplemental Table S6 and are presented as means 6 SE based on the estimates of label partitioning following metabolism of [U-14C]Glc in Table I.The coenzyme requirements for lipid assume the synthesis of triacylglycerol containing only oleoyl (C18:0) side chains. Costs for amino acidproduction are based on the assimilation of equal amounts of NO3

2 and NH4+. The ATP requirement for protein production is calculated from the

cost of synthesizing proteins from the component amino acids and is equivalent to 4 ATP per amino acid residue. Accumulating organic acids arewithdrawn directly from intermediate pools in the network and incur no additional cost. Generation of coenzymes by the central metabolic networkis determined from the net fluxes of the internal interconversions (i.e. excluding output steps) of the model defined in Figure 2 using conventionalreaction stoichiometries and ignoring the possible contribution of inorganic pyrophosphate as a phosphoryl donor. NADH and NADPH yields (initalics) assume conversion of isocitrate to 2-oxoglutarate by NADP+-dependent isocitrate dehydrogenase. ATP generation is attributed to substrate-level phosphorylation, and FADH2 production is assumed to be equivalent to 0.63 NADH. Values for coenzyme generation are means 6 SE ofestimates obtained from the best fit flux solutions of 501 Monte Carlo simulations; values in parentheses are determined from the global best fit fluxsolution of the network in which the oxidative steps of the PPP are located in both the cytosol and plastids, as detailed in Supplemental Table S5.

ProductCoenzyme Requirement (Molar Flux Relative to Glc Uptake)

ATP NADH NADPH

Suc 0.076 6 0.0039 2 2Starch 0.042 6 0.0070 2 2Cell wall 0.079 6 0.0112 20.039 6 0.0045 2Lipid 0.066 6 0.0045 0.068 6 0.0045 0.066 6 0.0045Ethanol 2 0.020 6 0.0023 2Amino acids 0.726 6 0.0090 0.144 6 0.0028 1.207 6 0.0141Protein 0.268 6 0.0285 2 2Total requirement 1.257 6 0.0332 0.193 6 0.0073 1.273 6 0.0148

Generated by the central metabolic network

0.999 6 0.0011 3.964 6 0.0024 0.771 6 0.00223.330 6 0.0021 1.407 6 0.0018

(1.001) (4.055) (0.770)(3.420) (1.405)





is insufficient to meet the NADPH requirements forbiosynthesis predicted by the network (Table III). Thisobservation is the result of two assumptions: (1) theassimilation of equal amounts of nitrate and ammo-nium (Table III); and (2) the NADPH dependence ofGlu synthesis (Supplemental Table S6). In this situa-tion, reductive biosynthesis will require other NADP+-dependent processes, and in the network consideredhere, the demand could be met by the operation ofNADP+-isocitrate dehydrogenase (Table III).

In contrast, the production of NADH is more thanenough for its utilization in reductive processes, in-cluding the demands for ATP synthesis. Indeed, if it isassumed that excess NADH (and FADH2) are reoxi-dized through the mitochondrial electron transportchain to produce ATP through oxidative phosphoryla-tion (at stoichiometries of 2.3–2.5 and 1.4–1.5 ATP perNADH and FADH2, respectively; Hinkle, 2005), thenthe maximum yield of ATP generated by the metabolicnetwork is over seven times that required to supportthe flux to biosynthetic products. The actual yield ofATP will depend on the extent of engagement of thealternative oxidase and the activity of mitochondrialuncoupling proteins, but the comparison suggests thatthe metabolic network has a considerable overcapacityfor ATP production. Of course, these calculationsignore the energy requirements of other metabolicprocesses, such as substrate cycling, turnover of pro-teins and other macromolecules, and synthesis andturnover of RNA, that are not included in thisnetwork. Moreover, ATP is used in other cellularprocesses, such as substrate uptake, transport of me-tabolites across intracellular membranes, and mainte-nance of transmembrane ion and electrical gradients.Such activities can account for over half of the cellularATP consumption in microorganisms (Nielsen et al.,2003), and the findings presented here suggest that aneven greater proportion of the potential chemicalenergy released during metabolism in actively grow-ing plant cells could be needed to meet these require-ments. Recently, analysis of energy demands in agenome-scale stoichiometric model of Arabidopsishas led to a similar conclusion that only a relativelysmall proportion of the potential ATP generated by thecell is used for biosynthesis (Poolman et al., 2009).

This conclusion has implications for the determi-nation of flux through plant metabolic networks byflux balance analysis. This approach depends on theselection of an appropriate objective function to ob-tain an optimum solution within the potential fluxspace defined by the stoichiometry of the biochemicalreaction network, plus other constraints, such as themetabolic outputs of the system (Kauffman et al.,2003). Typically, the optimization criterion in micro-organisms is maximization of yield of biomass, ATP,or a product of interest, or alternatively, maximiza-tion of flux (i.e. maximization of the rate of produc-tion of one or more components required for growth;Schuster et al., 2008). A recent study on developingendosperm of barley (Hordeum vulgare) used a refine-

ment of this approach in which the objective functionwas maximization of growth per unit of flux, which isa combination of maximization of biomass yield andminimization of overall flux (Grafahrend-Belau et al.,2009). The logic behind this optimization criterion isthat cellular metabolism has evolved to fulfill itsfunctions with the most economic use of availableresources (Holzhutter, 2004), but application of thiscriterion requires that the major energy-consumingprocesses have been adequately incorporated into themetabolic model. The demonstration that a largeproportion of the potential biochemical energy re-leased through respiration in heterotrophic plantcells is dissipated, or used by processes other thanthose directly associated with the synthesis of mac-romolecules and soluble components that accumu-late during growth, means that ATP expenditure forsuch purposes must be accurately assessed if fluxbalance analysis is to generate meaningful flux mapsof the network of central carbon metabolism inplants.

CONCLUSION

Steady-state MFA of multiple labeling experimentsin an Arabidopsis cell suspension shows that theexperimental observations can be explained by anyone of three different models of the subcellular com-partmentation of the oxidative PPP. Other biochemicalevidence favors the model in which the oxidative stepsare duplicated in the cytosol and plastids, and thisleads to the conclusion that the bulk of the fluxthrough the oxidative steps is cytosolic. This illustratesthe point that interesting features of the metabolicphenotype can be missed with an inappropriate modeland emphasizes the importance of basing the modelon all the available biochemical evidence.

MATERIALS AND METHODS

Cell Culture

Cell suspensions of Arabidopsis (Arabidopsis thaliana ecotype Landsberg

erecta) were maintained on an orbital shaker at 150 rpm in 250-mL Erlenmeyer

flasks sealed with a double layer of aluminum foil under a 16-h-light/8-h-dark

cycle at 22�C (Williams et al., 2008). Every 7 d, 15 mL of cell suspension was

subcultured into 100 mL of fresh Murashige and Skoog (MS) medium

supplemented with 3% (w/v) Glc, 0.5 mg L21 1-naphthaleneacetic acid, and

0.05 mg L21 kinetin. Heterotrophic cell suspensions were produced by

subculturing 15 mL of a 7-d-old light-grown cell suspension into 100 mL of

fresh medium and incubating in the dark at 22�C.

Chemicals

[1-14C]Glc (specific activity, 2.00 GBq mmol21), [6-14C]Glc (specific activity,

2.07 GBq mmol21), and [U-14C]Glc (specific activity, 10.4 GBq mmol21) were

purchased from GE Healthcare Life Sciences (http://www1.gelifesciences.

com/). [3,4-14C]Glc (specific activity, 1.11 GBq mmol21) was purchased from

American Radiolabeled Chemicals (http://www.arc-inc.com/), and [2-14C]

Glc (specific activity, 1.67 GBq mmol–1) was obtained from Perkin-Elmer NEN

(http://las.perkinelmer.com/). [2-13C]Glc (99 atom %) was purchased from

Omicron Biochemicals (http://www.omicronbio.com/), and [1-13C]Glc (99

Masakapalli et al.




atom%) and [U-13C6]Glc (99 atom%) were obtained from Isotec (http://www.

sigmaaldrich.com/chemistry/stable-isotopes-isotec.html). All enzymes were

from Roche Diagnostics (http://www.roche.com/). General chemicals and

chromatography resins were purchased from Sigma-Aldrich (http://www.

sigmaaldrich.com/) or Merck (http://www.merckbiosciences.co.uk/).

Incubation of Arabidopsis Cell Culture with13C-Labeled Glc

Cells were labeled by subculturing light-grown cells into medium in which

the unlabeled Glc was replaced with [1-13C]Glc, [2-13C]Glc, or amixture of 10%

[U-13C6]Glc and 90% unlabeled Glc. Cells were grown in the dark for 7 d at

22�C and then harvested by vacuum filtration, rinsed with 80 mL of MS

medium, frozen by immersion in liquid nitrogen, and stored at 280�C. Thisprocedure has been shown to have no discernible effect on the flux distribu-

tion in the core metabolic network of Arabidopsis (Kruger et al., 2007a) and to

lead to the establishment of an isotopic and metabolic state in the cell culture

(Williams et al., 2008).

Extraction Procedures after Stable Isotope Labeling

Soluble metabolites were extracted from frozen cells (approximately 8 g

fresh weight) using perchloric acid (Kruger et al., 2007a, 2008). Subsequently,

NMR spectra were recorded from samples dissolved in 3.2 mL of 2H2O

containing 10 mM EDTA, 25 mM 1,4-dioxane, and 10 mM KH2PO4/K2HPO4, pH

7.5. Starch and protein were extracted from the insoluble residue using

procedures that were similar to those described elsewhere (Williams et al.,

2008). Starchwas gelatinized by autoclaving awashed aliquot of the residue in

25 mM sodium acetate for 3 h, followed by enzymic digestion with a-amylase

and amyloglucosidase. The supernatant, containing Glc, was freeze dried and

redissolved in 3.2 mL of 2H2O with 25 mM 1,4-dioxane for NMR analysis.

Protein in the perchloric acid extraction pellet was hydrolyzed directly by

adding 2 mL of 6 M HCl and heating under vacuum at 110�C for 24 h in a 5-mL

Pierce hydrolysis tube (http://www.piercenet.com/). After digestion, the

protein hydrolysate was centrifuged and the supernatant was freeze dried.

The sample was redissolved in water and freeze dried again after adjusting

the pH to 7.5. Finally, the sample was redissolved in 3.2 mL of 2H2O containing

25 mM 1,4-dioxane for NMR analysis.

Quantification of Biosynthetic Outputs

Fluxes to biosynthetic products, and the rate of Glc consumption, were

obtained from the redistribution of label following metabolism of [U-14C]Glc

by cell suspensions.

Incubation of Arabidopsis Cell Culture with [U-14C]Glc

Cells were grownwith unlabeledGlc under the conditions used for the stable

isotope labeling experiments. After 6 d, 5-mL aliquots of the cell suspension

were transferred to sealed 100-mL conical flasks and supplemented with 185

kBq [U-14C]Glc (specific activity, 10.4 TBq mol21). After 12 or 24 h, the cells were

harvested by filtration and washed with two volumes of MS medium, and then

the cells and medium were frozen in liquid nitrogen. The production of 14CO2

was determined in replicate flasks inwhich the incubationwas terminated at the

appropriate time by injection of 2mL of 12 M formic acid to stopmetabolism and

release dissolved CO2 from the medium (Harrison and Kruger, 2008). Released14CO2 was collected in 1.0 mL of 10% (w/v) KOH in a vial suspended within

each sealed culture flask.

Extraction of Arabidopsis Cells after 14C Labeling

Frozen cells were combined with 1 mL of 100% methanol at 220�C and

frozen in liquid nitrogen. The sample was thawed on ice for 2 h and then

centrifuged at 8,500g for 10min at 5�C. The supernatantwas removed and stored

on ice. The pellet was resuspended in 1 mL of cold methanol, and after storage

on ice for 5 min, the suspension was centrifuged and the supernatant was

combined with the supernatant from the initial extraction. The insoluble cell

residue was then extracted twice with 1 mL of 100%methanol for 5 min at 70�C.Samples were centrifuged to separate methanol-soluble material from the

insoluble residue, and the supernatants were combined with that from the

cold extractions. Finally, the insoluble residue was washed with two further

1-mL aliquots of 100%methanol, and the insolublematerial was stored at280�C.

Ethanol Extraction from the Incubation Medium after14C Labeling

[14C]Ethanol was recovered from the cell culture medium by distillation

after adding 2 mL of 100% ethanol to the sample to aid recovery.

Fractionation of Methanolic Cell Extracts

Distilled water and chloroform were added to the methanolic fraction to

generate a CHCl3:CH3OH:water mixture (4:4:2, v/v). The mixture was centri-

fuged at 3,500g for 10 min at 20�C to promote phase separation. The aqueous

methanol phase was retained, and the chloroform phase was rinsed with a

further two washes of 2 mL of water plus 4 mL of methanol before storage at

220�C. The aqueous methanol fractions from each wash were pooled with the

initial aqueous methanol fraction, and this combined aqueous methanol

fraction was then dried by rotary evaporation at 25�C and resuspended in 5

mL of water. Total radioactivity in this fraction was measured before further

separation into basic, acidic, and neutral fractions by solid-phase extraction on

Dowex ion-exchange columns (Kruger et al., 2007a). The aqueous basic

fraction, consisting of amino acids, was freeze dried, resuspended in 1 mL of

water, and fractionated using an anion-exchange column to separate acidic

amino acids from neutral/basic amino acids (Cossins and Beevers, 1963). Each

neutral fraction was freeze dried, resuspended in 1 mL of water, and separated

into Suc, Glc, and Fru using thin-layer chromatography (Scott and Kruger, 1995).

The chloroform (lipid) fraction was separated into its components by

reverse-phase chromatography on an aminopropyl column (LC-NH2 SPE;

Supelco) as described by Kim and Salem (1990).

Total radioactivity in the methanol-insoluble material was determined by

incubating a 100-mL aliquot of insoluble residue overnight with 400 mL of

Solvable tissue solubilizer (Perkin-Elmer; http://las.perkinelmer.com) fol-

lowed by liquid scintillation counting. Insoluble material was further ana-

lyzed by autoclaving, digestion with amylase, amyloglucosidase, and

pronase, and separation into fractions by ion-exchange chromatography

(Malone et al., 2006). The resulting aqueous, acidic, and basic fractions

represented starch, cell wall material, and protein, respectively. Radioactivity

in the remaining undigested residue was determined by washing the residue

to remove the supernatant including digested material and enzymes, followed

by overnight incubation with tissue solubilizer (Kruger et al., 2007a).

Monitoring Release of 14CO2 by Arabidopsis Cell CultureMetabolizing Positionally Labeled [14C]Glc

A 5-d-old Arabidopsis cell suspension was incubated in the dark at 22�C in

a 100-mL conical flask in a total volume of 5 mL of culture medium

supplemented with 3.7 kBq [1-14C]-, [2-14C]-, [3,4-14C]-, or [6-14C]Glc. Each

flask was then sealed with a rubber bung and aerated on an orbital shaker at

100 rpm. Released 14CO2 was collected in 0.5 mL of 10% (w/v) KOH in a vial

suspended in the flask. The KOH solution was replaced at 2-h intervals for 12

h and again 24, 36, and 48 h after the beginning of the incubation. Ratios of14CO2 release from positionally labeled Glc were determined as described by

Harrison and Kruger (2008).

Determination of Radioactivity

Radioactivity was determined by liquid scintillation counting in a

Beckman LS 6500 scintillation counter. An appropriate aliquot of each fraction

was mixed with 4 volumes of Optiphase HiSafe scintillation fluid. Counting

efficiency was typically greater than 90%.

NMR Spectroscopy

One-dimensional 1H-decoupled 13C-NMR spectra were recorded at 20�C and

150.9 MHz on a Varian Unity Inova 600 spectrometer equipped with a 10-

mm-diameter broadband probe.Waltz 16 1H decouplingwas applied during the

acquisition time, and spectra were acquired with either a 6-s relaxation delay

and low-power frequency-modulated decoupling to generate the nuclear Over-

hauser effect or a 29-s relaxation delay without the nuclear Overhauser effect.

Large numbers of scans were generally required to generate spectra of sufficient





quality for quantitative analysis of the peak intensities, and spectra were

accumulated in blocks to check for unwanted spectral changes. For example, a

total acquisition time of 30 h was used for the 6-s spectra of the soluble

metabolites, and over 40 h was used for the 29-s spectra of protein hydrolysates.

Spectra of starch digests were obtained much more quickly, reflecting the high

Glc concentration in the samples. Spectra were processed using NUTS software

(AcornNMR), and in some instances, relative peak intensities in 6-s spectrawere

corrected using relative intensities in the fully relaxed spectra. Spectra were

processed after a Lorentzian-Gaussian transformation, with a line broadening of

21 Hz and a Gaussian factor of 0.1, and signal integration was performed after

baseline correction. Chemical shifts were measured relative to the 1,4-dioxane

signal at 67.30 ppm, and peaks were assigned on the basis of literature values

and authentic standards. Positional isotopomers were specified using 0 to

denote 12C, 1 to denote 13C, and X to denote either 12C or 13C.

Metabolic Modeling

Metabolic modeling was performed using 13C-FLUX (version 20050329;

Wiechert et al., 2001) according to the protocol outlined in Figures 4 and 5.

After defining the metabolic network, the free fluxes were fitted to the biomass

measurements and labeling data using the Donlp2 (P. Spelluci, Technische

Universitat Darmstadt, Germany) routine in 13C-FLUX. This sequential

quadratic programming algorithm is superior to the CooolEvoAlpha evolu-

tionary algorithm in 13C-FLUX, in terms of both computational efficiency and

avoidance of local minima (Wiechert et al., 2001). The optimizer was used in

conjunction with Monte Carlo simulations and bootstrap sampling of the

isotopomer abundances based on the measured values and their deviations to

generate solutions for the flux map. The bootstrap Monte Carlo sampling

method was invoked by the command: Donlp2 “model file name”.ftbl -a

0.00001 –m “number of simulations” -bootstrap -logAllF . “output file

name”.txt&watch “grep Resi “output file name”.txt|tail 220”. This strategy

of combining Donlp2 with bootstrap Monte Carlo sampling was first tested on

the network model distributed with 13C-FLUX and on a published model for

developing sunflower (Helianthus annuus) embryos (Alonso et al., 2007a) to

confirm that it was capable of generating the expected fluxes. Each simulation

generated a set of free fluxes after minimizing the sum of squared standard-

ized differences (the residuum) between the experimental measurements

and their predicted values. Principal component analysis of the free

fluxes obtained from 1,000 simulations was performed using SIMCA-P 11.5

(Umetrics) to identify the set of best fit feasible solutions with low values of

the residuum. This set of outputs was used to generate a mean flux solution

that was then used as a starting point for a further optimization run with

Donlp2. EstimateStat, the statistical analysis component of 13C-FLUX, was

used to compare the merits of different modeling schemes and to assign errors

to the flux values deduced by the numerical analysis.

Supplemental Data

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

Supplemental Figure S1. Metabolism of [U-14C]Glc by heterotrophic

Arabidopsis cells.

Supplemental Figure S2. Comparison of relative fractional abundance of

isotopomers in carbohydrates derived from cytosolic and plastidic

hexose phosphate pools.

Supplemental Figure S3. Distribution of flux estimates for selected steps

in the metabolic network from Monte Carlo simulations.

Supplemental Figure S4. Assessment of robustness of global best fit flux

solution.

Supplemental Figure S5. Relative expression levels of genes encoding

isozymes of the PPP and associated plastid envelope translocators in

Arabidopsis cells.

Supplemental Table S1. Calculation of fluxes to major metabolic end

products.

Supplemental Table S2. Amino acid composition of protein in cell

suspension culture.

Supplemental Table S3. Fluxes to amino acids.

Supplemental Table S4. Summary of biosynthetic output constraints

applied in flux determinations using 13C-FLUX.

Supplemental Table S5. Metabolic fluxes in heterotrophic Arabidopsis

cells.

Supplemental Table S6. Coenzyme stoichiometries for generation of

biosynthetic products from intermediates of the central metabolic net-

work.

Supplemental Table S7. Relative fractional abundance of isotopomers in

carbohydrates derived from cytosolic and plastidic hexose phosphate

pools.

Supplemental File S1. 13C-FLUX model of the central metabolic network

in Arabidopsis cell suspension culture (Microsoft Excel file).

Supplemental File S2. Positional isotopomer data sets obtained following

metabolism of [1-13C]Glc, [2-13C]Glc, and 10% [U-13C6]Glc by cell sus-

pension cultures of Arabidopsis (Microsoft Excel file).

Supplemental File S3. 13C-FLUX metabolic models of carbohydrate

oxidation containing different formulations of the oxidative PPP (Micro-

soft Excel file).

Supplemental File S4. Sensitivity analysis of oxidative PPP flux in the

biochemically preferred flux map of central metabolism in Arabidopsis

cell suspension culture (Microsoft Excel file).

Supplemental File S5. Redox analysis of the biosynthetic outputs of

heterotrophic Arabidopsis cells (Microsoft Excel file).

ACKNOWLEDGMENTS

We thank W. Wiechert (Lehrstuhl fur Simulationstechnik, Universitat

Siegen, Germany) for permission to use 13C-FLUX and T.C.R. Williams

(Department of Plant Sciences, University of Oxford) for helpful advice.

Received November 20, 2009; accepted November 24, 2009; published

November 25, 2009.

LITERATURE CITED

Allen DK, Ohlrogge JB, Shachar-Hill Y (2009) The role of light in soybean

seed filling metabolism. Plant J 58: 220–234

Allen DK, Shachar-Hill Y, Ohlrogge JB (2007) Compartment-specific

labeling information in 13C metabolic flux analysis of plants. Phyto-

chemistry 68: 2197–2210

Alonso AP, Goffman FD, Ohlrogge JB, Shachar-Hill Y (2007a) Carbon

conversion efficiency and central metabolic fluxes in developing sun-

flower (Helianthus annuus L.) embryos. Plant J 52: 296–308

Alonso AP, Raymond P, Rolin D, Dieuaide-Noubhani M (2007b) Substrate

cycles in the central metabolism of maize root tips under hypoxia.

Phytochemistry 68: 2222–2231

Alonso AP, Vigeolas H, Raymond P, Rolin D, Dieuaide-Noubhani M

(2005) 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

138: 2220–2232

Averill RH, Bailey-Serres J, Kruger NJ (1998) Co-operation between

cytosolic and plastidic oxidative pentose phosphate pathways revealed

by 6-phosphogluconate-deficient genotypes of maize. Plant J 14:

449–457

Carrari F, Urbanczyk-Wochniak E, Willmitzer L, Fernie AR (2003) Engi-

neering central metabolism in crop species: learning the system. Metab

Eng 5: 191–200

Cossins E, Beevers H (1963) Ethanol metabolism in plant tissues. Plant

Physiol 38: 375–380

Dieuaide-Noubhani M, Raffard G, Canioni P, Pradet A, Raymond P

(1995) Quantification of compartmented metabolic fluxes in maize root

tips using isotope distribution from 13C- or 14C-labeled glucose. J Biol

Chem 270: 13147–13159

Goffman FD, Alonso AP, Schwender J, Shachar-Hill Y, Ohlrogge JB

(2005) Light enables a very high efficiency of carbon storage in devel-

oping embryos of rapeseed. Plant Physiol 138: 2269–2279

Grafahrend-Belau E, Schreiber F, Koschutzki D, Junker BH (2009) Flux

Masakapalli et al.




balance analysis of barley seeds: a computational approach to study

systemic properties of central metabolism. Plant Physiol 149: 585–598

Harrison PW, Kruger NJ (2008) Validation of the design of feeding

experiments involving [14C]substrates used to monitor metabolic flux

in higher plants. Phytochemistry 69: 2920–2927

Hinkle PC (2005) P/O ratios of mitochondrial oxidative phosphorylation.

Biochim Biophys Acta 1706: 1–11

Ho CL, Saito K (2001) Molecular biology of the plastidic phosphorylated

serine biosynthetic pathway in Arabidopsis thaliana. Amino Acids 20:

243–259

Holzhutter HG (2004) The principle of flux minimization and its applica-

tion to estimate stationary fluxes in metabolic networks. Eur J Biochem

271: 2905–2922

Iyer VV, Sriram G, Fulton DB, Zhou R, Westgate ME, Shanks JV (2008)

Metabolic flux maps comparing the effect of temperature on protein and

oil biosynthesis in developing soybean cotyledons. Plant Cell Environ

31: 506–517

Joshi M, Seidel-Morgenstern A, Kremling A (2006a) Exploiting the

bootstrap method for quantifying parameter confidence intervals in

dynamical systems. Metab Eng 8: 447–455

Joshi V, Laubengayer KM, Schauer N, Fernie AR, Jander G (2006b) Two

Arabidopsis threonine aldolases are nonredundant and compete

with threonine deaminase for a common substrate pool. Plant Cell 18:

3564–3575

Junker BH, Lonien J, Heady LE, Rogers A, Schwender J (2007) Parallel

determination of enzyme activities and in vivo fluxes in Brassica napus

embryos grown on organic or inorganic nitrogen source. Phytochemis-

try 68: 2232–2242

Kauffman KJ, Prakash P, Edwards JS (2003) Advances in flux balance

analysis. Curr Opin Biotechnol 14: 491–496

Kelly EB, McMillen DP, Buono MJ (1990) Constructing confidence inter-

vals using the bootstrap: an application to a multi-product cost function.

Rev Econ Stat 72: 339–344

Kim H, Salem N (1990) Separation of lipid classes by solid phase extrac-

tion. J Lipid Res 31: 2285–2289

Kleijn RJ, van Winden WA, van Gulik WM, Heijnen JJ (2005) Revisiting

the 13C-label distribution of the non-oxidative branch of the pentose

phosphate pathway based upon kinetic and genetic evidence. FEBS J

272: 4970–4982

Krook J, Vreugdenhil D, Dijkema C, van der Plas LHW (1999) Su-

crose and starch metabolism in carrot (Daucus carota L.) cell sus-

pensions analysed by 13C-labelling: indications for a cytosol and a

plastid-localized oxidative pentose phosphate pathway. J Exp Bot 49:

1917–1924

Kruger NJ, Huddleston JE, Le Lay P, Brown ND, Ratcliffe RG (2007a)

Network flux analysis: impact of 13C-substrates on metabolism in

Arabidopsis thaliana cell suspension cultures. Phytochemistry 68:

2176–2188

Kruger NJ, Le Lay P, Ratcliffe RG (2007b) Vacuolar compartmentation

complicates the steady-state analysis of glucose metabolism and

forces reappraisal of sucrose cycling in plants. Phytochemistry 68:

2189–2196

Kruger NJ, Ratcliffe RG (2008) Metabolic organization in plants: a chal-

lenge for the metabolic engineer. In HJ Bohnert, H Nguyen, NG Lewis,

eds, Bioengineering and Molecular Biology of Plant Pathways, Vol 1.

Elsevier, Amsterdam, pp 1–27

Kruger NJ, Ratcliffe RG (2009) Insights into plant metabolic networks

from steady-state metabolic flux analysis. Biochimie 91: 697–702

Kruger NJ, Ratcliffe RG, Roscher A (2003) Quantitative approaches for

analyzing fluxes through plant metabolic networks using NMR and

stable isotope labeling. Phytochem Rev 2: 17–30

Kruger NJ, Troncoso-Ponce MA, Ratcliffe RG (2008) 1H-NMR metabolite

fingerprinting and metabolomic analysis of perchloric acid extracts

from plant tissues. Nat Protoc 3: 1001–1012

Kruger NJ, von Schaewen A (2003) The oxidative pentose phosphate

pathway: structure and organisation. Curr Opin Plant Biol 6: 236–246

Lehmann M, Schwarzlander M, Obata T, Sirikantaramas S, Burow M,

Olsen CE, Tohge T, Fricker MD, Møller BL, Fernie AR, et al (2009) The

metabolic response of Arabidopsis roots to oxidative stress is distinct

from that of heterotrophic cells in culture and highlights a complex

relationship between the levels of transcripts, metabolites and flux. Mol

Plant 2: 390–406

Libourel IGL, Shachar-Hill Y (2008) Metabolic flux analysis in plants: from

intelligent design to rational engineering. Annu Rev Plant Biol 59: 625–650

Lunn JE (2007) Compartmentation in plant metabolism. J Exp Bot 58: 35–47

Malone JG, Mittova V, Ratcliffe RG, Kruger NJ (2006) The response of

carbohydrate metabolism in potato tubers to low temperature. Plant

Cell Physiol 47: 1309–1322

Miyashita Y, Dolferus R, Ismond KP, Good AG (2007) Alanine amino-

transferase catalyses the breakdown of alanine after hypoxia in Arabi-

dopsis thaliana. Plant J 49: 92–102

Namba A (2004) Simulation studies on bootstrap empirical likelihood tests.

Comm Statist Simulation Comput 33: 99–108

Nielsen J, Villadsen J, Liden G (2003) Bioreaction Engineering Principles,

Ed 2. Springer, New York, pp 124–142

Poolman MG, Miguet L, Sweetlove LJ, Fell DA (2009) A genome-scale

metabolic model of Arabidopsis and some of its properties. Plant

Physiol 151: 1570–1581

Ratcliffe RG, Shachar-Hill Y (2005) Revealing metabolic phenotypes in

plants: inputs from NMR analysis. Biol Rev Camb Philos Soc 80: 27–43

Ratcliffe RG, Shachar-Hill Y (2006) Measuring multiple fluxes through

plant metabolic networks. Plant J 45: 490–511

Rontein D, Dieuaide-Noubhani M, Duforc EJ, Raymond P, Rolin D (2002)

The metabolic architecture of plant cells. J Biol Chem 277: 43948–43960

Roscher A, Kruger NJ, Ratcliffe RG (2000) Strategies for metabolic flux

analysis in plants using isotope labelling. J Biotechnol 77: 81–102

Scharte J, Schon H, Tjaden Z, Weis E, von Schaewen A (2009) Isoenzyme

replacement of glucose-6-phosphate dehydrogenase in the cytosol im-

proves stress tolerance in plants. Proc Natl Acad Sci USA 106: 8061–8066

Schuster S, Pfeiffer T, Fell DA (2008) Is maximization of molar yield in

metabolic networks favoured by evolution? J Theor Biol 252: 497–504

Schwender J (2008) Metabolic flux analysis as a tool in metabolic engi-

neering of plants. Curr Opin Biotechnol 19: 131–137

Schwender J, Ohlrogge JB, Shachar-Hill Y (2003) A flux model of glycol-

ysis and the oxidative pentosephosphate pathway in developing Bras-

sica napus embryos. J Biol Chem 278: 29442–29453

Schwender J, Shachar-Hill Y, Ohlrogge JB (2006) Mitochondrial metabo-

lism in developing embryos of Brassica napus. J Biol Chem 281: 34040–

34047

Scott P, Kruger NJ (1995) Influence of elevated fructose-2,6-bisphosphate

levels on starch mobilization in transgenic tobacco leaves in the dark.

Plant Physiol 108: 1569–1577

Selivanov VA, Meshalkina LE, Solovjeva ON, Kuchel PW, Ramos-

Montoya A, Kochetov GA, Lee PWN, Cascante M (2005) Rapid sim-

ulation and analysis of isotopomer distributions using constraints based

on enzyme mechanisms: an example from HT29 cancer cells. Bioinfor-

matics 21: 3558–3564

Sriram G, Fulton DB, Iyer VV, Peterson JM, Zhou R, Westgate ME,

Spalding MH, Shanks JV (2004) Quantification of compartmented

metabolic fluxes in developing soybean embryos by employing biosyn-

thetically directed fractional 13C labeling, two-dimensional [13C, 1H]

nuclear magnetic resonance, and comprehensive isotopomer balancing.

Plant Physiol 136: 3043–3057

Sriram G, Fulton DB, Shanks JV (2007) Flux quantification in central

carbon metabolism of Catharanthus roseus hairy roots by 13C labeling and

comprehensive bondomer balancing. Phytochemistry 68: 2243–2257

Sweetlove LJ, Fell D, Fernie AR (2008) Getting to grips with the plant

metabolic network. Biochem J 409: 27–41

Wakao S, Andre C, Benning C (2008) Functional analyses of cytosolic

glucose-6-phosphate dehydrogenases and their contribution to seed oil

accumulation in Arabidopsis. Plant Physiol 146: 277–288

Wiback SJ, Famili I, Greenberg HJ, Palsson BØ (2004) Monte Carlo

sampling can be used to determine the size and shape of the steady-state

flux space. J Theor Biol 228: 437–447

Wiechert W, Mollney M, Petersen S, de Graaf AA (2001) A universal

framework for 13C metabolic flux analysis. Metab Eng 3: 265–283

Williams TCR, Miguet L, Masakapalli SK, Kruger NJ, Sweetlove LJ,

Ratcliffe RG (2008) Metabolic network fluxes in heterotrophic Arabi-

dopsis cells: stability of the flux distribution under different oxygena-

tion conditions. Plant Physiol 148: 704–718





subcellular flux analysis of central metabolism in a ... · subcellular flux analysis of central...

Documents