34
Question d’examen La question d’examen pour la partie du cours sur “flux métaboliques” portera sur l’article dont la référence vous est fournie ci-dessous et fera appel à votre compréhension des concepts discutés pendant les cours. Plus précisément, la question portera sur la stratégie expérimentale et sur les données des expériences avec les substrats marqués. Vous ne serez pas questionnés sur les aspects techniques. 1

Question d’examen La question d’examen pour la …esilrch1.esi.umontreal.ca/~syguschj/cours/BCM6225/Lecture Notes... · Question d’examen La question d’examen pour la partie

  • Upload
    ngothu

  • View
    235

  • Download
    0

Embed Size (px)

Citation preview

Question d’examen

La question d’examen pour la partie du cours sur “flux métaboliques” portera sur l’article dont la référence vous est fournie ci-dessous et fera appel à votre compréhension des concepts discutés pendant les cours. Plus précisément, la

question portera sur la stratégie expérimentale et sur les données des expériences avec les substrats marqués. Vous ne serez pas questionnés sur les

aspects techniques.

1

Effect of Anaplerotic Fluxes and Amino Acid Availability onHepatic Lipoapoptosis*□S

Received for publication, July 28, 2009, and in revised form, September 3, 2009 Published, JBC Papers in Press, September 16, 2009, DOI 10.1074/jbc.M109.049478

Yasushi Noguchi‡1, Jamey D. Young‡1,2, Jose O. Aleman‡, Michael E. Hansen‡, Joanne K. Kelleher‡§,and Gregory Stephanopoulos‡3

From the ‡Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139 and the§Shriners Burn Hospital, Massachusetts General Hospital, Boston, Massachusetts 02115

To identify metabolic pathways involved in hepatic lipoapop-tosis, metabolic flux analysis using [U-13C5]glutamine as an iso-topic tracer was applied to quantify phenotypic changes inH4IIEC3 hepatoma cells treated with either palmitate alone(PA-cells) or bothpalmitate andoleate in combination (PA/OA-cells). Our results indicate that palmitate inhibited glycolysisand lactate dehydrogenase fluxes while activating citric acidcycle (CAC) flux and glutamine uptake. This decoupling of gly-colysis and CAC fluxes occurred during the period followingpalmitate exposure but preceding the onset of apoptosis. Oleateco-treatment restoredmost fluxes to their control levels, result-ing in steatotic lipid accumulation while preventing apoptosis.In addition, palmitate strongly increased the cytosolic NAD�/NADH ratio, whereas oleate co-treatment had the oppositeeffect on cellular redox. We next examined the influence ofamino acids on these free fatty acid-induced phenotypic changes.Increased medium amino acids enhanced reactive oxygen species(ROS) generation and apoptosis in PA-cells but not in PA/OA-cells. Overloading the medium with non-essential amino acidsinduced apoptosis, but essential amino acid overloading partiallyameliorated apoptosis. Glutamate was the most effective singleamino acid in promoting ROS. Amino acid overloading alsoincreased cellular palmitoyl-ceramide; however, ceramide synthe-sis inhibitors had no effect on measurable indicators of apoptosis.Our results indicate that free fatty acid-induced ROS generationandapoptosis are accompaniedby thedecouplingof glycolysis andCAC fluxes leading to abnormal cytosolic redox states. Aminoacids play a modulatory role in these processes via a mechanismthat does not involve ceramide accumulation.

Elevated serum free fatty acids (FFAs)4 cause hepatic apopto-sis, which is a prominent feature of non-alcoholic steatohepa-

titis and correlates with disease severity (1). Previous in vitrostudies in Chinese hamster ovary (CHO) cells (2, 3), cardiacmyocytes (4, 5), pancreatic �-cells (6), breast cancer cells (7),and hepatic cells (8, 9) have demonstrated that saturated fattyacids (SFAs) but not monounsaturated fatty acids (MUFAs)induce reactive oxygen species (ROS) generation and apoptosis,whereas MUFAs predominately induce steatosis. An earlierstudy showed thatMUFA co-treatment changed the palmitate-induced phenotype from apoptosis to steatosis by divertingSFAs into triglyceride synthesis and thereby reducing apoptosisin CHO cells (2). Increasing the desaturation of cellular fattyacids by stearoyl-CoA desaturase overexpression had an effectsimilar to that of oleate co-treatment in MIN6 cells (6).Ceramide accumulation has been considered a primary fac-

tor responsible for SFA-induced ROS generation and apopto-sis, because ceramide is synthesized de novo frompalmitate andserine and also has been shown to activate apoptotic signaling(10, 11). Recent studies, however, have reported that SFAs caninduce apoptosis through ROS formation (3) and endoplasmicreticulum stress without altering intracellular ceramide levels(12, 13). Thus, a consensus mechanism linking SFA-inducedmetabolic alterations to apoptosis has been difficult to estab-lish. To further address these questions, we hypothesized thatmetabolites or pathways other than those involved with cer-amide synthesis may influence ROS generation and the result-ing cellular phenotypes associated with elevated SFAs.The influence of non-lipid substrate availability on FFA-in-

duced pathology is another important issue addressed by thisstudy (14). In particular, amino acid availability is known toacutely and chronically regulate hepatic FFA metabolism viachanges in both gene expression and metabolic fluxes. Anexample of the former is provided by the effect of amino aciddeprivation to activate the global transcriptional regulatorGCN2 (15, 16). Changes in citric acid cycle (CAC) and anaple-rotic fluxes have also been observed as a result of varying extra-cellular amino acid concentrations (17, 18). We hypothesizedthat altering amino acid levels could be oneway tomodulate theresponse of liver cells to elevated FFA concentrations, therebyproviding additional insight into themetabolic deviations lead-ing to apoptosis.In this study, we first applied stable isotope-based compre-

hensive metabolic flux analysis to assess metabolic perturba-tions in H4IIEC3 hepatoma cells cultured with different typesof FFA (palmitate or oleate). Thenwe examined the influence ofamino acid availability on FFA-induced phenotypes by cultur-ing H4IIEC3 cells under varying levels of amino acids.

* This work was supported, in whole or in part, by National Institutes of HealthGrants DK075850 and ES013925.

□S The on-line version of this article (available at http://www.jbc.org) containssupplemental Tables S1–S9 and Figs. S1–S7.

1 Both of these authors have contributed equally to this work.2 Supported by National Institutes of Health Grant F32 DK072856.3 To whom correspondence should be addressed: Dept. of Chemical Engi-

neering, Massachusetts Institute of Technology, Cambridge, MA 02139.Tel.: 617-258-0398; Fax: 617-253-3122; E-mail: [email protected].

4 The abbreviations used are: FFA, free fatty acid; CHO, Chinese hamsterovary; DMEM, Dulbecco’s modified Eagle’s medium; EAA, essential aminoacid; GC, gas chromatography; MS, mass spectrometry; MUFA, monounsat-urated fatty acid; NEAA, non-essential amino acid; PA-cell, H4IIEC3 hepatomacell treated with palmitate alone; PA/OA-cell, H4IIEC3 hepatoma cell treatedwith both palmitate and oleate in combination; ROS, reactive oxygen species;SFA, saturated fatty acid; PK, pyruvate kinase; CAC, citric acid cycle.

THE JOURNAL OF BIOLOGICAL CHEMISTRY VOL. 284, NO. 48, pp. 33425–33436, November 27, 2009© 2009 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in the U.S.A.

NOVEMBER 27, 2009 • VOLUME 284 • NUMBER 48 JOURNAL OF BIOLOGICAL CHEMISTRY 33425

at BIB

LIO D

E LA

SA

NT

E, on F

ebruary 23, 2010w

ww

.jbc.orgD

ownloaded from

http://www.jbc.org/content/suppl/2009/09/16/M109.049478.DC1.htmlSupplemental Material can be found at:

EXPERIMENTAL PROCEDURES

Materials—Fumonisin-B1, cycloserine, sphingomyelinase(Bacillus cereus) and free amino acids were purchased fromSigma. Fatty acid-free bovine serum albumin was purchasedfrom JRH Biosciences (Lenexa, KS). C16 and C17 ceramidewere purchased from Avanti Polar Lipids (Alabaster, AL).Amino acid-free Dulbecco’s modified Eagle’s medium (DMEM)was a gift of Ajinomoto Co., Inc. (Tokyo, Japan). [U-13C5]-Glutamine (99 atom % 13C) was purchased from Cambridge Iso-tope Laboratories (Andover, MA). All silylation reagents used inthis study were purchased from Pierce.Human and Rat Plasma Amino Acids—To formulate our

growth medium to mimic the composition of physiologicalfluids, plasma amino acid concentrations were measured inhuman and rat samples using an L-8800 automatic amino acidanalyzer (Hitachi, Tokyo, Japan). A total of 39 healthy volunteersamples and 29 rat samples were used. Blood samples werecollected after overnight fasting. Plasma sampleswere preparedusing EDTA as anticoagulant, treated with 2 volumes of 5%(w/w) trichloroacetic acid, and then centrifuged to remove pro-tein as precipitate. The samples obtained were filtered throughanUltrafree-MCcentrifugal filter (Millipore, Billerica,MA). Toprepare deproteinized tissue extracts, tissues were homoge-nized with 5% trichloroacetic acid and processed with the sameprotocol used for plasma. The samples were kept at 4 °C duringall steps to minimize chemical reactions of thiol-metabolitesand stored at �80 °C. The studies of human and rat aminoacid profiling were conducted by Ajinomoto Co., Inc., andprotocols were approved by Ajinomoto Ethical and AnimalCare Committees.Cell Culture—H4IIEC3 cells (American Type Culture Col-

lection, Manassas, VA), rat liver hepatomas, were cultured inDMEM supplemented with 10% fetal bovine serum, 50units/ml penicillin, and 50 units/ml streptomycin sulfate. Toexamine the influence of medium amino acids on FFA-inducedhepatic phenotypes, the medium was changed to a customizedmedium containing 25 mM glucose, 2.15 mM amino acids,amino acid-free DMEM, 20 mM HEPES, and 0.5% serumreplacement 3 with or without FFA-bovine serum albumincomplex (palmitate, palmitate plus oleate). Based on a previouspaper (19), amino acid composition was designed to resemblehuman and rat plasma amino acids (Table 1). Total amino acidcontent or the relative amino acid composition was varied insome experiments to examine the effects of these parameters.For isotopic analysis of metabolic flux, [U-13C5]glutamine

was used as a tracer substrate. Cells were preconditioned for 6 hin unlabeled medium before washing with phosphate-bufferedsaline and replenishing with medium containing labeled gluta-mine. Both medium and cellular metabolites were analyzed byGC-MS at 3, 6, and 9 h.ROS Measurement—Cellular ROS generation was measured

using 2�,7�-dichlorodihydrofluoresceindiacetate (Invitrogen)as described previously (20). 2�,7�-Dichlorodihydrofluorescein-diacetate is a non-polar compound that readily diffuses intocells, where it is hydrolyzed by intracellular esterases to thenonfluorescent dichlorodihydrofluorescein and trapped withinthe cells. In the presence of a proper oxidant, dichlorodihy-

drofluorescein is oxidized to the highly fluorescent 2,7-dichlo-rofluorescein. Cells treated with various fatty acids or aminoacids were followed by incubation with 10 �M 2�,7�-dichlorodi-hydrofluoresceindiacetate for 30 min at 37 °C in the dark andthen resuspended with Hanks’ balanced salt solution. The fluo-rescence intensity was measured at an excitation/emissionwavelength of 490/530 nm using a FusionTM universal micro-plate analyzer (PerkinElmer Life Sciences).Nile Red Staining—Cells were cultured in 96-well plates with

or without FFA. After a 24-h incubation, cells were stained byNile red using AdipoRed reagent (Cambrex, East Rutherford,NJ). Intracellular lipid accumulation was observed using a fluo-rescence microscope and also quantified using a microplateanalyzer at an excitation/emission wavelength of 485/590 nm.Detection of Apoptotic Markers—Caspase-3/7 activities were

measured using the Apo-ONE homogeneous caspase-3/7 assaykit (Promega, Madison, WI) according to the manufacturer’sprotocol. Briefly, cells were cultured in 96-well plates with orwithout FFA. After a 16-h incubation, cells were lysed withbuffer containing caspase substrate benzyloxycarbonyl-DEVD-R100 and incubated at room temperature. Caspase-3/7 activi-ties were detected at an excitation/emission wavelength of 485/535 nm. For DNA laddering analysis, cells were cultured in12-well plates, and DNAwas extracted using a Suicide trackTM

DNA ladder isolation kit (EMD Chemicals, San Diego, CA).The isolated DNA fragments were separated in 1.5% agarosegels and detected by SYBR gold (Invitrogen).Cell Viability—Cultured cells in 96-well plates were washed

twice with phosphate-buffered saline and then incubated withserum-free DMEM for 4 h at 37 °C. Cellular metabolic capacitywas measured based on resazurin reduction using theCellTiter-Blue� cell viability assay (Promega).Metabolite Extraction—We employed a biphasic extraction

protocol, with non-polar metabolites partitioning into a chlo-roform phase and polar metabolites partitioning into a metha-nol/water phase. Cultured cells in 100-mm dishes were washedtwice with ice-cold phosphate-buffered saline and immediatelyquenchedwith 1ml of precooledmethanol (�80 °C). Cellswerecollected in glass tubes with a cell scraper (BD Biosciences).Dishes were resuspended with 1 ml of ice-cold water, and theremaining cells were combined in the same tubes. For proteindetermination, 20�l of cell suspensionwas taken from the sam-ple tubes and stored at �80 °C. Afterward, 15.8 nmol of trihep-tadecanoin, 2.94 nmol of 5�-cholestane, and 3.5 nmol ofN-heptadecanoyl-sphingosine (C17 ceramide) in 30 �l of chlo-roform (non-polar internal standards) and also 9.9 nmol of rib-itol and 11.5 nmol of norvaline in 30 �l of methanol (polarinternal standards) were added. After the addition of 1 ml ofchloroform, samples were shaken for 30 min at room tempera-ture, and then 3ml of chloroform and 2ml of water were added.Vortexed samples were centrifuged at 4,000 � g for 30 min atroom temperature. Two 2-ml extracts from the methanol/wa-ter phase or non-polar samples from the chloroform phasewere separately collected to new tubes and then evaporated todryness. All samples were stored at�80 °Cwhile awaiting anal-ysis. Protein contents in each sample were quantified using aBCA assay kit (Pierce).

Metabolic Flux Analysis of Hepatic Lipoapoptosis

33426 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 284 • NUMBER 48 • NOVEMBER 27, 2009

at BIB

LIO D

E LA

SA

NT

E, on F

ebruary 23, 2010w

ww

.jbc.orgD

ownloaded from

http://www.jbc.org/content/suppl/2009/09/16/M109.049478.DC1.htmlSupplemental Material can be found at:

Polar Metabolite Analysis—Derivatization of polar metabo-lites was performed according to previous reports (21, 22) withthe following modifications. One of each pair of dried sam-ples was dissolved in 30 �l of methoxyamine hydrochloride(20 mg/ml in pyridine). Sample solutions were sonicated for15 min at room temperature and incubated for 120 min at37 °C. Afterward, 70 �l of N-methyl-N-tert-butyldimethyl-silytrifluoroacetamide (MBTSTFA) plus 1% tert-butyldi-methylchlorosilane (TBDMCS) was added, and the solutionswere further incubated overnight at room temperature.Gas chromatography-mass spectrometry (GC-MS) analy-

sis was performed using an Agilent 6890N gas chromato-graph equipped with a 30-meter DB-35ms capillary columnconnected to an Agilent 5975B mass spectrometer operatingunder ionization by electron impact at 70 eV. An injection vol-ume of 1 �l was introduced in splitless mode at an injectiontemperature of 270 °C. The temperature of the MS source andquadrupole were held at 230 and 150 °C, respectively. Thedetector was set to scan over the mass range 50–550m/z. TheGC temperature programwas optimized to follow a large num-ber of metabolites simultaneously: 5 min at 90 °C, 60-min rampto 280 °C, and held for 5 min at 280 °C (total 70 min/run).Non-polar Metabolite Analysis—For ceramide and free fatty

acid analysis, dried non-polar samples were dissolved in 1.55mlof isooctane/methanol/ethyl acetate (20:10:1). To remove tri-glyceride, samples were applied to a silica gel packed Poly-Prepcolumn (Bio-Rad) as described previously (23, 24). Eluted freelipid fractions were evaporated to dryness. For ceramide anal-ysis, sampleswere redissolved in 250�l of phosphate buffer (0.1M, pH 7.0) containing 0.1 unit of sphingomyelinase and 1 mM

MgCl2. Free lipid fractions were extracted three times in 500 �lof isooctane/ethyl acetate (3:1) and then evaporated to dryness.Samples were dissolved in 150 �l of N,O-bis(trimethylsilyl)tri-fluoroacetamide (BSTFA) plus 1% trimethylchlorosilane(TMCS), acetonitrile (4:1) and then incubated overnight atroom temperature. GC-MS analysis of free lipids was per-formedwith the following parameter settings. The temperatureof the injection port, MS source, and quadrupole were set at310, 230, and 150 °C, respectively. The GC temperature pro-gram was set as follows: 3 min at 130 °C, 4-min ramp to 190 °C,3 min at 190 °C, 12.3-min ramp to 264 °C, 5 min at 264 °C,5.75-min ramp to 287 °C, 8 min at 287 °C, 4.6-min ramp to310 °C, 3min at 310 °C, 4.7-min ramp to 325 °C, and 16.6min at325 °C (total 70 min/run).For total cellular fatty acid analysis, dried non-polar samples

were dissolved in 250 �l of 0.5 N KOH in methanol and thenincubated at 70 °C for 1 h. After the addition of 250 �l of 14%trifluoroborane inmethanol, samples were further incubated at70 °C for 2h. 250 �l of saturated NaCl water was added to theresulting samples, and fatty acid methyl esters were extractedtwice in 500 �l of hexane. The temperature program for fattyacid methyl esters was set as follows: 5 min at 100 °C, 5-minramp to 175 °C, 1min at 175 °C, 11-min ramp to 208 °C, 3.6minat 208 °C, 1.4-min ramp to 215 °C, 4 min at 215 °C, 2-min rampto 215 °C, 7-min ramp to 255 °C, 3-min ramp to 300 °C, and 2min at 300 °C (total 45 min/run).Cellular Redox—Cytosolic NAD�/NADH ratio was calcu-

lated from the lactate/pyruvate ratio obtained from GC-MS

profiling data, assuming that Keq � 1.11 � 10�4 for lactatedehydrogenase as described (25). Cellular NAD�/NADH ratiowas also determined enzymatically using an NAD�/NADHquantification kit (Biovision, Mountain View, CA) accordingto the manufacturer’s protocol. Cultured cells in 24-wellplates were washed with ice-cold phosphate-buffered salineand then quenched with 200 �l of extraction buffer. Sampleswere vortexed and centrifuged at 16,000 � g for 5 min atroom temperature. The NAD�/NADH ratio in extractedsolutions was determined colorimetrically using a micro-plate analyzer at 450 nm.GC-MS Data Analysis—All GC-MS data were analyzed

according to Styczynski et al. (26). Briefly, mass spectra wereprocessed by the AMDIS software (available from the NationalInstitute of Standards and Technology web site). The resultingELU files were further analyzed by the SpectConnect software(available on the MIT web site) developed in our laboratory toidentify well conserved peaks among multiple GC-MS chro-matograms. Metabolite identification of electron impact-MSpeaks was performed using in-house standard libraries alongwith the NIST05 MS library described in Ref. 27. Two-dimen-sional clustering of metabolic flux distributions was performedusing JMP (SAS Institute Inc., Cary, NC).GC-MS Nomenclature—In discussing isotopic labeling data

obtained by GC-MS, we distinguish between multiply labeledmolecules of a given metabolite using the following terminol-ogy. Positional isotopomers are molecules with identical globalisotopic composition but differing in the position of their heavyatoms. Mass isotopomers differ in the total number of theirheavy atoms, resulting in different molecular weights.We referto the mass isotopomers of a metabolite as M�0, M�1, M�2,etc. in order of increasing weight.Extracellular Metabolite Determination—Medium glucose,

lactate, and glutamine were enzymatically determined using aglucose assay kit (glucose oxidase; Sigma), lactate assay kit (Bio-Vision, Mountain View, CA), and L-glutamine/ammonia rapidassay kit (Megazyme,Wicklow, Ireland), respectively. Allmeas-urements were quantified by colorimetric 96-well plate assays.To estimate metabolite uptake or secretion, we obtained timecoursemeasurements at 3-h intervals and computed the rate ofchange during each interval by differencing.Metabolic Flux Determination—The labeling patterns of

metabolic intermediates and by-products that emerge uponfeeding an isotopically labeled substrate to cultured cells aredirectly determined by the relative pathway fluxes in their bio-chemical network (28). This enables the reconstruction ofmetabolic flux maps by minimizing the lack of fit between sim-ulated and measured GC-MS data. We applied Metran, a soft-ware package developed in our laboratory, to estimate meta-bolic fluxes from isotope labeling data obtained byGC-MS (29).Metran relies upon the recently developed elementary metab-olite unit framework to efficiently simulate the effects of vary-ing fluxes on the labeling state of measurable metabolites (30,31). The flux parameters of the network model are iterativelyadjusted using a Levenberg-Marquardt algorithm until optimalagreement with the experiment is obtained (32). All resultswere subjected to a �2 statistical test to assess goodness of fit,and accurate 95% confidence intervals were computed for all

Metabolic Flux Analysis of Hepatic Lipoapoptosis

NOVEMBER 27, 2009 • VOLUME 284 • NUMBER 48 JOURNAL OF BIOLOGICAL CHEMISTRY 33427

at BIB

LIO D

E LA

SA

NT

E, on F

ebruary 23, 2010w

ww

.jbc.orgD

ownloaded from

http://www.jbc.org/content/suppl/2009/09/16/M109.049478.DC1.htmlSupplemental Material can be found at:

estimated parameters by evaluating the sensitivity of the sumof squared residuals to flux variations (33). Refer to thesupplemental material for a detailed discussion of the modelformulation used to estimate fluxes.Flux results are presented in terms of net and exchange

fluxes. All net fluxes were expressed relative to a pyruvatekinase (PK) flux of 100 to facilitate flux comparisons acrosstreatments. The actual PK flux in each condition was deter-mined from the labeling of pyruvate and CAC intermediatesalong with the measured [U-13C5]glutamine uptake ratebecause the PK flux provides the main source of unlabeledmaterial that dilutes the 13C tracer. Carbon loss via oxidativepentose phosphate pathway activity or glycogen depositionwasassumed to account for any mismatch between glucose uptakeand PK flux. Exchange fluxes represent the extent of reactionreversibility and are computed as min(vf,vr), where vf is the for-ward reaction rate and vr is the rate in the reverse direction.Exchange fluxes were scaled according to the transformation,

Vexch�0,100� � 100 �

vexch

vexch � vref(Eq. 1)

where vref is the reference (PK) flux value (34).Statistics—Unless otherwise indicated, data are expressed as

mean � S.D. Statistical significance was determined by Dun-nett’s or Tukey’s honestly significant differences test accordingto each experimental design.

RESULTS

FFA Treatments Induce Hepatic ROS Generation and LipidAccumulation at Physiologic Amino Acid Concentrations—Inthe present study, we used serum-free medium to mimic the invivo amino acid composition of human and rat plasma and touncover physiologically relevant metabolic perturbationscaused by FFA treatments. In contrast, typical DMEM culturemediumcontains unrealistically high amino acid levels andmaylead to non-representative phenotypes (Table 1). We first con-firmed the practicality of our medium by cultivating H4IIEC3cells with FFA (100 or 400 �M) and then observing cellularlipids, ROS, and cell viability with Nile red, DCF fluorescence,and resazurin reduction, respectively. Intracellular lipid drop-lets were observed when cells were incubated for 24 h withpalmitate or oleate, but the effect was more pronounced inoleate-treated cells (Fig. 1A). On the other hand, 400�M palmi-tate, but not oleate, induced ROS generation and reduced cellviability (Fig. 1B). Concurrent exposure to 50�Moleate in addi-tion to 400 �M palmitate strongly induced lipid accumulationand also alleviated ROS in comparison with cells treated withpalmitate alone (Fig. 1, A and B). The effects varied with FFAconcentration because 100�M palmitate failed to impact eitherROS generation or cell viability (Fig. 1B).Palmitate Leads to Decoupling of Glycolysis and CAC Fluxes—

To test whether the induction of apoptotic or steatotic pheno-types by FFA was associated with changes in central carbonmetabolism, we carried out metabolic flux analysis of centralcarbon pathways in preapoptotic cells (at 3, 6, and 9 h) using[U-13C5]glutamine as an isotopic tracer in the presence of phys-iologic amino acid levels. Here, we confirmed that apoptotic

indicators, such as caspase-3/7 activity and DNA laddering,were not changed at least up to 10 h (data not shown). Thenetwork model used in this study is shown in Fig. 2A. We firstdetermined external fluxes of glucose uptake, glutamine up-take, and lactate production by measuring extracellular metab-olites as described under “Experimental Procedures” (Fig. 2B).Although PA-cells exhibited qualitatively lower glucose uptakecompared with untreated cells and PA/OA-cells, significantdifferences were not detected due to the inherent variability ofmeasuring small changes in the high glucose concentration ofourmedium (Fig. 2B, top).On the other hand, glutamine uptakewas significantly higher in PA-cells (Fig. 2B, middle). Lactateproduction of both FFA-treated cultures decreased from 3 to9 h, whereas that of untreated cells remained relatively constantthroughout the experiment (Fig. 2B, bottom).Next, we calculated a total of 19 net and seven exchange

fluxes from isotopic labeling patterns of intracellular metabo-lites, as described under “Experimental Procedures” (Fig. 2, Cand D). The ratio of CAC flux to glycolysis is most stronglydetermined by the GC-MS measurements of CAC intermedi-ates and related amino acids. To understand this, one mustrecognize that when [U-13C5]glutamine enters the CAC, it willinitially give rise to the fully labeled M�5 mass isotopomer of�-ketoglutarate and thenM�4 succinate following decarboxy-lation. As this labeled material continues to cycle, it will lead tolower mass isotopomers, such as M�1 succinate. Increaseddilution from unlabeled sources (mainly glucose) results indecreased abundance of these lower mass isotopomers relativeto their fully labeled precursors. A good example of this is foundin the succinate labeling data shown in Fig. 3, where we foundthat the ratio of M�1/M�4 mass isotopomers was highest inPA-cells compared with untreated cells and PA/OA-cells. Sim-ilar trends were evident in other intermediates, such as malateand aspartate, indicating that exposure to palmitate caused

TABLE 1Comparison of physiological and medium amino acids

Amino acidPlasma Culture medium

Human Rat DMEM This study

�M �M

EssentialHistidine 60–120 45–90 200 50Isoleucine 40–105 30–115 800 50Leucine 85–185 40–195 800 100Lysine 125–310 80–285 800 200Methionine 15–40 30–100 200 50Phenylalanine 40–75 34–85 400 50Threonine 70–215 55–270 800 100Tryptophan 45–80 30–105 100 50Valine 140–360 55–235 800 150

Non-essentialAlanine 220–610 220–460 0 250Arginine 45–150 36–143 400 80Asparagine 35–90 35–58 0 50Aspartate 3–15 38–90 0 20Cysteine 10–45 10–56 200 50Glutamate 20–120 42–94 0 50Glutamine 375–760 375–700 4000 450Glycine 160–390 105–460 400 150Proline 90–190 100–220 0 100Serine 70–215 110–210 400 100Tyrosine 40–85 30–100 400 50

Total 1876–3471 1875–3500 10700 2150EAA/NEAA 0.39–0.68 0.30–0.66 0.84 0.59

Metabolic Flux Analysis of Hepatic Lipoapoptosis

33428 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 284 • NUMBER 48 • NOVEMBER 27, 2009

at BIB

LIO D

E LA

SA

NT

E, on F

ebruary 23, 2010w

ww

.jbc.orgD

ownloaded from

http://www.jbc.org/content/suppl/2009/09/16/M109.049478.DC1.htmlSupplemental Material can be found at:

CAC flux to increase relative to glycolysis and that this effectwas reversed by oleate co-treatment.Correlation analysis revealed that untreated cells and

PA/OA-cells exhibit similarmetabolic flux distributions, whereeach condition was hierarchically clustered based on the simi-larity of theirmeasured flux patterns (Fig. 2C, left). On the otherhand, the flux distribution of PA-cells diverged markedly fromthe other groups, particularly at the 9 h time point, wherepalmitate treatment appears to strongly down-regulate glyco-lytic and lactate dehydrogenase fluxes while up-regulatingmitochondrial CAC and anaplerotic fluxes (Fig. 2C, right). FFAtreatments also altered exchange fluxes (Fig. 2D, left), whichwas detected by the reduced isotopic equilibration of glutamateand aspartate with their corresponding keto acids at 9 h (Fig.2D, right). This could indicate reduced aspartate/malate shuttleactivity because this cycle results in exchange betweenGlu/Akgand Asp/Mal. Taken as a whole, these data indicate that palmi-

tate strongly induced metabolicdecoupling by down-regulating gly-colysis and up-regulating CACfluxes.Palmitate and Oleate Induce

Opposing Changes to Cellular Redox—We estimated cytosolic NAD�/NADH from the lactate/pyruvateratio obtained byGC-MS analysis asindicated under “Experimental Pro-cedures.” The NAD�/NADH ratiowas maintained at high levels inPA-cells throughout the cultivation,whereas in contrast, it was consis-tently low in PA/OA-cells (Fig. 4, Aand B). These data indicate thatmetabolic changes induced by bothpalmitate and oleate are associatedwith changes in cellular redox.To test whether the observed

cytosolic flux and redox changesinduced by oleate co-treatmentaffect triglyceride synthesis, we firstdetermined total cellular fatty acidsand FFA. Total cellular palmitateand oleate contents were increasedin PA/OA-cells relative to PA-cells(Table 2). However, free palmitateand stearate as well as total FFAwere decreased (Table 2). We nextdetermined free intracellular glyc-erol 3-phosphate aswell as free glyc-erol in palmitate-treated cells bothin the presence and absence ofoleate. The results clearly show thatoleate co-treatment significantlyincreased the glycerol 3-phosphateand free glycerol pool sizes (Table3). These data indicate that oleateco-treatment reduced the free SFApool by enhancing triglyceride in-

corporation. This is consistent with the ability of oleate torestore cytosolic redox because glycerol 3-phosphate produc-tion depends upon reducing power supplied by cytosolicNADH.Amino Acids Modulate FFA-induced Metabolic Alterations—

To test the effects of amino acid availability on FFA-inducedmetabolic changes, we supplemented H4IIEC3 cells with vari-ous amino acid mixtures and measured the impact on centralcarbon fluxes, ROS generation, and apoptosis. First, non-essen-tial amino acid (NEAA) or essential amino acid (EAA)mixtureswere separately added into the basal medium shown in Table 1.The NEAA and EAA mixtures were prepared according to theamino acid composition of the basalmedium. EAA supplemen-tation significantly increased glucose uptake and lactate dehy-drogenase fluxes while decreasing glutamine uptake (Fig. 5, Aand B). On the other hand, NEAA supplementation signifi-cantly enhanced palmitate-induced flux alterations by reducing

FIGURE 1. FFA-induced apoptosis and steatosis in H4IIEC3 cells under physiologic amino acid condi-tions. A, microscopic observations of FFA-induced lipid accumulation and ROS generation in H4IIEC3 cellsat physiologic amino acid levels. Cells were incubated with 400 �M palmitate, 400 �M oleate, 400 �M

palmitate plus 50 �M oleate or without FFA (none) for either 12 h (for ROS detection) or 24 h (for lipiddetection). Cellular lipids and ROS were determined using Nile red and DCF fluorescence, respectively.B, quantification of FFA-induced lipid accumulation, ROS generation, and cell viability. H4IIEC3 cells wereincubated with 100 or 400 �M palmitate or without FFA (white bar). Also, cells were incubated with 400 �M

palmitate with 10 or 50 �M oleate. Lipids, ROS and cell viability were determined using Nile red, DCFfluorescence, and resazurin reduction, respectively. Values are expressed as mean � S.E. (n � 3). *, p 0.05for the comparison with untreated cells (white bars).

Metabolic Flux Analysis of Hepatic Lipoapoptosis

NOVEMBER 27, 2009 • VOLUME 284 • NUMBER 48 JOURNAL OF BIOLOGICAL CHEMISTRY 33429

at BIB

LIO D

E LA

SA

NT

E, on F

ebruary 23, 2010w

ww

.jbc.orgD

ownloaded from

http://www.jbc.org/content/suppl/2009/09/16/M109.049478.DC1.htmlSupplemental Material can be found at:

glucose uptake and increasing glu-tamine uptake (Fig. 5A), leading tofurther decoupling between glycol-ysis and CAC fluxes (Fig. 5B).Therefore, these data indicate thatmedium amino acid availability canmodulate the decoupling betweencentral metabolic pathways inpalmitate-treated hepatic cells.Amino Acid Levels Affect FFA-in-

duced Hepatic Apoptosis—Meta-bolic flux analysis indicated theimportance of anaplerosis andamino acid availability in the earlyphase of palmitate-induced apopto-sis. To further explore this result, wevaried the total concentration ofmedium amino acids (1.08, 2.15, or8.60 mM) in the presence of 400 �M

palmitate. Cell viability, cellular lip-ids, and ROS generation were mon-itored using resazurin reduction,Nile red, and DCF fluorescence,

FIGURE 2. Stable isotopic metabolic flux analysis of central carbon pathways in H4IIEC3 cells. Metabolic flux analysis of central carbon pathways using[U-13C5]glutamine at 3, 6, and 9 h. A, a metabolic network model for flux estimation in hepatoma cultures. A total of 19 net and seven exchange fluxes werecalculated as described under “Experimental Procedures.” AcCoA, acetyl-CoA; Akg, �-ketoglutarate; Cit, citrate; Fum, fumarate; G6P, glucose 6-phosphate; Lac,lactate; Mal, malate; Pyr, pyruvate; Suc, succinate; T3P, triose 3-phosphate; AST, aspartate aminotransferase; ACL, ATP-citrate lyase; ADH, �-ketoglutaratedehydrogenase; CS, citrate synthase; FUM, fumarase; GDH, glutamate dehydrogenase; GLN, glutaminase; IDH, isocitrate dehydrogenase; LDH, lactate dehy-drogenase; ME, malic enzyme; PDH, pyruvate dehydrogenase; PGI, phosphoglucose isomerase; SDH, succinate dehydrogenase. B, external fluxes of glucoseuptake, glutamine uptake, and lactate production calculated from extracellular glucose, glutamine, and lactate measurements, respectively. Values areexpressed as �mol/h/106 cells based on measurements over the 3-h intervals 0 –3, 3– 6, and 6 –9 h. C and D, hierarchical clustering analysis of net (C, left) andexchange fluxes (D, left) under different FFA treatments. All net flux values are normalized to the PK flux. The net and exchange fluxes are represented by thecolor range from 0 (intense green) to 2 (intense red) in C and from 0 (green) to 100% (red) in D, respectively. Representative fluxes are shown on the right-hand sidesof C and D. C (right), net fluxes expressed relative to a PK flux of 100. D (right), amino acid turnover. The first plot represents the extent of equilibration betweenGlu and mitochondrial Akg and the second between Asp and Mal. The first is computed from the GDH exchange flux, whereas the second is calculated from theAsp dilution flux. *, p 0.05 for the comparison with untreated cells (white bars) or as indicated.

FIGURE 3. Ratio of M�1 to M�4 mass isotopomers of CAC intermediates. M�1/M�4 peak ratios of succi-nate, malate, and aspartate were determined by GC-MS and corrected for natural abundance using the methodof Fernandez et al. (46). Higher peak ratios indicate elevated CAC flux relative to glycolysis. Values are expressedas mean � S.E. (n � 3). *, p 0.05; **, p 0.01 for the comparison with untreated cells (white bars) or asindicated.

Metabolic Flux Analysis of Hepatic Lipoapoptosis

33430 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 284 • NUMBER 48 • NOVEMBER 27, 2009

at BIB

LIO D

E LA

SA

NT

E, on F

ebruary 23, 2010w

ww

.jbc.orgD

ownloaded from

http://www.jbc.org/content/suppl/2009/09/16/M109.049478.DC1.htmlSupplemental Material can be found at:

respectively. Amino acids promoted ROS generation in PA-cells, but ROS levels remained low in PA/OA-cells and did notrespond to changing amino acid concentrations (Fig. 6A). Low-ering medium amino acids below physiological levels signifi-cantly reduced ROS formation in PA-cells, returning ROS tothat of FFA-untreated cells (Fig. 6A). Similar effects wereobserved in stearate-treated H4IIEC3 cells and palmitate-treated mouse hepatocytes (data not shown). Furthermore,

oleate normalized amino acid-dependent caspase-3/7 activationand DNA fragmentation in palmi-tate-treated H4IIEC3 cells (Fig. 6B).Thus, these data indicate that aminoacid and oleate availabilities simul-taneously influence palmitate-in-duced apoptosis and steatosis.To assess the modulatory effects

of different classes of amino acidson the palmitate-induced apoptoticphenotype, H4IIEC3 cultures weresupplemented withNEAA and EAAmixtures as previously described.EAA supplementation partially nor-malized caspase-3/7 activity and cellviability in a dose-dependent man-ner, whereas NEAA supplementa-tion had no significant effect on

these readouts (Fig. 7).To analyze the relationship between cellular amino acid

pools and apoptosis, an elevated amount of each single aminoacid (800 �M) was added to low (“1⁄2�” � 1.08 mM) amino acidmedium in the presence or absence of palmitate. Low aminoacid medium slightly reduced cell viability independent ofpalmitate co-treatment (data not shown). In palmitate-un-treated cells, the addition of most single amino acidsimproved cell viability (Fig. 8A, top). However, in the pres-ence of palmitate, glutamate supplementation led to a signif-icant decrease in cell viability (Fig. 8A, bottom). In addition,glutamate significantly induced ROS generation andcaspase-3/7 activity (Fig. 8B). Several other NEAAs exhib-ited similar effects in the presence of palmitate, but theirimpacts were less pronounced and in most cases were notstatistically significant. These data suggest that the intracel-lular glutamate pool or possibly anaplerotic flux into theCAC via glutamate can be an important factor in controllingthe palmitate-induced apoptotic process.Influence of Amino Acids on FFA-induced Apoptosis Is Inde-

pendent of Ceramide—Ceramide, which is a known apoptosis-inducingmetabolite, is synthesized de novo from palmitate andserine. To examine the contribution of ceramide toward theapoptotic phenotype in PA-cells, we introduced fumonisin-B1and cycloserine, known ceramide synthesis inhibitors, to ourcultures and observed caspase-3/7 activity under different lev-els of amino acids. Increased medium amino acids enhancedintracellular palmitoyl-ceramide content in a dose-dependentmanner (Fig. 9A). Both ceramide synthesis inhibitors com-pletely suppressed C16 ceramide content (Fig. 9A) but failed torepress caspase-3/7 activation and ROS formation (Fig. 9B).These results suggest that amino acid availability impacts cer-amide levels in palmitate-treated cells but that the modulatoryeffect of amino acids on the apoptotic phenotype is not medi-ated by ceramide.

DISCUSSION

Themechanisms by which FFAs induce apoptosis in hepato-cytes are of importance because of their potential relevance to

FIGURE 4. Palmitate disturbs cytosolic redox in H4IIEC3 cells. A, lactate/pyruvate ratio was determined fromGC-MS analysis. B, cytosolic NAD�/NADH was calculated from the lactate/pyruvate values, assuming Keq �1.11 � 10�4 for lactate dehydrogenase. Values in A and B are expressed as mean � S.D. (n � 5). *, p 0.05 forthe comparison with untreated cells (white circles).

TABLE 2Fatty acid composition in triglyceridesData represent mean � S.D.

Fatty acidTreatment

None 400 �Mpalmitate

400 �M palmitate �50 �M oleate

nmol/mg proteinTotal lipids14:0 3.0 � 0.2a 6.5 � 0.7b 8.6 � 0.5b16:0 27.4 � 1.7a 114.3 � 8.2b 150.4 � 3.8c16:1n-7 3.5 � 0.2a 36.1 � 2.1b 36.6 � 1.5b18:0 27.0 � 1.5a 36.8 � 1.8b 41.3 � 1.4b18:1n-9 10.3 � 1.5a 18.1 � 1.3a 43.0 � 1.3b20:0 0.2 � 0.0 0.2 � 0.0 0.3 � 0.020:1 0.1 � 0.0 0.1 � 0.0 0.2 � 0.020:3 0.3 � 0.1 0.3 � 0.1 0.3 � 0.020:4 1.1 � 0.0a 1.7 � 0.1b 1.8 � 0.1b

Free lipids14:0 1.4 � 0.3a 3.3 � 0.3b 1.9 � 0.2a16:0 12.6 � 2.7a 31.4 � 2.7c 21.4 � 1.5b18:0 13.1 � 2.9a 31.8 � 1.9c 20.9 � 1.5b18:1n-9 1.7 � 0.5a 3.9 � 0.3b 4.3 � 0.6b20:0 0.3 � 0.1 0.8 � 0.1 0.5 � 0.120:4 0.01 � 0.01 0.04 � 0.02 0.02 � 0.01

Total free 29.0 � 6.5a 71.2 � 4.9c 49.0 � 3.4ba–c Means in columns without a common letter differ; p 0.05.

TABLE 3Cellular free glycerol and glycerol 3-phosphateMetabolites were measured at 12 h. Data represent mean � S.E.

Treatment

None 400 �Mpalmitate

400 �M palmitate �50 �M oleate

nmol/mg proteinGlycerol 3-phosphate 1.4 � 0.1a 1.8 � 0.1b 4.6 � 0.4c

Free glycerol 1.5 � 0.3a 2.5 � 0.4b 6.6 � 1.2ca–c Means in columns without a common letter differ; p 0.05.

Metabolic Flux Analysis of Hepatic Lipoapoptosis

NOVEMBER 27, 2009 • VOLUME 284 • NUMBER 48 JOURNAL OF BIOLOGICAL CHEMISTRY 33431

at BIB

LIO D

E LA

SA

NT

E, on F

ebruary 23, 2010w

ww

.jbc.orgD

ownloaded from

http://www.jbc.org/content/suppl/2009/09/16/M109.049478.DC1.htmlSupplemental Material can be found at:

non-alcoholic steatohepatitis (1, 35). Previous studies of FFA-induced apoptosis have focused on neutral lipid deposition (2),de novo ceramide synthesis (36), ROS generation (3), andrecently ER stress (12). ROS have a critical role in most pro-posed mechanisms; however, the mechanism of ROS genera-tion remains unclear. We analyzed metabolic pathways in pre-apoptotic cells to elucidate physiological events leading to ROSformation and apoptosis. In addition, we expected thatmanipulation of extracellular amino acid levels would pro-vide new insights into FFA-induced metabolic perturbationsbecause of their ability to modulate lipid homeostasis. Tothis end, we used GC-MS-based stable isotopic analysis ofmetabolic fluxes in central carbon pathways to uncover met-abolic disturbances preceding the onset of apoptotic or stea-totic phenotypes. Our results clearly point to the activationof CAC fluxes by palmitate (PA-cells) concomitant withreduced glycolysis and increased cytosolic NAD�/NADHratio. The decoupling between glycolysis and CAC fluxeswas normalized by oleate co-supplementation (PA/OA-cells). Also, it was found that amino acids play a modulatoryrole in these processes, possibly through the control of cen-tral carbon fluxes. A schematic showing the proposed path-ways of FFA-induced apoptosis is presented in Fig. 10.A recent study demonstrated the application of metabolic

flux analysis to long term palmitate-treated human hepatomaHepG2 cells using conventional stoichiometric modeling withmedium metabolite analysis (37). According to this work, 1–3days of palmitate treatment failed to change glycolysis andCAC

fluxes, whereas increased �-oxida-tion and ketone body formationwere observed (37). However, werealized that there are at least twopoints to consider in regard to thisanalysis. First, FFA-induced apop-tosis normally completes within24 h, culminating in cell death.Therefore, the time scale of theaforementioned experiment wastoo long to observe changes in theflux phenotype that precede theonset of apoptosis. Second, oncecellshave undergone apoptosis, they mayrelease intracellular metabolitesinto the medium. This could invali-date measurements of extracellularsubstrate depletion and productaccumulation that are used in thesubsequent flux analysis. In thepresent study, we demonstrate met-abolic flux analysis of central carbonpathways using 13C5 glutamine as alabeled substrate from short (3 h) tolong (9 h) periods in both PA- andPA/OA-cells. However, the longtime points still precede the onset ofapoptosis. Our data indicate thatpalmitate can induce decoupling ofcentral metabolism at 9 h by inhib-

FIGURE 5. Amino acid levels affect FFA-induced metabolic alterations inH4IIEC3 cells. Effect of medium EAA and NEAA availability on palmitate-treated cells. A, external fluxes expressed as �mol/h/106 cells. B, internal netfluxes relative to a PK flux of 100. Metabolic flux analysis of central carbonpathways were determined using [U-13C5]glutamine as described in the leg-end to Fig. 2. H4IIEC3 cells were incubated with 400 �M palmitate. MediumEAA (0.8 mM) and NEAA (1.35 mM) mixtures were prepared based on thephysiological amino acid compositions presented in Table 1 and supple-mented into amino acid-free medium. *, p 0.05 for the comparison withcontrol PA-treated cells (black bar or symbols). ACL, ATP-citrate lyase; CS, cit-rate synthase; FUM, fumarase; GDH, glutamate dehydrogenase; ME, malicenzyme; SDH, succinate dehydrogenase; LDH, lactate dehydrogenase.

FIGURE 6. Amino acid levels affect FFA-induced ROS generation and apoptosis in H4IIEC3 cells. Influenceof amino acid availability on FFA-induced ROS generation (A) and also caspase-3/7 activity and DNA laddering(B). H4IIEC3 cells were incubated with 400 �M palmitate both with and without 50 �M oleate and under varyingconcentrations of amino acids (1.08, 2.15, or 8.6 mM). DCF fluorescence was determined after a 12-h incubation,whereas caspase-3/7 activity and DNA laddering were measured after a 24-h incubation. In the electrophoresisgel images (B, right panels), the label of each lane corresponds to the conditions in the bar graph of caspase-3/7activity (B, left panels) except for the size marker (M). Amino acid concentrations are expressed as a magnifica-tion ratio in comparison with their physiological level (2.15 mM). Values are expressed as mean � S.E. (n � 3). *,p 0.05; **, p 0.01 for the comparison with untreated cells (white bars) or as indicated.

Metabolic Flux Analysis of Hepatic Lipoapoptosis

33432 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 284 • NUMBER 48 • NOVEMBER 27, 2009

at BIB

LIO D

E LA

SA

NT

E, on F

ebruary 23, 2010w

ww

.jbc.orgD

ownloaded from

http://www.jbc.org/content/suppl/2009/09/16/M109.049478.DC1.htmlSupplemental Material can be found at:

iting glycolysis and activating CAC fluxes. Furthermore,increased CAC flux seems to derive from increased glutamineuptake rather than increased fatty acid oxidation. This mightexplain the finding of Srivistava andChan (37) that inhibition of�-oxidation in HepG2 cells failed to suppress palmitatelipotoxicity.Oleate co-treatment has been reported to prevent palmitate-

induced apoptosis while promoting steatosis by divertingpalmitate away from apoptotic pathways and toward triglycer-ide synthesis (2). However, the mechanism by which oleatechanges themetabolic fate of palmitate is still unclear. Our datareveal that decoupling of glycolysis and CAC fluxes by palmi-tate is normalized by oleate co-treatment in PA/OA-cells (Fig.10). Particularly, diminished lactate dehydrogenase flux in PA-cells was restored in PA/OA-cells. Our data also reveal thatoleate co-treatment increases the intracellular glycerol 3-phos-phate and free glycerol pools, which are required for triglycer-ide synthesis. Both the production of lactate by lactate dehydro-genase and the production of glycerol-3-phosphate fromDHAP depend upon cytosolic NADH. Thus, our data indicate

that restoration of glycolytic fluxes by oleate plays a critical rolein promoting triglyceride synthesis and steatosis while reduc-ing apoptosis in PA/OA-cells, at least in part, by normalizingthe cytosolic redox state (Fig. 10).Studies of glycolysis inhibition by FFAs in hepatocytes have

been previously reported (38, 39). An earlier study showed thatFFA treatment in rat hepatocytes inhibits pyruvate formationfrom glucose, suggesting that FFA oxidation competes withglucose utilization (38). The authors attributed this result to theinhibitory effect of increased citrate levels on phosphofructoki-nase activity. They argue that increased flux to acetyl-CoAderived from fatty acid oxidation will not only produce theobserved increases in citrate concentration but will alsoenhance flux through pyruvate carboxylase, thus shifting thecell toward a more gluconeogenic state. Our data indicate thatpalmitate treatment could increase not only pyruvate carbox-ylase flux relative to glycolysis but also many other CAC andanaplerotic fluxes. Another recent study examined the influ-ence of FFAs on glucokinase activity and glycolysis in rat hepa-tocytes (39). In that study, 150 �M palmitate lowered the glu-cokinase activity of rat hepatocytes in the presence of 2 nMinsulin. Oleate had no effect on glucokinase activity, indicatingthat the influence of FFAs on glycolysis is species-dependent. Inthe same study, palmitate-treated hepatocytes did not showchanges in the overall glycolytic rate in response to glucokinasedown-regulation. However, our hypothesis is still reasonable,considering the fact that treatment of H4IIEC3 cells with lowerpalmitate levels (100 �M) had no effect on cellular phenotypesin our study. Higher palmitate levels may be required to reduceglycolysis in hepatocytes as well. Inherent differences betweenhepatocyte and hepatoma metabolisms should also be consid-ered in this regard. Hansson et al. (40) compared glucose andFFA metabolism in McA-RH7777 hepatomas and rat hepato-cytes using isotopic tracers. They showed that although FFAuptake was higher in hepatomas than in hepatocytes, �-oxida-tion flux was lower in hepatomas. In the same study, glucoseuptake was lower, but its oxidation was higher in hepatomas.Also, triglyceride synthesis was higher in hepatomas. Consid-ering these results, hepatomas might be more sensitive to freelipid toxicity because of their weak capacity for FFA oxidation.In addition, hepatomas seem to depend on glucose, rather thanfatty acids, as their major energy source, indicating the impor-tance of glycolysis in hepatomametabolism. In fact, our prelim-inary data show that induction of cell death and ROS formationby palmitate is much stronger in hepatomas than in hepato-cytes (supplemental Fig. S1).Although the experiments presented in this article were con-

ducted in high glucose medium, in preliminary studies weexamined both low (5 mM) and high (25 mM) concentrationsand did not observe differences in the ability of palmitate toinduce ROS and cell death (data not shown). Furthermore, aprevious study on myocyte cells showed no difference betweenlow and high glucose in regard to palmitate-induced apoptosis(4). Thus, the effects of palmitate do not appear to stem fromglucose toxicity or to depend strongly on changes in glucoseconcentration.EAAs affect fatty acid synthase genes inHepG2 cells (41), and

recent studies show systematic regulation of lipid pathways by

FIGURE 7. NEAA, but not EAA, promotes FFA-induced apoptosis inH4IIEC3 cells. Shown is the effect of medium NEAA and EAA levels on palmi-tate-induced apoptosis. H4IIEC3 cells were incubated with or without 400 �M

palmitate under varying amino acid compositions for 24 h, and then cell via-bility and caspase-3/7 activity were determined. Medium EAA (0.8 mM) andNEAA (1.35 mM) mixtures were prepared based on the physiological aminoacid compositions presented in Table 1 and supplemented into amino acid-free medium according to the indicated magnifications. Values are expressedas mean � S.E. (n � 3). *, p 0.05; **, p 0.01 for the comparison withuntreated cells (white bars) or as indicated.

Metabolic Flux Analysis of Hepatic Lipoapoptosis

NOVEMBER 27, 2009 • VOLUME 284 • NUMBER 48 JOURNAL OF BIOLOGICAL CHEMISTRY 33433

at BIB

LIO D

E LA

SA

NT

E, on F

ebruary 23, 2010w

ww

.jbc.orgD

ownloaded from

http://www.jbc.org/content/suppl/2009/09/16/M109.049478.DC1.htmlSupplemental Material can be found at:

EAA through a GCN2-dependent pathway (15). Furthermore,amino acids are known to affectmetabolic fluxes in glucose andlipid pathways (18, 42). It was reported that the NEAAs gluta-mine, alanine, and proline promote de novo lipogenesis in hepa-tocyte and mouse hepatoma cells (42). Thus, the influence ofamino acids on lipid metabolism and related metabolic disor-ders is an issue of current importance. However, there has notbeen a systematic study investigating how amino acids affectlipid toxicity. In the present study, we focused on the role ofamino acids in controlling metabolic fluxes, rather than geneexpression, in lipoapoptotic cells because we assumed thatbreakdown of metabolic homeostasis by FFAs could triggerROS generation and apoptosis and also that these effects couldbe modulated by the medium amino acid composition. In pre-liminary studies, we examined both EAA and NEAA supple-mentation in FFA-untreated H4IIEC3 cells but did not observechanges in ROS production or cell viability (data not shown).However, our present data reveal that increasing the totalamino acid concentration promotes ROS generation and apop-tosis in palmitate-treated cells. Furthermore, NEAA supple-mentation was shown to exacerbate the palmitate-induceddecoupling of glycolysis and CAC fluxes, whereas EAA supple-mentation protected against these effects. Thus, the totalamino acid concentration and, in particular, the abundance ofNEAA relative to EAA appear to modulate the metabolic andapoptotic effects of palmitate. Induction of ROS generation byNEAA was also confirmed in mouse primary hepatocytes (seesupplemental Fig. S1).We further examined the effects of single amino acid over-

loading in palmitate-treated cells and found that glutamate wasthe most effective single amino acid in promoting palmitate-

induced ROS generation. Of course,it is unlikely that such high extracel-lular glutamate availability isdirectly involved in FFA-inducedapoptosis in vivo due to its lowerabundance in circulation. However,it is possible that altered amin-otransferase activities or changes inother amino acid catabolic path-ways might affect the intracellularglutamate pool and thereby promoteFFA-induced apoptosis. It should benoted that the effect of glutamate topromote ROS generation andcaspase activation and to reduce cellviability only occurred in palmitate-treated cells, indicating that theeffect is dependent on palmitate.As for EAA, hepatocytes are

known to have unique limitations intheir amino acid catabolic path-ways; for instance, they have no oronly weak branched-chain aminoacid aminotransferase activities(43). In addition, lysine, the mostabundant ketogenic amino acid, isoxidized only in the mitochondrial

matrix (44) and is expected to compete with FFA oxidation. Ofcourse, our present data and the previously reported studies arenot enough to explain differences in the efficacy of each aminoacid, but it is highly likely that the anaplerotic flux partitioningof amino acids into the CAC could be an important factor.To investigate the contribution of palmitate oxidation to

central pathways, we also employed metabolic flux analysisusing 13C16 palmitate inH4IIEC3 cells (data not shown). Only avery small amount of labeling, however, was detected in CACintermediates, indicating that the oxidation of extracellularpalmitate was not a major contributor to the enhanced CACfluxes we observed. On the contrary, our data indicate that theeffects of palmitate to increase CAC activity are mediated pri-marily through increased anaplerotic flux from glutamine. Thislends further support to the hypothesis that differences inanapleroticmetabolismmay be a key factor behind the ability ofcertain classes of amino acids tomodulate lipotoxicity in differ-ent ways. In addition, isotopic exchange from �-ketoglutarateto glutamate and from malate to aspartate was significantlylower in both PA- and PA/OA-cells. This indicates that FFAtreatments could reduce glutamate and aspartate exchangepathways, such as the aspartate/malate shuttle. Therefore, theconsequences of reduced aspartate/malate shuttling to altercytosolic NAD�/NADHmight be another possible explanationfor the observed effects of amino acid availability.Ceramide synthesis inhibitors had no effect on palmitate-

induced ROS generation and caspase-3/7 activation underincreased levels of medium amino acids, although they com-pletely suppressed cellular ceramide. Where NEAA loadingeffectively increased cellular ceramide, EAA failed to bringabout a change (data not shown). This indicates that cellular

FIGURE 8. Cellular glutamate metabolism may be important in FFA-induced ROS generation and apop-tosis. Shown is the effect of single amino acid overloading on palmitate-induced ROS generation and apop-tosis. Each single amino acid (800 �M) was supplemented into 0.5� basal amino acid medium (1.08 mM) with400 �M palmitate. After a 12-h incubation, cell viability (A), ROS generation (B, top), and caspase-3/7 activity(B, bottom) were determined (n � 6). White, gray, and black bars, vehicle-, EAA-, and NEAA-treated cells, respec-tively. Values are expressed as mean � S.E. *, p 0.05; **, p 0.01 for the comparison with vehicle condition(white bar).

Metabolic Flux Analysis of Hepatic Lipoapoptosis

33434 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 284 • NUMBER 48 • NOVEMBER 27, 2009

at BIB

LIO D

E LA

SA

NT

E, on F

ebruary 23, 2010w

ww

.jbc.orgD

ownloaded from

http://www.jbc.org/content/suppl/2009/09/16/M109.049478.DC1.htmlSupplemental Material can be found at:

ceramide is influenced bymediumamino acid availability, but itis not involved in both palmitate- and amino acid-induced phe-notypic changes, at least in our model. Therefore, our data sug-gest that palmitate toxicity can bemodulated by amino acids aswell as oleate, not through ceramide synthesis but possibly

through activation of CAC fluxes (Fig. 10). This suggests thatmanipulation of dietary amino acids could be a possible way totreat NAFLD and non-alcoholic steatohepatitis.Only a fewmetabolomic studies have been reported inmam-

malian cells (45). This may be due to practical difficulties sur-rounding sample preparation and data analysis. Our laboratoryhas previously developed two computational programs to facil-itate metabolomic data analysis: SpectConnect for GC-MS-based metabolic profiling (26) and Metran for stable isotopicflux analysis (29). In the present study, we applied these toolsfor pathway analysis in lipoapoptotic cells and successfullyidentified phenotype-specific changes. We believe that ourmethods can be powerful tools for metabolic discovery in vari-ous mammalian cell models.

Acknowledgments—We thank Dr. Takehana and Ajinomoto Co., Inc.for the gift of amino acid-free DMEM. Also, we thank Dr. Kimura andDr. Ando for human and rat plasma amino acid profiles.

REFERENCES1. Wieckowska, A.,McCullough, A. J., and Feldstein, A. E. (2007)Hepatology

46, 582–5892. Listenberger, L. L., Han, X., Lewis, S. E., Cases, S., Farese, R. V., Jr., Ory,

D. S., and Schaffer, J. E. (2003)Proc. Natl. Acad. Sci. U.S.A. 100, 3077–30823. Listenberger, L. L., Ory, D. S., and Schaffer, J. E. (2001) J. Biol. Chem. 276,

14890–148954. Miller, T. A., LeBrasseur, N. K., Cote, G. M., Trucillo, M. P., Pimentel,

D. R., Ido, Y., Ruderman, N. B., and Sawyer, D. B. (2005) Biochem. Biophys.Res. Commun. 336, 309–315

5. Okere, I. C., Chandler, M. P., McElfresh, T. A., Rennison, J. H., Sharov, V.,Sabbah, H. N., Tserng, K. Y., Hoit, B. D., Ernsberger, P., Young, M. E., andStanley, W. C. (2006) Am. J. Physiol. Heart Circ. Physiol. 291, H38–H44

6. Busch, A. K., Gurisik, E., Cordery, D. V., Sudlow, M., Denyer, G. S., Lay-butt, D. R., Hughes, W. E., and Biden, T. J. (2005)Diabetes 54, 2917–2924

FIGURE 9. Influence of amino acid availability on FFA-induced apoptosisis independent of ceramide. A, the influence of medium amino acids and denovo ceramide synthesis inhibition on intracellular palmitoyl-ceramide inH4IIEC3 cells. Ceramide was determined by GC-MS analysis as describedunder “Experimental Procedures.” Values are expressed as mean � S.E. (n �3). *, p 0.05 for the comparison with untreated cells. B, effect of de novoceramide synthesis inhibitors on apoptosis in palmitate-treated cells. Cellswere incubated with either 1 mM cycloserine or 50 �M fumonisin-B1 undervarying amino acid levels with or without 400 �M palmitate. DCF fluorescenceand caspase-3/7 activity were measured. Values are expressed as mean � S.E.(n � 4). Amino acid concentrations are expressed as a magnification ratio incomparison with physiological amino acids (2.15 mM). **, p 0.01 for thecomparison with each palmitate-untreated control (white bars).

FIGURE 10. Proposed metabolic pathways causing SFA-induced apopto-sis and loci of amino acid control. SFAs inhibit glycolysis flux and promotethe decoupling of glycolysis and CAC (TCA cycle) fluxes associated withdecreased cytosolic NAD�/NADH. NEAAs, such as glutamate, stronglyenhance this process, probably through CAC activation. High cytosolicNAD�/NADH prevents lactate and glycerol 3-phosphate formation frompyruvate and DHAP, respectively. The decreased glycerol pool might alsoreduce triglyceride synthesis and increase FFA accumulation and ceramidesynthesis. Ceramide synthesis is dependent upon non-essential amino acidavailability but is not involved in apoptotic pathways. The MUFA oleate canprevent SFA-induced decoupling of glycolysis and CAC fluxes and normalizecytosolic redox possibly by reversing glycolysis inhibition.

Metabolic Flux Analysis of Hepatic Lipoapoptosis

NOVEMBER 27, 2009 • VOLUME 284 • NUMBER 48 JOURNAL OF BIOLOGICAL CHEMISTRY 33435

at BIB

LIO D

E LA

SA

NT

E, on F

ebruary 23, 2010w

ww

.jbc.orgD

ownloaded from

http://www.jbc.org/content/suppl/2009/09/16/M109.049478.DC1.htmlSupplemental Material can be found at:

7. Hardy, S., El-Assaad, W., Przybytkowski, E., Joly, E., Prentki, M., and Lan-gelier, Y. (2003) J. Biol. Chem. 278, 31861–31870

8. Barreyro, F. J., Kobayashi, S., Bronk, S. F., Werneburg, N. W., Malhi, H.,and Gores, G. J. (2007) J. Biol. Chem. 282, 27141–27154

9. Malhi, H., Bronk, S. F., Werneburg, N. W., and Gores, G. J. (2006) J. Biol.Chem. 281, 12093–12101

10. Unger, R. H., and Zhou, Y. T. (2001) Diabetes 50, Suppl. 1, S118–S12111. Zhou, Y. T., Grayburn, P., Karim, A., Shimabukuro,M., Higa, M., Baetens,

D., Orci, L., and Unger, R. H. (2000) Proc. Natl. Acad. Sci. U.S.A. 97,1784–1789

12. Guo, W., Wong, S., Xie, W., Lei, T., and Luo, Z. (2007) Am. J. Physiol.Endocrinol. Metab. 293, E576–E586

13. Wei, Y., Wang, D., Topczewski, F., and Pagliassotti, M. J. (2006) Am. J.Physiol. Endocrinol. Metab. 291, E275–E281

14. Zivkovic, A.M.,German, J. B., and Sanyal, A. J. (2007)Am. J. Clin. Nutr.86,285–300

15. Guo, F., and Cavener, D. R. (2007) Cell Metab. 5, 103–11416. Kimball, S. R., and Jefferson, L. S. (2006) Am. J. Clin. Nutr. 83, 500S–507S17. Chan, C., Berthiaume, F., Lee, K., and Yarmush, M. L. (2003)Metab. Eng.

5, 1–1518. Wong, M. S., Raab, R. M., Rigoutsos, I., Stephanopoulos, G. N., and Kelle-

her, J. K. (2004) Physiol. Genomics 16, 247–25519. Jousse, C., Bruhat, A., Ferrara, M., and Fafournoux, P. (1998) Biochem. J.

334, 147–15320. Sudhandiran, G., and Shaha, C. (2003) J. Biol. Chem. 278, 25120–2513221. Antoniewicz, M. R., Kelleher, J. K., and Stephanopoulos, G. (2007) Anal.

Chem. 79, 7554–755922. Fiehn, O., Kopka, J., Dormann, P., Altmann, T., Trethewey, R. N., and

Willmitzer, L. (2000) Nat. Biotechnol. 18, 1157–116123. Tserng, K. Y., and Griffin, R. (2003) Anal. Biochem. 323, 84–9324. Vieu, C., Terce, F., Chevy, F., Rolland, C., Barbaras, R., Chap, H., Wolf, C.,

Perret, B., and Collet, X. (2002) J. Lipid Res. 43, 510–52225. Siess, E. A., Brocks, D. G., Lattke, H. K., and Wieland, O. H. (1977) Bio-

chem. J. 166, 225–23526. Styczynski,M. P.,Moxley, J. F., Tong, L. V.,Walther, J. L., Jensen, K. L., and

Stephanopoulos, G. N. (2007) Anal. Chem. 79, 966–97327. Tikunov, Y., Lommen, A., de Vos, C. H., Verhoeven, H. A., Bino, R. J., Hall,

R. D., and Bovy, A. G. (2005) Plant Physiol. 139, 1125–113728. Wiechert,W., Mollney, M., Isermann, N.,Wurzel, M., and de Graaf, A. A.

(1999) Biotechnol. Bioeng. 66, 69–8529. Yoo,H., Antoniewicz,M. R., Stephanopoulos, G., andKelleher, J. K. (2008)

J. Biol. Chem. 283, 20621–2062730. Antoniewicz, M. R., Kraynie, D. F., Laffend, L. A., Gonzalez-Lergier, J.,

Kelleher, J. K., and Stephanopoulos, G. (2007)Metab. Eng. 9, 277–29231. Young, J. D., Walther, J. L., Antoniewicz, M. R., Yoo, H., and Stephano-

poulos, G. (2008) Biotechnol. Bioeng. 99, 686–69932. Madsen, K., Nielsen, H. B., and Tingleff, O. (2004)Methods for Non-linear

Least Squares Problems, 2nd Ed., pp. 24–29, Technical University of Den-mark, Lyngby, Denmark

33. Antoniewicz, M. R., Kelleher, J. K., and Stephanopoulos, G. (2006)Metab.Eng. 8, 324–337

34. Wiechert, W., Siefke, C., de Graaf, A. A., and Marx, A. (1997) Biotechnol.Bioeng. 55, 118–135

35. Farrell, G. C., and Larter, C. Z. (2006)Hepatology 43, Suppl. 1., S99–S11236. Turpin, S. M., Lancaster, G. I., Darby, I., Febbraio, M. A., and Watt, M. J.

(2006) Am. J. Physiol. Endocrinol. Metab. 291, E1341–E135037. Srivastava, S., and Chan, C. (2008) Biotechnol. Bioeng. 99, 399–41038. Morand, C., Besson, C., Demigne, C., and Remesy, C. (1994) Arch. Bio-

chem. Biophys. 309, 254–26039. Swagell, C. D., Morris, C. P., and Henly, D. C. (2006) Nutrition 22,

672–67840. Hansson, P. K., Asztely, A. K., Clapham, J. C., and Schreyer, S. A. (2004)

Biochim. Biophys. Acta 1684, 54–6241. Dudek, S. M., and Semenkovich, C. F. (1995) J. Biol. Chem. 270,

29323–2932942. Baquet, A., Lavoinne, A., and Hue, L. (1991) Biochem. J. 273, 57–6243. Suryawan, A., Hawes, J. W., Harris, R. A., Shimomura, Y., Jenkins, A. E.,

and Hutson, S. M. (1998) Am. J. Clin. Nutr. 68, 72–8144. Blemings, K. P., Crenshaw, T. D., Swick, R.W., and Benevenga, N. J. (1994)

J. Nutr. 124, 1215–122145. Ramanathan, A., Wang, C., and Schreiber, S. L. (2005) Proc. Natl. Acad.

Sci. U.S.A. 102, 5992–599746. Fernandez, C. A., Des Rosiers, C., Previs, S. F., David, F., and Brunengra-

ber, H. (1996) J. Mass Spectrom. 31, 255–262

Metabolic Flux Analysis of Hepatic Lipoapoptosis

33436 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 284 • NUMBER 48 • NOVEMBER 27, 2009

at BIB

LIO D

E LA

SA

NT

E, on F

ebruary 23, 2010w

ww

.jbc.orgD

ownloaded from

http://www.jbc.org/content/suppl/2009/09/16/M109.049478.DC1.htmlSupplemental Material can be found at:

Supplementary Materials

Lipotoxicity in mouse hepatocytes

mouse hepatocyte

B

AH4IIEC3

Figure S-1: Influence of amino acid availability on FFA-induced lipid accumulation, cellviability and ROS generation in mouse hepatocytes. (A) H4IIEC3 cells and mouse hepato-cytes were incubated with 400 mM palmitate or without FFA. Also, cells were incubatedwith 400 mM palmitate with 50 mM oleate. ROS and cell viability were determined usingDCF fluorescence and resazurin reduction, respectively. Values are expressed as mean +/-SEM (n = 3). (B) Mouse hepatocytes were incubated with or without 400 mM palmitateunder varying amino acid compositions for 24h. DCF fluorescence was determined. MediumEAA (=0.8 mM) and NEAA (=1.35 mM) mixtures were prepared based on the physiologi-cal amino acid compositions presented in Table 1 and added to amino-acid-free medium asindicated by the magnification factor. In the bar graph, values are expressed as mean +/-SEM (n = 3). *p < 0.05 and **p < 0.01 for the comparison with untreated cells (white bar)or as indicated.

1

Metabolic flux analysis (MFA)

The metabolic network model includes the reactions listed in Table S-1 and is based on the

following assumptions:

1. The intracellular labeling is at isotopic quasi-steady state. As a result, metabolite pool

size data are not required by the model. Even though metabolism is changing over

time, these changes are slow relative to the isotope labeling dynamics. By sampling

intermediate pools that label rapidly, we are able to apply steady-state MFA at each

time point to obtain a series of snapshots that describe the trajectory of flux changes.

2. Cell growth is negligible over the course of the experiment.

3. There is no reincorporation of labeled CO2.

4. Fatty acid oxidation is negligible. This assumption was confirmed by the inability of

[U-13C]palmitate to produce significant labeling of central carbon metabolites (data

not shown).

5. Succinate and fumarate are symmetric molecules that do not retain any particular

orientation when metabolized by TCA cycle enzymes.

6. Extracellular glucose and lactate are assumed to be unlabeled throughout the course

of the experiment, as well as glutamine derived from endogenous protein turnover.

7. Exchange between free and protein-bound glutamine occurs with zero net flux in either

direction. This allows for amino acid turnover without net synthesis or degradation of

protein.

8. Mitochondrial malic enzyme provides the dominant route by which malate is oxidized

to pyruvate. This enzyme is known to be highly active in several different hepatoma

cell lines (1).

9. Separate cytosolic and mitochondrial pools of pyruvate are included in the model. All

other metabolites are considered to be isotopically equilibrated between these compart-

2

Table S-1: A complete list of reactions and atom transitions for the rat hepatoma metabolicnetwork. Dot suffixes differentiate between separate pools of a given metabolite and make use ofthe following notation: .c, cytosolic; .d, dilution (unlabeled); .m, mitochondrial; .p, protein; .x,extracellular. Any species that does not have a dot suffix is either (i) only present in a single poolor (ii) corresponds to a combined intracellular pool. Note that the dilution fluxes are not actualmetabolic fluxes but are pseudo-fluxes introduced to describe the mixing of compartmentalizedpools that occurs upon extraction of intracellular metabolites (see text for details). Refer to theNomenclature section for a list of metabolite abbreviations.

GlycolysisGlc (abcdef) −→ G6P (abcdef)G6P (abcdef) −→ Lost to glycogen or oxPPPG6P (abcdef) −→ T3P (cba) + T3P (def)T3P (abc) −→ Pyr.c (abc)

Pyruvate metabolismPyr.c (abc) ←→ Lac (abc)Pyr.m (abc) + CO2 (d) −→ Mal (abcd)Pyr.c (abc) −→ AcCoA (bc) + CO2 (a)Mal (abcd) −→ Pyr.m (abc) + CO2 (d)Pyr.m (abc) ←→ Pyr.c (abc)

TCA cycleAcCoA (ab) + Mal (cdef) −→ Cit.m (fedbac)Cit.m (abcdef) ←→ Akg (abcde) + CO2 (f)Akg (abcde) −→ Suc.m (1/2 bcde + 1/2 edcb) + CO2 (a)Suc.m (1/2 abcd + 1/2 dcba) ←→ Fum (1/2 abcd + 1/2 dcba)Fum (1/2 abcd + 1/2 dcba) ←→ Mal (abcd)

Amino acid metabolismGln.x (abcde) −→ Gln (abcde)Gln.p (abcde) ←→ Gln (abcde)Gln (abdce) −→ Glu (abcde)Glu (abcde) ←→ Akg (abcde)

Citrate export and cleavageCit.m (fedbac) −→ Fatty acids (ab) + Mal (cdef)

Pool dilution/mixingCit.d (abcdef) −→ Cit (abcdef)Cit.m (abcdef) −→ Cit (abcdef)Suc.d (abcd) −→ Suc (abcd)Suc.m (abcd) −→ Suc (abcd)Pyr.c (abc) −→ Pyr (abc)Pyr.m (abc) −→ Pyr (abc)Ala.d (abc) −→ Ala (abc)Pyr.m (abc) −→ Ala (abc)Asp.d (abcd) −→ Asp (abcd)Mal (abcd) −→ Asp (abcd)

3

ments or, alternatively, to only exist in a single compartment. Although the model

includes multiple pyruvate pools with possibly different mass isotopomer distributions

(MIDs), it does not necessitate data for each compartment. Rather, we define addi-

tional model parameters that represent the relative contribution from each compart-

ment toward the observed labeling. These so-called “dilution” or “mixing” fluxes mimic

the commingling of segregated pools that occurs upon extraction and sampling of in-

tracellular metabolites. The model accounts for dilution fluxes in a way that does not

perturb the flux distribution in the remainder of the network, since these pseudo-fluxes

do not participate directly in metabolism.

10. The labeling of both citrate and succinate does not appear to be equilibrated with

the remaining TCA cycle metabolites in some experiments. In the case of citrate,

this only occurred when fatty acids were added to the medium. A substantial amount

of unlabeled citrate was detected in the growth medium of these cultures, which was

likely present in the BSA used to solubilize the fatty acids. This unlabeled citrate

would therefore provide a source of isotope dilution. Succinate is often found to have

decreased enrichment in tracer studies, possibly due to subcellular compartmentation

(2). Lack of equilibration in both citrate and succinate is handled by introducing a

corresponding dilution flux into the model for each metabolite.

11. Dilution fluxes are also employed to model the lack of equilibration in the free amino

acid pools of alanine and aspartate. This accounts for slow labeling due to exchange

with unlabeled pools in protein and in the growth medium.

12. Cytosolic pyruvate is assumed to be the substrate of the pyruvate dehydrogenase

(PDH) enzyme. This assumption is supported by evidence from membrane associ-

ation studies showing that PDH binds to the inner mitochondrial membrane (3). Con-

versely, the malic enzyme (ME) and pyruvate carboxylase (PC) reactions are assumed

to communicate only with the mitochondrial pyruvate pool. These assumptions are

necessary to explain a particular feature of the GC-MS data, namely the presence of

M+3 pyruvate that is formed from decarboxylation of fully labeled malate via ME.

4

Table S-2: Measurements used for flux determination. (a) Extracellular concentrations ofglucose, lactate and glutamine were measured enzymatically and used to determine extra-cellular uptake and secretion rates. (b) GC-MS ions used to assess metabolite labeling.Standard errors of measurement (SEM) were determined based on the lack of agreementbetween measured and theoretically computed mass isotopomer distributions obtained fromunlabeled cell extracts.

(a) Measured external fluxesMetabolite Net FluxGlucose Glc −→ G6PLactate Pyr.c −→ LacGlutamine Gln.x −→ Gln

(b) Measured GC-MS ionsMetabolite Mass Carbons Composition SEM, mol%Pyr 174 1 2 3 C6H12O3NSi 0.5Ala 232 2 3 C10H26ONSi2 0.2Ala 260 1 2 3 C11H26O2NSi2 0.3Suc 289 1 2 3 4 C12H25O4Si2 0.3Mal 419 1 2 3 4 C18H39O5Si3 0.5Asp 302 1 2 C14H32O2NSi2 0.3Asp 390 2 3 4 C17H40O3NSi3 0.4Asp 418 1 2 3 4 C18H40O4NSi3 0.6Glu 330 2 3 4 5 C16H36O2NSi2 0.3Glu 432 1 2 3 4 5 C19H42O4NSi3 0.4Gln 431 1 2 3 4 5 C19H43N2O3Si3 0.4Cit 459 1 2 3 4 5 6 C20H39O6Si3 0.6

If this (M+3)-labeled pyruvate were metabolized by PDH to become (M+2)-labeled

AcCoA, it would result in excessive M+2 labeling of TCA cycle intermediates beyond

a level that can be reconciled with the GC-MS measurements. Therefore, the AcCoA

must be formed from a (cytosolic) pyruvate pool that is mostly unlabeled and does not

reflect the high M+3 labeling of the measured pyruvate pool.

The measurements used for flux determination are listed in Table S-2. Metabolic fluxes

were estimated by minimizing the lack-of-fit between measured and simulated MIDs using

least-squares regression. To ensure that the final solution was the global optimum, flux

estimation was repeated at least ten times starting from random initial values. Figs. S-2

through S-4 show the experimental GC-MS measurements for each condition described in

Fig. 2 of the main article (9h time points only) along with best-fit model simulations. Each

5

fit was statistically accepted based on a χ2 test at the 95% confidence level. A full listing of

the optimal parameter estimates can be found in Tables S-4 through S-6.

In Fig. 5 of the main article, where the effects of amino acid concentrations on metabolic

fluxes were investigated, the Pyr174 and Gln431 ions could not be reliably measured due to

low abundance. Therefore, a simplified MFA model was applied to assess fluxes based on this

reduced measurement set. The reactions in the simplified network are shown in Table S-3,

and best-fit model simulations and optimal parameter estimates can be found in Figs. S-5

through S-7 and Tables S-7 through S-9, respectively.

Nomenclature

AcCoA, acetyl-CoA; Akg, alpha-ketoglutarate; Ala, alanine; Asp, aspartate; Cit, citrate;

Fum, fumarate; G6P, glucose-6-phosphate; Glc, glucose; Gln, glutamine; Glu, glutamate;

Lac, lactate; Mal, malate; oxPPP, oxidative pentose phosphate pathway; Pyr, pyruvate;

Suc, succinate; T3P, triose-3-phosphate.

References

[1] Moreadith, R. W. and Lehninger, A. L. (1984) J. Biol. Chem. 259, 6215–21.

[2] Chatham, J. C., Bouchard, B., and Des Rosiers, C. (2003) Mol. Cell. Biochem. 249,

105–112.

[3] Sumegi, B. and Srere, P. A. (1984) J. Biol. Chem. 259(24), 15040–5.

6

Table S-3: Simplified network used to investigate the effects of amino acid concentrations onmetabolic fluxes.

GlycolysisGlc (abcdef) −→ G6P (abcdef)G6P (abcdef) −→ Lost to glycogen or oxPPPG6P (abcdef) −→ T3P (cba) + T3P (def)T3P (abc) −→ Pyr (abc)

Pyruvate metabolismPyr (abc) ←→ Lac (abc)Pyr (abc) + CO2 (d) −→ Mal (abcd)Pyr (abc) −→ AcCoA (bc) + CO2 (a)Mal (abcd) −→ Pyr (abc) + CO2 (d)

TCA cycleAcCoA (ab) + Mal (cdef) −→ Cit (fedbac)Cit (abcdef) ←→ Akg (abcde) + CO2 (f)Akg (abcde) −→ Suc (1/2 bcde + 1/2 edcb) + CO2 (a)Suc (1/2 abcd + 1/2 dcba) ←→ Fum (1/2 abcd + 1/2 dcba)Fum (1/2 abcd + 1/2 dcba) ←→ Mal (abcd)

Amino acid metabolismGln.x (abcde) −→ Gln (abcde)Gln (abdce) −→ Glu (abcde)Glu (abcde) ←→ Akg (abcde)

Citrate export and cleavageCit (fedbac) −→ Fatty acids (ab) + Mal (cdef)

Pool dilution/mixingAla.d (abc) −→ Ala (abc)Pyr (abc) −→ Ala (abc)Asp.d (abcd) −→ Asp (abcd)Mal (abcd) −→ Asp (abcd)

7

Detailed MFA results for Fig. 2 of main article

8

m/z

0 1 2 3 0

0.0140.027 0.94 0.95 0.96

Pyr

174

0 1 2 0

0.0110.023 0.96 0.97 0.98

Ala

232

0 1 2 3 0

0.0110.021 0.96 0.97 0.98

Ala

260

0 1 2 3 4 0

0.051 0.1 0.79 0.83 0.86

Su

c289

0 1 2 3 4 0

0.0340.068 0.83 0.85 0.88

Mal

419

0 1 2 0

0.0410.082 0.89 0.92 0.94

Asp

302

0 1 2 3 0

0.030.060.87 0.90.92

Asp

390

0 1 2 3 4 0

0.0250.051 0.87 0.89 0.9

Asp

418

0 1 2 3 4 0 0.1 0.20.670.73 0.8

Glu

330

0 1 2 3 4 5 0 0.1 0.20.660.730.79

Glu

432

0 1 2 3 4 5 0

0.170.33 0.60.720.83

Gln

431

0 1 2 3 4 5 6 0

0.0240.048 0.81 0.83 0.85

Cit

459

0 1 2 3 4 5 0

0.150.310.630.730.84

Gln

431.

x

Simulated Measured

Figure S-2: Measured and optimally fitted mass isotopomer distributions from 9h untreatedcells. Data shown are corrected for natural isotope abundance. SSR = 34.0 < 95.0 =χ2

inv,0.95(70).

9

m/z

0 1 2 0

0.0098

0.02

0.97

0.98

0.99A

la23

2

0 1 2 3 0

0.0088

0.018

0.96

0.97

0.98

Ala

260

0 1 2 3 4 0

0.063

0.13

0.72

0.76

0.81

Su

c289

0 1 2 3 4 0

0.046

0.093

0.77

0.8

0.83

Mal

419

0 1 2 0

0.045

0.089

0.88

0.91

0.94A

sp30

2

0 1 2 3 0

0.031

0.062

0.86

0.88

0.91

Asp

390

0 1 2 3 4 0

0.029

0.058

0.84

0.87

0.89

Asp

418

0 1 2 3 4 0

0.12

0.24

0.36

0.49

0.61

Glu

330

0 1 2 3 4 5 0

0.032

0.064

0.28

0.47

0.66G

lu43

2

0 1 2 3 4 5 0

0.16

0.33

0.61

0.72

0.83

Gln

431

0123456 0

0.025

0.05

0.84

0.85

0.87

Cit

459

0 1 2 3 4 5 0

0.16

0.32

0.62

0.72

Gln

431.

x

Simulated Measured

Figure S-3: Measured and optimally fitted mass isotopomer distributions from 9h palmitate-treated cells. Data shown are corrected for natural isotope abundance. SSR = 55.9 < 90.3 =χ2

inv,0.95(66).

10

m/z

0 1 2 0

0.01

0.02

0.98

0.99

1A

la23

2

0 1 2 3 0

0.0087

0.017

0.97

0.98

0.99

Ala

260

0 1 2 3 4 0

0.069

0.14

0.73

0.78

0.83

Su

c289

0 1 2 3 4 0

0.053

0.11

0.75

0.79

0.82

Mal

419

0 1 2 0

0.047

0.095

0.87

0.9

0.94A

sp30

2

0 1 2 3 0

0.037

0.073

0.85

0.88

0.9

Asp

390

0 1 2 3 4 0

0.033

0.065

0.84

0.86

0.89

Asp

418

0 1 2 3 4 0

0.045

0.089

0.29

0.47

0.66

Glu

330

0 1 2 3 4 5 0

0.037

0.074

0.29

0.47

0.65

Glu

432

0 1 2 3 4 5 0

0.15

0.29

0.64

0.74

0.84

Gln

431

0123456 0

0.028

0.057

0.78

0.8

0.82

Cit

459

0 1 2 3 4 5 0

0.14

0.28

0.66

0.75

0.85

Gln

431.

x

Simulated Measured

Figure S-4: Measured and optimally fitted mass isotopomer distributions from 9hpalmitate/oleate-treated cells. Data shown are corrected for natural isotope abundance.SSR = 60.7 < 89.2 = χ2

inv,0.95(65).

11

Table S-4: Best-fit parameter estimates and 95% confidence intervals for 9h untreated cells.Net flux values are relative to a pyruvate kinase (T3P → Pyr) flux of 100 (actual PK fluxwas 0.40 µmol/h/106 cells). Exchange fluxes are scaled according to the transformation

v[0,100]exch = 100 × vexch/(vexch + vref), where vref is the PK flux. Dilution fluxes represent

percentage of the total sampled pool.

Parameter Value IntervalNet fluxesGlc → G6P 61.7 [44.2 , 85.8]G6P → Sink 11.7 [ 0.0 , 37.3]G6P → T3P + T3P 50.0 [35.6 , 59.5]T3P → Pyr.c 100.0 [71.1 , 118.9]Pyr.c → Lac 44.5 [36.2 , 52.7]Pyr.m + CO2 → Mal 55.4 [41.9 , 86.1]Pyr.c → AcCoA + CO2 65.8 [38.2 , 84.4]Mal → Pyr.m + CO2 65.6 [50.4 , 97.7]Pyr.m → Pyr.c 10.2 [ 8.2 , 12.3]AcCoA + Mal → Cit.m 65.8 [38.2 , 84.4]Cit.m → Akg + CO2 44.7 [34.6 , 55.9]Akg → Suc.m + CO2 54.9 [42.9 , 67.9]Suc.m → Fum 54.9 [42.9 , 67.9]Fum → Mal 54.9 [42.9 , 67.9]Gln.x → Gln 10.2 [ 8.2 , 12.3]Gln.p → Gln -0.0 [ 0.0 , 0.0]Gln → Glu 10.2 [ 8.2 , 12.3]Glu → Akg 10.2 [ 8.2 , 12.3]Cit.m → Fatty acids + Mal 21.1 [ 0.0 , 35.2]

Exchange fluxesPyr.c ↔ Lac 87.9 [70.7 , 100.0]Pyr.m ↔ Pyr.c 57.6 [35.8 , 72.2]Cit.m ↔ Akg + CO2 20.1 [10.0 , 27.4]Suc.m ↔ Fum 100.0 [ 0.0 , 100.0]Fum ↔ Mal 6.1 [ 0.0 , 100.0]Gln.p ↔ Gln 0.3 [ 0.1 , 0.5]Glu ↔ Akg 51.9 [44.3 , 63.7]

Dilution/mixing fluxesCit.d → Cit 5.3 [ 0.0 , 14.6]Cit.m → Cit 94.7 [85.4 , 100.0]Suc.d → Suc 23.6 [12.1 , 32.0]Suc.m → Suc 76.4 [68.0 , 87.9]Pyr.c → Pyr 9.0 [ 0.0 , 49.0]Pyr.m → Pyr 91.0 [51.0 , 100.0]Pyr.m → Ala 60.8 [43.4 , 79.5]Ala.d → Ala 39.2 [20.5 , 56.6]Mal → Asp 76.4 [71.4 , 81.8]Asp.d → Asp 23.6 [18.2 , 28.6]

12

Table S-5: Best-fit parameter estimates and 95% confidence intervals for 9h palmitate-treatedcells. Net flux values are relative to a pyruvate kinase (T3P → Pyr) flux of 100 (actual PKflux was 0.22 µmol/h/106 cells). Exchange fluxes are scaled according to the transformation

v[0,100]exch = 100 × vexch/(vexch + vref), where vref is the PK flux. Dilution fluxes represent

percentage of the total sampled pool.

Parameter Value IntervalNet fluxesGlc → G6P 56.9 [ 42.1 , 101.3]G6P → Sink 6.9 [ 0.0 , 52.3]G6P → T3P + T3P 50.0 [ 41.7 , 65.8]T3P → Pyr.c 100.0 [ 83.4 , 131.7]Pyr.c → Lac 5.3 [ 0.0 , 11.3]Pyr.m + CO2 → Mal 139.8 [112.8 , 168.5]Pyr.c → AcCoA + CO2 127.2 [108.1 , 160.3]Mal → Pyr.m + CO2 172.3 [142.6 , 204.1]Pyr.m → Pyr.c 32.6 [ 28.3 , 36.9]AcCoA + Mal → Cit.m 127.2 [108.1 , 160.3]Cit.m → Akg + CO2 127.2 [108.1 , 147.7]Akg → Suc.m + CO2 159.8 [136.8 , 183.9]Suc.m → Fum 159.8 [136.8 , 183.9]Fum → Mal 159.8 [136.8 , 183.9]Gln.x → Gln 32.6 [ 28.3 , 36.9]Gln.p → Gln -0.0 [ 0.0 , 0.0]Gln → Glu 32.6 [ 28.3 , 36.9]Glu → Akg 32.6 [ 28.3 , 36.9]Cit.m → Fatty acids + Mal 0.0 [ 0.0 , 31.6]

Exchange fluxesPyr.c ↔ Lac 70.0 [ 65.6 , 74.1]Pyr.m ↔ Pyr.c 100.0 [ 90.4 , 100.0]Cit.m ↔ Akg + CO2 29.4 [ 18.6 , 41.6]Suc.m ↔ Fum 15.2 [ 0.0 , 100.0]Fum ↔ Mal 87.2 [ 0.0 , 100.0]Gln.p ↔ Gln 0.3 [ 0.0 , 0.9]Glu ↔ Akg 43.9 [ 40.1 , 47.5]

Dilution/mixing fluxesCit.d → Cit 46.0 [ 40.0 , 52.0]Cit.m → Cit 54.0 [ 48.0 , 60.0]Suc.d → Suc 24.7 [ 10.0 , 30.3]Suc.m → Suc 75.3 [ 69.7 , 90.0]Pyr.c → Pyr 41.0 [ 0.0 , 100.0]Pyr.m → Pyr 59.0 [ 0.0 , 100.0]Pyr.m → Ala 40.0 [ 31.4 , 50.8]Ala.d → Ala 60.0 [ 49.2 , 68.6]Mal → Asp 62.2 [ 58.9 , 65.6]Asp.d → Asp 37.8 [ 34.4 , 41.1]

13

Table S-6: Best-fit parameter estimates and 95% confidence intervals for 9h palmitate/oleate-treated cells. Net flux values are relative to a pyruvate kinase (T3P → Pyr) flux of 100(actual PK flux was 0.29 µmol/h/106 cells). Exchange fluxes are scaled according to the

transformation v[0,100]exch = 100× vexch/(vexch + vref), where vref is the PK flux. Dilution fluxes

represent percentage of the total sampled pool.

Parameter Value IntervalNet fluxesGlc → G6P 68.1 [46.9 , 91.7]G6P → Sink 18.1 [ 0.0 , 42.8]G6P → T3P + T3P 50.0 [42.8 , 59.4]T3P → Pyr.c 100.0 [85.5 , 118.8]Pyr.c → Lac 58.8 [45.9 , 71.6]Pyr.m + CO2 → Mal 74.9 [56.3 , 88.2]Pyr.c → AcCoA + CO2 59.8 [51.7 , 76.0]Mal → Pyr.m + CO2 93.5 [74.1 , 108.3]Pyr.m → Pyr.c 18.6 [16.4 , 20.7]AcCoA + Mal → Cit.m 59.8 [51.7 , 76.0]Cit.m → Akg + CO2 59.8 [51.5 , 68.6]Akg → Suc.m + CO2 78.4 [68.3 , 89.0]Suc.m → Fum 78.4 [68.3 , 89.0]Fum → Mal 78.4 [68.3 , 89.0]Gln.x → Gln 18.6 [16.4 , 20.7]Gln.p → Gln -0.0 [ 0.0 , 0.0]Gln → Glu 18.6 [16.4 , 20.7]Glu → Akg 18.6 [16.4 , 20.7]Cit.m → Fatty acids + Mal 0.0 [ 0.0 , 17.2]

Exchange fluxesPyr.c ↔ Lac 68.5 [56.5 , 80.6]Pyr.m ↔ Pyr.c 73.4 [59.4 , 100.0]Cit.m ↔ Akg + CO2 27.9 [21.2 , 36.8]Suc.m ↔ Fum 0.0 [ 0.0 , 26.7]Fum ↔ Mal 98.1 [ 0.0 , 100.0]Gln.p ↔ Gln 0.2 [ 0.0 , 0.6]Glu ↔ Akg 35.6 [32.2 , 38.9]

Dilution/mixing fluxesCit.d → Cit 31.0 [25.1 , 36.8]Cit.m → Cit 69.0 [63.2 , 74.9]Suc.d → Suc 35.1 [25.9 , 38.5]Suc.m → Suc 64.9 [61.5 , 74.1]Pyr.c → Pyr 81.1 [ 0.0 , 100.0]Pyr.m → Pyr 18.9 [ 0.0 , 100.0]Pyr.m → Ala 23.8 [16.7 , 50.4]Ala.d → Ala 76.2 [49.6 , 83.3]Mal → Asp 61.2 [58.1 , 64.4]Asp.d → Asp 38.8 [35.6 , 41.9]

14

Detailed MFA results for Fig. 5 of main article

15

m/z

0 1 2 0

0.0110.023 0.96 0.97 0.98

Ala

232

0 1 2 3 0

0.0110.021 0.96 0.97 0.98

Ala

260

0 1 2 3 4 0

0.087 0.17 0.61 0.67 0.72

Su

c289

0 1 2 3 4 0

0.059 0.12 0.68 0.73 0.77

Mal

419

0 1 2 0

0.088 0.18 0.75 0.81 0.87

Asp

302

0 1 2 3 0

0.066 0.13 0.71 0.76 0.8

Asp

390

0 1 2 3 4 0

0.057 0.11 0.7 0.74 0.78

Asp

418

0 1 2 3 4 0 0.10.210.560.63 0.7

Glu

330

0 1 2 3 4 5 0 0.1 0.20.550.620.69

Glu

432

0 1 2 3 4 5 6 0

0.055 0.11 0.64 0.68 0.72

Cit

459

Simulated Measured

Figure S-5: Measured and optimally fitted mass isotopomer distributions from 9h palmitate-treated cells. Data shown are corrected for natural isotope abundance. SSR = 36.0 < 72.6 =χ2

inv,0.95(51).

16

m/z

0 1 2 0

0.0130.025 0.96 0.97 0.98

Ala

232

0 1 2 3 0

0.0097 0.019 0.96 0.97 0.98

Ala

260

0 1 2 3 4 0

0.087 0.17 0.59 0.65 0.71

Su

c289

0 1 2 3 4 0

0.059 0.12 0.67 0.72 0.76

Mal

419

0 1 2 0

0.086 0.17 0.75 0.81 0.87

Asp

302

0 1 2 3 0

0.064 0.13 0.7 0.75 0.79

Asp

390

0 1 2 3 4 0

0.057 0.11 0.69 0.73 0.77

Asp

418

0 1 2 3 4 0

0.098 0.2 0.56 0.62 0.69

Glu

330

0 1 2 3 4 5 0

0.097 0.19 0.54 0.61 0.67

Glu

432

0 1 2 3 4 5 6 0

0.065 0.13 0.62 0.67 0.71

Cit

459

Simulated Measured

Figure S-6: Measured and optimally fitted mass isotopomer distributions from 9h palmitate-treated, EAA-supplemented cells. Data shown are corrected for natural isotope abundance.SSR = 52.9 < 72.6 = χ2

inv,0.95(51).

17

m/z

0 1 2 0

0.0110.023 0.96 0.97 0.98

Ala

232

0 1 2 3 0

0.0110.022 0.96 0.97 0.98

Ala

260

0 1 2 3 4 0

0.089 0.18 0.61 0.67 0.73

Su

c289

0 1 2 3 4 0

0.058 0.12 0.69 0.74 0.78

Mal

419

0 1 2 0

0.088 0.18 0.75 0.81 0.87

Asp

302

0 1 2 3 0

0.067 0.13 0.71 0.76 0.8

Asp

390

0 1 2 3 4 0

0.058 0.12 0.7 0.74 0.78

Asp

418

0 1 2 3 4 0

0.110.210.570.640.71

Glu

330

0 1 2 3 4 5 0 0.10.210.560.63 0.7

Glu

432

0 1 2 3 4 5 6 0

0.055 0.11 0.64 0.69 0.73

Cit

459

Simulated Measured

Figure S-7: Measured and optimally fitted mass isotopomer distributions from 9h palmitate-treated, NEAA-supplemented cells. Data shown are corrected for natural isotope abundance.SSR = 38.0 < 72.6 = χ2

inv,0.95(51).

18

Table S-7: Best-fit parameter estimates and 95% confidence intervals for 9h palmitate-treatedcells. Net flux values are relative to a pyruvate kinase (T3P → Pyr) flux of 100 (actual PKflux was 0.28 µmol/h/106 cells). Exchange fluxes are scaled according to the transformation

v[0,100]exch = 100 × vexch/(vexch + vref), where vref is the PK flux. Dilution fluxes represent

percentage of the total sampled pool.

Parameter Value IntervalGlc → G6P 50.0 [ 47.2 , 52.7]G6P → Sink 0.0 [ 0.0 , 8.3]G6P → T3P + T3P 50.0 [ 41.7 , 52.8]T3P → Pyr 100.0 [ 83.4 , 105.6]Pyr → Lac 7.4 [ 6.9 , 7.8]Pyr + CO2 → Mal 42.4 [ 34.0 , 52.9]Pyr → AcCoA + CO2 117.5 [100.8 , 123.2]Mal → Pyr + CO2 67.3 [ 58.6 , 77.9]Fatty acids + Mal → Cit 117.5 [100.8 , 123.2]Cit → Akg + CO2 61.3 [ 53.6 , 69.9]Akg → Suc + CO2 86.2 [ 78.3 , 94.9]Suc → Fum 86.2 [ 78.3 , 94.9]Fum → Mal 86.2 [ 78.3 , 94.9]Gln.x → Gln 24.9 [ 23.9 , 25.8]Gln → Glu 24.9 [ 23.9 , 25.8]Glu → Akg 24.9 [ 23.9 , 25.8]Cit → Fatty acids + Mal 56.2 [ 41.9 , 65.0]

Exchange fluxesPyr ↔ Lac 55.3 [ 43.3 , 65.5]Cit ↔ Akg + CO2 21.0 [ 16.4 , 25.7]Suc ↔ Fum 27.2 [ 17.5 , 100.0]Fum ↔ Mal 92.2 [ 20.4 , 100.0]Glu ↔ Akg 89.1 [ 83.8 , 94.8]

Dilution/mixing fluxesPyr → Ala 61.2 [ 46.8 , 81.0]Ala.d → Ala 38.8 [ 19.0 , 53.2]Mal → Asp 96.9 [ 93.6 , 100.0]Asp.d → Asp 3.1 [ 0.0 , 6.4]

19

Table S-8: Best-fit parameter estimates and 95% confidence intervals for 9h palmitate-treated, EAA-supplemented cells. Net flux values are relative to a pyruvate kinase (T3P→ Pyr) flux of 100 (actual PK flux was 0.33 µmol/h/106 cells). Exchange fluxes are scaled

according to the transformation v[0,100]exch = 100× vexch/(vexch + vref), where vref is the PK flux.

Dilution fluxes represent percentage of the total sampled pool.

Parameter Value IntervalGlc → G6P 50.0 [46.1 , 53.9]G6P → Sink 0.0 [ 0.0 , 10.1]G6P → T3P + T3P 50.0 [40.7 , 53.9]T3P → Pyr 100.0 [81.5 , 107.7]Pyr → Lac 24.7 [22.1 , 27.3]Pyr + CO2 → Mal 33.7 [26.5 , 42.5]Pyr → AcCoA + CO2 93.4 [75.0 , 101.6]Mal → Pyr + CO2 51.8 [44.5 , 60.7]Fatty acids + Mal → Cit 93.4 [75.0 , 101.6]Cit → Akg + CO2 59.4 [49.9 , 68.3]Akg → Suc + CO2 77.5 [67.9 , 86.5]Suc → Fum 77.5 [67.9 , 86.5]Fum → Mal 77.5 [67.9 , 86.5]Gln.x → Gln 18.1 [17.6 , 18.6]Gln → Glu 18.1 [17.6 , 18.6]Glu → Akg 18.1 [17.6 , 18.6]Cit → Fatty acids + Mal 34.0 [20.5 , 43.5]

Exchange fluxesPyr ↔ Lac 42.3 [26.8 , 54.6]Cit ↔ Akg + CO2 18.0 [13.3 , 22.6]Suc ↔ Fum 24.0 [ 8.5 , 100.0]Fum ↔ Mal 62.8 [ 8.7 , 100.0]Glu ↔ Akg 87.1 [81.5 , 93.4]

Dilution/mixing fluxesPyr → Ala 56.6 [42.3 , 74.9]Ala.d → Ala 43.4 [25.1 , 57.7]Mal → Asp 96.0 [92.7 , 99.3]Asp.d → Asp 4.0 [ 0.7 , 7.3]

20

Table S-9: Best-fit parameter estimates and 95% confidence intervals for 9h palmitate-treated, NEAA-supplemented cells. Net flux values are relative to a pyruvate kinase (T3P→ Pyr) flux of 100 (actual PK flux was 0.24 µmol/h/106 cells). Exchange fluxes are scaled

according to the transformation v[0,100]exch = 100× vexch/(vexch + vref), where vref is the PK flux.

Dilution fluxes represent percentage of the total sampled pool.

Parameter Value IntervalNet fluxesGlc → G6P 50.0 [ 47.9 , 52.1]G6P → Sink -0.0 [ 0.0 , 3.9]G6P → T3P + T3P 50.0 [ 46.0 , 52.1]T3P → Pyr 100.0 [ 91.9 , 104.2]Pyr → Lac 5.4 [ 4.8 , 6.0]Pyr + CO2 → Mal 61.9 [ 51.3 , 71.7]Pyr → AcCoA + CO2 128.4 [120.2 , 132.7]Mal → Pyr + CO2 95.7 [ 84.8 , 105.6]Fatty acids + Mal → Cit 128.4 [120.2 , 132.7]Cit → Akg + CO2 71.5 [ 63.3 , 80.5]Akg → Suc + CO2 105.3 [ 96.9 , 114.4]Suc → Fum 105.3 [ 96.9 , 114.4]Fum → Mal 105.3 [ 96.9 , 114.4]Gln.x → Gln 33.8 [ 33.0 , 34.5]Gln → Glu 33.8 [ 33.0 , 34.5]Glu → Akg 33.8 [ 33.0 , 34.5]Cit → Fatty acids + Mal 56.8 [ 47.1 , 65.7]

Exchange fluxesPyr ↔ Lac 55.3 [ 47.0 , 62.2]Cit ↔ Akg + CO2 21.6 [ 17.6 , 25.8]Suc ↔ Fum 100.0 [ 19.1 , 100.0]Fum ↔ Mal 26.2 [ 18.1 , 100.0]Glu ↔ Akg 91.6 [ 86.9 , 96.6]

Dilution/mixing fluxesPyr → Ala 47.1 [ 38.8 , 57.2]Ala.d → Ala 52.9 [ 42.8 , 61.2]Mal → Asp 99.5 [ 96.4 , 100.0]Asp.d → Asp 0.5 [ 0.0 , 3.6]

21