EMBO Molecular Medicine

cross-journal focus

Frontiers in Metabolism

cross-journal focus

emboj.embopress.org | embor.embopress.org | embomolmed.embopress.org | msb.embopress.org

Maria PolychronidouEditormaria.polychronidou@embo.org | T +49 6221 8891 410

Maria received her PhD from the University of Heidelberg, where she studied the role of nuclear membrane proteins in development and aging. During her post-doctoral work, she focused on the analysis of tissue-specific regulatory functions of Hox transcription factors using a combination of computational and genome-wide methods.

Thomas LembergerChief Editorthomas.lemberger@embo.org | T +49 6221 8891 413

Thomas completed his PhD at the University of Lausanne, where he studied hormonal regulation of gene expression by nuclear receptors. He moved then to Heidelberg where his research focused on the regulation of transcription in the brain.

Roberto BuccioneEditorr.buccione@embomolmed.org | T +49 6221 8891 310

Roberto Buccione completed his PhD at the University of l’Aquila, Italy studying the process of oogenesis in mammals. After continuing these studies as a post-doctoral researcher at the Jackson Laboratory, Bar Harbor ME, USA, he joined the Mario Negri Sud research institute in S. Maria Imbaro, Italy, where he lead a research group focused on the cell biology of tumour cell invasion. He joined EMBO Molecular Medicine as a Scientific Editor in October 2012.

EMBO Molecular Medicine

Céline CarretEditorc.carret@embomolmed.org | T +49 6221 8891 310

Céline Carret completed her PhD at the University of Montpellier, France, characterising host immunodominant antigens to fight babesiosis, a parasitic disease caused by a unicellular Apicomplexan parasite closely related to the malaria agent Plasmodium. She further developed her post-doctoral career on malaria working at the Wellcome Trust Sanger Institute in Cambridge, UK and Instituto de Medicina Molecular in Lisbon, Portugal. Céline joined EMBO Molecular Medicine as a Scientific Editor in March 2011.

EMBO Molecular Medicine

Stefanie DimmelerChief Editoreditor@embomolmed.org | T +49 6221 8891 310Stefanie Dimmeler is Professor of Experimental Medicine and Director of the Institute of Cardiovascular Regeneration, Center for Molecular Medicine at the University of Frankfurt, Germany. Her group elucidates the basic mechanisms underlying cardiovascular disease and vessel growth with the aim to develop new cellular and pharmacological therapies for improving the treatment of cardiovascular disease. Her current ongoing research focuses on epigenetic mechanisms that control cardiovascular repair, specifically the function of histone modifying enzymes and microRNAs. She received several international prizes including the Leibniz Award 2005, the Award of the Jung Foundation 2007 and the FEBS award in 2006.

EMBO Molecular Medicine

EDITORS

Frontiers in Metabolism

Molecular Systems Biology

A computational study of the Warburg effect identifies metabolic targets inhibiting cancer migrationMolecular Systems Biology (2014) 10:744Yizhak K, Le Dévédec SE, Rogkoti VM, Baenke F, de Boer VC, Frezza C, Schulze A, van de Water B, Ruppin E.DOI: 10.15252/msb.20134993

EMBO Molecular Medicine

Principles of targeting endothelial cell metabolism to treat angiogenesis and endothelial cell dysfunction in diseaseEMBO Molecular Medicine (2014) 6 (9):1105-20Goveia J, Stapor P, Carmeliet P.DOI: 10.15252/emmm.201404156

EMBO Molecular Medicine

The heart-liver metabolic axis: defective communication exacerbates diseaseEMBO Molecular Medicine (2014) 6 (4):436-8Baskin KK, Bookout AL, Olson EN.DOI: 10.1002/emmm.201303800

EMBO Molecular Medicine

Salvaging hope: Is increasing NAD+ a key to treating mitochondrial myopathy?EMBO Molecular Medicine (2014) 6 (6):705-7Lightowlers RN, Chrzanowska-Lightowlers ZMA.DOI: 10.15252/emmm.201404179

EMBO Molecular Medicine

Long-term therapeutic silencing of miR-33 increases circulating triglyceride levels and hepatic lipid accumulation in miceEMBO Molecular Medicine (2014) 6 (9):1133-41Goedeke L, Salerno A, Ramírez CM, Guo L, Allen RM, Yin X, Langley SR, Esau C, Wanschel A, Fisher EA, Suárez Y, Baldán A, Mayr M, Fernández-Hernando C.DOI: 10.15252/emmm.201404046

Molecular Systems Biology

Phenotypic bistability in Escherichia coli’s central carbon metabolismMolecular Systems Biology (2014) 10:736Kotte O, Volkmer B, Radzikowski JL, Heinemann M.DOI: 10.15252/msb.20135022

CONTENTS

continued overleaf

EMBO Molecular Medicine

Macrophages and β-cells are responsible for CXCR2-mediated neutrophil infiltration of the pancreas during autoimmune diabetesEMBO Molecular Medicine (2014) 6 (8):1090-104Diana J, Lehuen A.DOI: 10.15252/emmm.201404144

EMBO Molecular Medicine

Deoxypyrimidine monophosphate bypass therapy for thymidine kinase 2 deficiencyEMBO Molecular Medicine (2014) 6 (8):1016-27Garone C, Garcia-Diaz B, Emmanuele V, Lopez LC, Tadesse S, Akman HO, Tanji K, Quinzii CM, Hirano M.DOI: 10.15252/emmm.201404092

Molecular Systems Biology

Screen for mitochondrial DNA copy number maintenance genes reveals essential role for ATP synthaseMolecular Systems Biology (2014) 10:734Fukuoh A, Cannino G, Gerards M, Buckley S, Kazancioglu S, Scialo F, Lihavainen E, Ribeiro A, Dufour E, Jacobs HT.DOI: 10.15252/msb.20145117

EMBO Molecular Medicine

Constitutive hippocampal cholesterol loss underlies poor cognition in old rodentsEMBO Molecular Medicine (2014) 6 (7):902-17Martin MG, Ahmed T, Korovaichuk A, Venero C, Menchón SA, Salas I, Munck S, Herreras O, Balschun D, Dotti CG.DOI: 10.15252/emmm.201303711

Molecular Systems Biology

Genetic regulation of mouse liver metabolite levelsMolecular Systems Biology (2014) 10:730Ghazalpour A, Bennett BJ, Shih D, Che N, Orozco L, Pan C, Hagopian R, He A, Kayne P, Yang WP, Kirchgessner T, Lusis AJ.DOI: 10.15252/msb.20135004

CONTENTS

continued overleaf

EMBO Molecular Medicine

Thioredoxin-interacting protein regulates protein disulfide isomerases and endoplasmic reticulum stressEMBO Molecular Medicine (2014) 6 (6):732-43Lee S, Min Kim S, Dotimas J, Li L, Feener EP, Baldus S, Myers RB, Chutkow WA, Patwari P, Yoshioka J, Lee RT. DOI: 10.15252/emmm.201302561

Molecular Systems Biology

Metabolic shifts toward glutamine regulate tumor growth, invasion and bioenergetics in ovarian cancerMolecular Systems Biology (2014) 10:728Yang L, Moss T, Mangala LS, Marini J, Zhao H, Wahlig S, Armaiz-Pena G, Jiang D, Achreja A, Win J, Roopaimoole R, Rodriguez-Aguayo C, Mercado-Uribe I, Lopez-Berestein G, Liu J, Tsukamoto T, Sood AK, Ram PT, Nagrath D.DOI: 10.1002/msb.20134892

EMBO Molecular Medicine

Embelin inhibits endothelial mitochondrial respiration and impairs neoangiogenesis during tumor growth and wound healingEMBO Molecular Medicine (2014) 6 (5):624-39Coutelle O, Hornig-Do HT, Witt A, Andree M, Schiffmann LM, Piekarek M, Brinkmann K, Seeger JM, Liwschitz M, Miwa S, Hallek M, Krönke M, Trifunovic A, Eming SA, Wiesner RJ, Hacker UT, Kashkar H.DOI: 10.1002/emmm.201303016

EMBO Molecular Medicine

Mitochondrial response to nutrient availability and its role in metabolic diseaseEMBO Molecular Medicine (2014) 6 (5):580-9Gao AW, Cantó C, Houtkooper RH.DOI: 10.1002/emmm.201303782

Molecular Systems Biology

Non-enzymatic glycolysis and pentose phosphate pathway-like reactions in a plausible Archean oceanMolecular Systems Biology (2014) 10:725Keller MA, Turchyn AV, Ralser M.DOI: 10.1002/msb.20145228

CONTENTS

continued overleaf

EMBO Molecular Medicine

Metabolic crosstalk between the heart and liver impacts familial hypertrophic cardiomyopathyEMBO Molecular Medicine (2014) 6 (4):482-95Magida JA, Leinwand LA.DOI: 10.1002/emmm.201302852

EMBO Molecular Medicine

Effective treatment of mitochondrial myopathy by nicotinamide riboside, a vitamin B3EMBO Molecular Medicine (2014) 6 (6):721-31Khan NA, Auranen M, Paetau I, Pirinen E, Euro L, Forsström S, Pasila L, Velagapudi V, Carroll CJ, Auwerx J, Suomalainen A. DOI: 10.1002/emmm.201403943

Molecular Systems Biology

Identification of anticancer drugs for hepatocellular carcinoma through personalized genome-scale metabolic modelingMolecular Systems Biology (2014) 10:721Agren R, Mardinoglu A, Asplund A, Kampf C, Uhlen M, Nielsen J.DOI: 10.1002/msb.145122

Molecular Systems Biology

Glutamine-driven oxidative phosphorylation is a major ATP source in transformed mammalian cells in both normoxia and hypoxiaMolecular Systems Biology (2013) 9:712Fan J, Kamphorst JJ, Mathew R, Chung MK, White E, Shlomi T, Rabinowitz JD.DOI: 10.1038/msb.2013.65

CONTENTS

Article

A computational study of the Warburg effectidentifies metabolic targets inhibitingcancer migrationKeren Yizhak1,*,†, Sylvia E Le Dévédec2,†, Vasiliki Maria Rogkoti2, Franziska Baenke3, Vincent C de Boer4,

Christian Frezza5, Almut Schulze3, Bob van de Water2,‡ & Eytan Ruppin1,6,‡,**

Abstract

Over the last decade, the field of cancer metabolism has mainlyfocused on studying the role of tumorigenic metabolic rewiringin supporting cancer proliferation. Here, we perform the firstgenome-scale computational study of the metabolic underpin-nings of cancer migration. We build genome-scale metabolicmodels of the NCI-60 cell lines that capture the Warburg effect(aerobic glycolysis) typically occurring in cancer cells. The extentof the Warburg effect in each of these cell line models is quan-tified by the ratio of glycolytic to oxidative ATP flux (AFR),which is found to be highly positively associated with cancercell migration. We hence predicted that targeting genes thatmitigate the Warburg effect by reducing the AFR may specifi-cally inhibit cancer migration. By testing the anti-migratoryeffects of silencing such 17 top predicted genes in four breastand lung cancer cell lines, we find that up to 13 of these novelpredictions significantly attenuate cell migration either in all orone cell line only, while having almost no effect on cell prolifer-ation. Furthermore, in accordance with the predictions, a signifi-cant reduction is observed in the ratio between experimentallymeasured ECAR and OCR levels following these perturbations.Inhibiting anti-migratory targets is a promising future avenue intreating cancer since it may decrease cytotoxic-related sideeffects that plague current anti-proliferative treatments.Furthermore, it may reduce cytotoxic-related clonal selection ofmore aggressive cancer cells and the likelihood of emergingresistance.

Keywords cancer cell migration; cellular metabolism; genome-scale

metabolic modeling

Subject Categories Genome-Scale & Integrative Biology; Metabolism;

Computational Biology

DOI 10.15252/msb.20134993 | Received 18 November 2013 | Revised 6 July

2014 | Accepted 7 July 2014

Mol Syst Biol. (2014) 10: 744

Introduction

Altered tumor metabolism has become a generally regarded hall-

mark of cancer (Hanahan & Weinberg, 2011). The initial recognition

that metabolism is altered in cancer can be traced back to Otto

Warburg’s early studies, showing that transformed cells consume

glucose at an abnormally high rate and largely reduce it to lactate,

even in the presence of oxygen (Warburg, 1956). Over the last

decade, much of the field of cancer metabolism has focused on the

role of the Warburg effect in supporting cancer proliferation (Vander

Heiden et al, 2009). However, the role of this process in supporting

other fundamental cancer phenotypes such as cellular migration has

received far less attention.

Contemporary cytotoxic cancer treatment has been mainly

based on drugs that kill proliferating cells generally unselectively

and are therefore accompanied by many undesirable side effects.

Drug targets that can inhibit migration but leave cellular prolifer-

ation relatively spared may be able to avoid such side effects.

Such targets may have the additional benefit of reducing the

selection for more resistant clones that occurs due to the elimi-

nation of treatment-sensitive cells. The growing availability of

high-throughput measurements for a range of cancer cells

presents an opportunity to study a wider scope of dysregulated

metabolism across many different cancers. Here, we aim to

1 The Blavatnik School of Computer Science, Tel-Aviv University, Tel-Aviv, Israel2 Division of Toxicology, Leiden Academic Centre for Drug Research, Leiden University, Leiden, The Netherlands3 Gene Expression Analysis Laboratory, Cancer Research UK, London Research Institute, London, UK4 Laboratory Genetic Metabolic Diseases, Academic Medical Center, Amsterdam, The Netherlands5 MRC Cancer Unit, Hutchison/MRC Research Centre, University of Cambridge, Cambridge, UK6 The Sackler School of Medicine, Tel-Aviv University, Tel-Aviv, Israel

*Corresponding author. Tel: +972 3 6405378; E-mail: kerenyiz@post.tau.ac.il**Corresponding author. Tel: +972 3 6406528; E-mail: ruppin@post.tau.ac.il†These authors contributed equally to this study‡These authors contributed equally to this study

ª 2014 The Authors. Published under the terms of the CC BY 4.0 license Molecular Systems Biology 10: 744 | 2014 1

observed after all glycolysis inhibitors was lower than the corre-

sponding increase in A549 cells (Fig 1C).

Quantifying the Warburg effect and its relation to proliferationand migration across the NCI-60 cell lines

While ECAR and OCR are the commonly used measures for experi-

mentally quantifying the bioenergetic capacity of the cell and thus

the Warburg effect, the genome-wide scope of GSMMs enables us to

examine other putative measures as well. One promising such

measure we examined is the ratio between the ATP flux rate in the

glycolysis versus its flux rate in OXPHOS (AFR). Clearly, higher AFR

values denote more ‘Warburgian’ cell lines and vice versa. A

comparison of our new AFR metric versus the aforementioned

state-of-the-art ECAR/OCR ratio (EOR) (Materials and Methods and

Supplementary Dataset S2) showed a significant correlation across

the NCI-60 models (Spearman correlation R = 0.66, P-value = 2e�8).

Testing both measures using a genome-wide NCI-60 drug response

dataset (Scherf et al, 2000), we find that the model-predicted wild-

type AFR levels across all cell line models are significantly corre-

lated (Spearman P-value < 0.05; FDR corrected with a = 0.05) with

Gi50 values of 30% of the compounds across these cell lines

(empiric P-value < 9.9e�4), whereas the model-predicted EOR

measure accomplish this task for only 19% of the compounds

(Materials and Methods). Interestingly, we find that out of the 30%

AFR-Gi50-correlated compounds, 97% are positively correlated,

suggesting that the more ‘Warburgian’ cell lines are less responsive

and therefore require higher dosage of compound to suppress their

0 0.5 1

0 0.5

0.5

0.5

1

1

1

0

0

C

3BrPA(HK2)

Iodoacetate(G3PDH)

Fluoride (Enolase)

Glucose

G6P

G3P

2PG

PEP

Pyruvate

Lactate

Oxamate(LDH)

1,3BPG

A

B

Figure 1. A comparison between experimental and predicted in silicomeasurements of lactate secretion (or ECAR) and OCR across different cancer cell lines.

A Measured versus predicted lactate secretion rates across the 59 cell lines available at Jain et al (2012).B Measured versus predicted lactate secretion rates in hypoxic (red) and normoxic (blue) conditions for four breast cancer cell lines: T47D, MCF7, BT549, and Hs578T.

Bars represent the measured lactate secretion rates and the line represents the corresponding predicted rates. Error bars represent SD; number of samples forexperimental data (bars) is n = 7; number of samples for predicted data (line) is n = 1000.

C Predicted ECAR and OCR by the A549 and H460 cell line models following inhibitory perturbations in the glycolytic pathway. The models predictions show a decreasein ECAR (red line) and an increase in OCR (blue line). As found experimentally, the predicted OCR increase in H460 cells is lower than that found for A549 cells. Thex-axes represent the level of inhibition imposed, starting from a zero to a maximal inhibition (Materials and Methods). The specific perturbations include 3BpRA thatinhibits the enzyme hexokinase 2; Iodoacetate that inhibits the enzyme glycerol-3-phosphate dehydrogenase; Fluoride that inhibits the enzyme enolase; andOxamate that inhibits the enzyme lactate dehydrogenase.

ª 2014 The Authors Molecular Systems Biology 10: 744 | 2014

Keren Yizhak et al Identifying anti-migratory metabolic drug targets Molecular Systems Biology

3

integrate pertaining data with a genome-scale mechanistic model

of human metabolism to study the role of the Warburg effect in

tumor progression and its potential association with cellular

migration.

Genome-scale metabolic modeling is an increasingly widely

used computational framework for studying metabolism. Given

the genome-scale metabolic model (GSMM) of a species along-

side contextual information such as growth media and ‘omics’

data, one can obtain a fairly accurate prediction of numerous

metabolic phenotypes, including growth rates, nutrient uptake

rates, gene essentiality, and more (Covert et al, 2004). GSMMs

have been used for various applications (Oberhardt et al, 2009;

Chandrasekaran & Price, 2010; Jensen & Papin, 2010; Szappanos

et al, 2011; Wessely et al, 2011; Lerman et al, 2012; Nogales

et al, 2012; Schuetz et al, 2012) including drug discovery

(Trawick & Schilling, 2006; Oberhardt et al, 2013; Yizhak et al,

2013) and metabolic engineering (Burgard et al, 2003; Pharkya

et al, 2004). Over the last few years, GSMMs have been success-

fully used for modeling human metabolism as well (Duarte et al,

2007; Ma et al, 2007; Shlomi et al, 2008; Gille et al, 2010; Lewis

et al, 2010; Mardinoglu et al, 2013). Specifically, GSMM models

of cancer cells have been reconstructed and applied for predict-

ing selective drug targets, as well as for studying the role of

tumor suppressors and oxidative stress (Folger et al, 2011; Frezza

et al, 2011; Agren et al, 2012, 2014; Jerby et al, 2012; Goldstein

et al, 2013; Gatto et al, 2014). In the context of studying the

Warburg effect, the original human metabolic model does not

predict forced lactate secretion under maximal biomass produc-

tion rate, even when oxygen consumption rate equals zero.

This renders it unsuitable for studying the Warburg effect as is,

as already noted by (Shlomi et al, 2011). While the addition of

solvent capacity constraints has been shown to overcome this

hurdle in principle (Shlomi et al, 2011), this addition requires

enzymatic kinetic data which are still largely absent on a

genome-scale.

In this study, we utilize individual genome-scale metabolic

models tailored separately to each of the NCI-60 cancer cell lines

to study the role of the Warburg effect in supporting cancer cellu-

lar migratory capacity. We first test and validate the individual

models against both existing and novel bioenergetic experimental

data. Then, we examine the extent of the Warburg effect occur-

ring in a given cancer cell line, by quantifying the glycolytic to

oxidative ATP flux ratio (AFR). We find that the AFR is highly

positively correlated with cancer cell migration, emphasizing the

role of glycolytic flux in supporting the more aggressive meta-

static stages of tumor development. To determine whether a

causal relation exists between AFR levels and cell migration, we

predict gene silencing that reduce this ratio. These potential

targets are then filtered further to exclude those predicted to

result in cell lethality. Reassuringly, the predicted targets are

found to be significantly more highly expressed in metastatic and

high-grade breast cancer tumors. Experimental investigation of

the top predicted targets via siRNA-mediated knockdown shows

that a significant portion of them truly attenuate cancer cell

migration without inducing a lethal effect. Furthermore, in accor-

dance with the predictions, a significant reduction is observed in

the ratio between ECAR and OCR levels following these genes

silencing perturbations.

Results

Stoichiometric and flux capacity constraints successfully capturethe coupling of high cell proliferation rate to lactate secretionacross individual NCI-60 cancer models

As a starting point for this study, we developed a set of metabolic

models specific for each of the NCI-60 cell lines. We built these

models using a new algorithm we have recently developed termed

PRIME, for building individual models of cells from pertaining

omics data (Yizhak et al, submitted, Supplementary Information

and Supplementary Fig S1). PRIME uses the generic human model

as a scaffold and sets maximal flux capacity constraints over a

subset of its growth-associated reactions according to the expression

levels of their corresponding catalyzing enzymes in each of the

target cell lines.

An important hallmark of cancerous cells is the production of

lactate through the Warburg effect (Warburg, 1956). As a first step

in validating the basic function of our NCI-60 models, we assessed

whether maximizing biomass forces production of lactate, which

would signify proper coupling of biomass production with lactate

output as seen in cancer cells. We found that the models indeed

must secrete lactate under biomass maximization (Supplementary

Information and Supplementary Fig S2). Hence, in contrast to the

original generic model of human metabolism, they enable us to

systematically assess the extent of lactate secretion and study the

Warburg effect across a wide range of cancer cell lines without

needing to add (mostly unknown) solvent capacity constraints, thus

identifying its functional correlates on a genome scale.

Comparing predicted versus experimentally measuredbioenergetics capacity

We compared the predicted lactate secretion rates across all cell

lines to those measured experimentally by Jain et al (Jain et al,

2012), obtaining a moderate but significant correlation (Spearman

correlation R = 0.36, P-value = 5.7e�3, Fig 1A, Materials and Meth-

ods). To further test the models’ performance under different envi-

ronmental conditions, we measured lactate secretion rates in four

breast cancer cell lines, T47D, MCF7, BT549, and Hs578T (Supple-

mentary Dataset S1), under both normoxic and hypoxic conditions

(see Materials and Methods). Utilizing the corresponding cell line

models from the NCI-60 set, we found a high correlation between

measured and predicted lactate secretion levels across both condi-

tions (Spearman correlation R = 0.95, P-value = 1.1e�3, Fig 1B).

The ratio of glycolytic versus oxidative capacity in a cell can be

quantified using its extracellular acidification rate (ECAR, a proxy of

lactate secretion) and its oxygen consumption rate (OCR). To

further examine how well our cell line models capture measured

Warburg-related activity in response to genetic perturbations, we

utilized measured ECAR and OCR levels in response to perturba-

tions in two NCI-60 lung cancer cell lines (A549 and H460), and

compared the results to predictions from our models (Materials and

Methods) (Wu et al, 2007). Qualitatively similar ECAR and OCR

changes are found in response to various enzymatic perturbations

along the glycolytic pathway. Specifically, increased glycolytic inhi-

bition resulted in reduced ECAR and elevated OCR levels in both

cells, while the maximum cellular respiration increase in H460 cells

Molecular Systems Biology 10: 744 | 2014 ª 2014 The Authors

Molecular Systems Biology Identifying anti-migratory metabolic drug targets Keren Yizhak et al

2

observed after all glycolysis inhibitors was lower than the corre-

sponding increase in A549 cells (Fig 1C).

Quantifying the Warburg effect and its relation to proliferationand migration across the NCI-60 cell lines

While ECAR and OCR are the commonly used measures for experi-

mentally quantifying the bioenergetic capacity of the cell and thus

the Warburg effect, the genome-wide scope of GSMMs enables us to

examine other putative measures as well. One promising such

measure we examined is the ratio between the ATP flux rate in the

glycolysis versus its flux rate in OXPHOS (AFR). Clearly, higher AFR

values denote more ‘Warburgian’ cell lines and vice versa. A

comparison of our new AFR metric versus the aforementioned

state-of-the-art ECAR/OCR ratio (EOR) (Materials and Methods and

Supplementary Dataset S2) showed a significant correlation across

the NCI-60 models (Spearman correlation R = 0.66, P-value = 2e�8).

Testing both measures using a genome-wide NCI-60 drug response

dataset (Scherf et al, 2000), we find that the model-predicted wild-

type AFR levels across all cell line models are significantly corre-

lated (Spearman P-value < 0.05; FDR corrected with a = 0.05) with

Gi50 values of 30% of the compounds across these cell lines

(empiric P-value < 9.9e�4), whereas the model-predicted EOR

measure accomplish this task for only 19% of the compounds

(Materials and Methods). Interestingly, we find that out of the 30%

AFR-Gi50-correlated compounds, 97% are positively correlated,

suggesting that the more ‘Warburgian’ cell lines are less responsive

and therefore require higher dosage of compound to suppress their

0 0.5 1

0 0.5

0.5

0.5

1

1

1

0

0

C

3BrPA(HK2)

Iodoacetate(G3PDH)

Fluoride (Enolase)

Glucose

G6P

G3P

2PG

PEP

Pyruvate

Lactate

Oxamate(LDH)

1,3BPG

A

B

Figure 1. A comparison between experimental and predicted in silicomeasurements of lactate secretion (or ECAR) and OCR across different cancer cell lines.

A Measured versus predicted lactate secretion rates across the 59 cell lines available at Jain et al (2012).B Measured versus predicted lactate secretion rates in hypoxic (red) and normoxic (blue) conditions for four breast cancer cell lines: T47D, MCF7, BT549, and Hs578T.

Bars represent the measured lactate secretion rates and the line represents the corresponding predicted rates. Error bars represent SD; number of samples forexperimental data (bars) is n = 7; number of samples for predicted data (line) is n = 1000.

C Predicted ECAR and OCR by the A549 and H460 cell line models following inhibitory perturbations in the glycolytic pathway. The models predictions show a decreasein ECAR (red line) and an increase in OCR (blue line). As found experimentally, the predicted OCR increase in H460 cells is lower than that found for A549 cells. Thex-axes represent the level of inhibition imposed, starting from a zero to a maximal inhibition (Materials and Methods). The specific perturbations include 3BpRA thatinhibits the enzyme hexokinase 2; Iodoacetate that inhibits the enzyme glycerol-3-phosphate dehydrogenase; Fluoride that inhibits the enzyme enolase; andOxamate that inhibits the enzyme lactate dehydrogenase.

ª 2014 The Authors Molecular Systems Biology 10: 744 | 2014

Keren Yizhak et al Identifying anti-migratory metabolic drug targets Molecular Systems Biology

3

integrate pertaining data with a genome-scale mechanistic model

of human metabolism to study the role of the Warburg effect in

tumor progression and its potential association with cellular

migration.

Genome-scale metabolic modeling is an increasingly widely

used computational framework for studying metabolism. Given

the genome-scale metabolic model (GSMM) of a species along-

side contextual information such as growth media and ‘omics’

data, one can obtain a fairly accurate prediction of numerous

metabolic phenotypes, including growth rates, nutrient uptake

rates, gene essentiality, and more (Covert et al, 2004). GSMMs

have been used for various applications (Oberhardt et al, 2009;

Chandrasekaran & Price, 2010; Jensen & Papin, 2010; Szappanos

et al, 2011; Wessely et al, 2011; Lerman et al, 2012; Nogales

et al, 2012; Schuetz et al, 2012) including drug discovery

(Trawick & Schilling, 2006; Oberhardt et al, 2013; Yizhak et al,

2013) and metabolic engineering (Burgard et al, 2003; Pharkya

et al, 2004). Over the last few years, GSMMs have been success-

fully used for modeling human metabolism as well (Duarte et al,

2007; Ma et al, 2007; Shlomi et al, 2008; Gille et al, 2010; Lewis

et al, 2010; Mardinoglu et al, 2013). Specifically, GSMM models

of cancer cells have been reconstructed and applied for predict-

ing selective drug targets, as well as for studying the role of

tumor suppressors and oxidative stress (Folger et al, 2011; Frezza

et al, 2011; Agren et al, 2012, 2014; Jerby et al, 2012; Goldstein

et al, 2013; Gatto et al, 2014). In the context of studying the

Warburg effect, the original human metabolic model does not

predict forced lactate secretion under maximal biomass produc-

tion rate, even when oxygen consumption rate equals zero.

This renders it unsuitable for studying the Warburg effect as is,

as already noted by (Shlomi et al, 2011). While the addition of

solvent capacity constraints has been shown to overcome this

hurdle in principle (Shlomi et al, 2011), this addition requires

enzymatic kinetic data which are still largely absent on a

genome-scale.

In this study, we utilize individual genome-scale metabolic

models tailored separately to each of the NCI-60 cancer cell lines

to study the role of the Warburg effect in supporting cancer cellu-

lar migratory capacity. We first test and validate the individual

models against both existing and novel bioenergetic experimental

data. Then, we examine the extent of the Warburg effect occur-

ring in a given cancer cell line, by quantifying the glycolytic to

oxidative ATP flux ratio (AFR). We find that the AFR is highly

positively correlated with cancer cell migration, emphasizing the

role of glycolytic flux in supporting the more aggressive meta-

static stages of tumor development. To determine whether a

causal relation exists between AFR levels and cell migration, we

predict gene silencing that reduce this ratio. These potential

targets are then filtered further to exclude those predicted to

result in cell lethality. Reassuringly, the predicted targets are

found to be significantly more highly expressed in metastatic and

high-grade breast cancer tumors. Experimental investigation of

the top predicted targets via siRNA-mediated knockdown shows

that a significant portion of them truly attenuate cancer cell

migration without inducing a lethal effect. Furthermore, in accor-

dance with the predictions, a significant reduction is observed in

the ratio between ECAR and OCR levels following these genes

silencing perturbations.

Results

Stoichiometric and flux capacity constraints successfully capturethe coupling of high cell proliferation rate to lactate secretionacross individual NCI-60 cancer models

As a starting point for this study, we developed a set of metabolic

models specific for each of the NCI-60 cell lines. We built these

models using a new algorithm we have recently developed termed

PRIME, for building individual models of cells from pertaining

omics data (Yizhak et al, submitted, Supplementary Information

and Supplementary Fig S1). PRIME uses the generic human model

as a scaffold and sets maximal flux capacity constraints over a

subset of its growth-associated reactions according to the expression

levels of their corresponding catalyzing enzymes in each of the

target cell lines.

An important hallmark of cancerous cells is the production of

lactate through the Warburg effect (Warburg, 1956). As a first step

in validating the basic function of our NCI-60 models, we assessed

whether maximizing biomass forces production of lactate, which

would signify proper coupling of biomass production with lactate

output as seen in cancer cells. We found that the models indeed

must secrete lactate under biomass maximization (Supplementary

Information and Supplementary Fig S2). Hence, in contrast to the

original generic model of human metabolism, they enable us to

systematically assess the extent of lactate secretion and study the

Warburg effect across a wide range of cancer cell lines without

needing to add (mostly unknown) solvent capacity constraints, thus

identifying its functional correlates on a genome scale.

Comparing predicted versus experimentally measuredbioenergetics capacity

We compared the predicted lactate secretion rates across all cell

lines to those measured experimentally by Jain et al (Jain et al,

2012), obtaining a moderate but significant correlation (Spearman

correlation R = 0.36, P-value = 5.7e�3, Fig 1A, Materials and Meth-

ods). To further test the models’ performance under different envi-

ronmental conditions, we measured lactate secretion rates in four

breast cancer cell lines, T47D, MCF7, BT549, and Hs578T (Supple-

mentary Dataset S1), under both normoxic and hypoxic conditions

(see Materials and Methods). Utilizing the corresponding cell line

models from the NCI-60 set, we found a high correlation between

measured and predicted lactate secretion levels across both condi-

tions (Spearman correlation R = 0.95, P-value = 1.1e�3, Fig 1B).

The ratio of glycolytic versus oxidative capacity in a cell can be

quantified using its extracellular acidification rate (ECAR, a proxy of

lactate secretion) and its oxygen consumption rate (OCR). To

further examine how well our cell line models capture measured

Warburg-related activity in response to genetic perturbations, we

utilized measured ECAR and OCR levels in response to perturba-

tions in two NCI-60 lung cancer cell lines (A549 and H460), and

compared the results to predictions from our models (Materials and

Methods) (Wu et al, 2007). Qualitatively similar ECAR and OCR

changes are found in response to various enzymatic perturbations

along the glycolytic pathway. Specifically, increased glycolytic inhi-

bition resulted in reduced ECAR and elevated OCR levels in both

cells, while the maximum cellular respiration increase in H460 cells

Molecular Systems Biology 10: 744 | 2014 ª 2014 The Authors

Molecular Systems Biology Identifying anti-migratory metabolic drug targets Keren Yizhak et al

2

Fig 2D and Supplementary Table S1), it correlates even more

strongly in the positive direction with cancer cell migration

(Spearman correlation of R = 0.88, P-value = 0.03, Fig 2D and

Supplementary Table S1). Controlling for the cell lines’ measured

growth rates, this correlation becomes even more significant (partial

Spearman correlation of R = 0.96, P-value = 7e�3, Supplementary

Table S1). Overall, this finding suggests that glycolytic flux correlates

with migration rather than with growth, while OXPHOS flux exhibits

the opposite behavior. A similar association between lactate secretion

and growth rate has been recently found in an experimental study

by Jain et al (Jain et al, 2012) across the entire NCI-60 collection

(Spearman correlation of R = �0.22, P = 0.09). Furthermore,

previous studies have shown that high concentrations of lactate

correlate with a high incidence of distant metastasis (Hirschhaeuser

et al, 2011). The overall picture portrayed by these correlations is that

while glycolytic carbon diverted to biosynthetic pathways may

support cell proliferation, non-diverted glycolytic carbon supports cell

migration and metastasis (Supplementary Fig S4).

Predicting drug targets that revert the AFR and hence mayinhibit cancer migration

The congruence between AFR levels and disease severity led us to

ask if we could build upon this association to identify potential new

drug targets. We searched for drug targets predicted to reduce the

AFR ratio by simulating the knockout of each metabolic reaction

across the NCI-60 models, and examining the effects of the knock-

outs on biomass production, lactate secretion, and the AFR. As

lactate secretion is a basic indicator of the Warburg effect, we first

identified a set of 113 reactions whose knockout is predicted to

abolish lactate secretion rate in all cancer cell lines under biomass

maximization. Interestingly, the set of enzymes catalyzing these

reactions is significantly more highly expressed in the NCI-60 cell

lines than the background metabolic genes (one-sided Wilcoxon

P-value < 1.6e�8), indicating the potential oncogenic nature of

these genes.

To avoid selecting for drug-resistant clones it would be advanta-

geous to develop drugs that reduce the virulence of cancer cells but

avoid killing them. The knockout of 12 of 113 lactate-reducing reac-

tions reduces the AFR but relatively spares biomass production

(Materials and Methods and Supplementary Table S2). Importantly,

the knockout of these 12 reactions according to models of healthy

lymphoblast cells built by PRIME (Choy et al, 2008) also spares

their biomass production (Materials and Methods). Moreover, we

found that none of the lymphoblast cell lines show the forced lactate

secretion that is observed in cancer cells. While the Warburg effect

is sometimes referred in the literature as occurring in highly prolifer-

ating cells in general, our analysis finds that this phenomenon is

apparently more prominent in cancer cells, at least with regard to

the lymphoblastoid cell population studied here.

The final list of predicted gene targets includes 17 metabolic

enzymes that are associated with the final 12 reactions, spanning

glycolysis, serine, and methionine metabolism (Fig 3A). 10 of the

predicted targets have significantly higher expression levels in meta-

static versus non-metastatic breast cancer patients (Chang et al,

2005) (one-sided Wilcoxon P-value < 0.05, Fig 3B). Moreover, 9 of

the predicted targets exhibit higher expression levels in grade 3

tumors than in grade 1 tumors (Miller et al, 2005) (one-sided

Wilcoxon P-value < 0.05, Fig 3C). Finally, lower expression of nine

of the predicted targets is significantly associated with improved

long-term survival (Curtis et al, 2012) (log-rank P-value < 0.05,

Fig 3D), testifying for their potential role as therapeutic targets. All

P-values are corrected for multiple hypothesis using FDR with a = 0.05.

siRNA-mediated gene knockdown experiments testing thepredicted targets

To experimentally test our predictions we silenced the 17 predicted

AFR-reducing genes and examined their phenotypic effects in the

MDA-MB-231, MDA-MB-435, BT549, and A549 cell lines. Knock-

down experiments were performed with SmartPools from Dharma-

con using a live cell migration and fixed proliferation assays

(Materials and Methods). 8–13 out of the 17 enzymes (8–10 out of

12 metabolic reactions) were found to significantly attenuate migra-

tion speed in each cell line (two-sided t-test P-value < 0.05, FDR

corrected with a = 0.05, Fig 4, Materials and Methods and Supple-

mentary Dataset S4). This result is highly significant as only 17% of

the metabolic genes were found to impair cell migration in a siRNA

screen of 190 metabolic genes (Fokkelman M, Rogkoti VM et al,

unpublished data, Bernoulli P-value in the range of 3.9e�3 and

1.18e�7). Of note, the association between the gene expression of

the predicted targets and the measured migration speed is insignifi-

cant for all targets but one, testifying for the inherent value of our

model-based prediction analysis (Supplementary Table S3). It

should also be noted that the knockdown of the three splices of the

enolase gene have almost no significant effect on these cells’ migra-

tion speed, possibly because of isoenzymes backup mechanisms.

Importantly, most of the gene knockdown experiments do not

manifest any significant effects on cell proliferation (Fig 4). In

accordance with the findings of Simpson et al (Simpson et al,

2008), we found that the correlation between the reduction in

migration speed and reduction in proliferation rate is mostly

insignificant (Supplementary Dataset S4), suggesting that the

reduced migration observed is not simply a consequence of

common mechanisms hindering proliferation, but rather that it

occurs due to the disruption of distinct migratory-associated

metabolic pathways.

ECAR and OCR levels following selected gene silencing

To further study the association between reduced AFR levels and

impaired cell migration we used the Seahorse XF96 extracellular

flux analyzer to measure both ECAR and OCR fluxes in the MDA-

MB-231 cell line, following knockdown of a selected group of targets

(Materials and Methods and Supplementary Fig S6). As the AFR

measure is very difficult to measure experimentally, we tested the

conventionally measured EOR (ECAR/OCR) as its proxy. We

focused on a subset of seven genes (Fig 5) whose knockdown is

predicted to have the highest effect on cell migration and span all

three predicted metabolic pathways. As shown in Fig 5, a significant

EOR reduction versus the control is found for all seven examined

genes (two-sided t-test P-value < 0.05, FDR corrected with a = 0.05,

Materials and Methods and Supplementary Table S4). The silencing

of the four glycolytic genes (HK2, PGAM1, PGK2, and GAPDH)

results in both decreased ECAR and increased OCR levels, while

the silencing of the serine- and methionine-associated genes

ª 2014 The Authors Molecular Systems Biology 10: 744 | 2014

Keren Yizhak et al Identifying anti-migratory metabolic drug targets Molecular Systems Biology

5

growth. The effect of most of these compounds is also negatively

correlated with the cells’ growth rates, suggesting that slowly

proliferating cells are more resistant to treatment (similar results were

previously shown for compounds targeting cell growth (Penault-

Llorca et al, 2009; Vincent-Salomon et al, 2004)). Interestingly, the

response to many compounds in this dataset shows a significant

association with the AFR measure while having no association with

the cells’ growth rate. 133 such compounds were identified (Supple-

mentary Dataset S3), possibly suggesting that their mechanism

might be related to the Warburg level of the cells rather than to their

proliferation. Finally, predicted AFR values correctly separate

between epithelial and mesenchymal breast cancer cell lines (with

the more aggressive mesenchymal cell lines exhibiting larger

Warburg effect (Sarrio et al, 2008), Fig 2A). Once again, the AFR

was more predictive of this experimental observation than the EOR

(Supplementary Dataset S2).

We next turned to our primary objective of examining the rela-

tion between the Warburg effect and tumor proliferation and migra-

tion. To this end, we experimentally measured the migration speed

of six NCI-60 breast cancer cell lines (Fig 2B and C, Materials and

Methods, Supplementary Fig S3, and Supplementary Dataset S2)

and utilized publically available measured growth rates for these

cell lines. While the AFR correlates markedly negatively with cell

growth rate (Spearman correlation of R = �0.55, P-value = 4.53e�6,

* *

*

A

Basa

lLu

min

al

DB

C

Figure 2. Association between AFR levels and cell proliferation and migration.

A The 20 cell lines that are predicted to exhibit the Warburg effect to the greatest/least extent according to the AFR measure. The x-axis and y-axis represent the meanand SD of the normalized ATP flux rate in glycolysis and OXPHOS, respectively (Materials and Methods). The AFR measure correctly separates between mesenchymal(orange) and epithelial cell lines (green), showing that the former (which are known to be more aggressive) have higher AFR levels.

B We analyzed a panel of six breast cancer cell lines for their migration capacity using live cell imaging. Differential Interference Contrast (DIC) images of the six celllines in the order of their respective migration speed (from low to high), scale bar is 100 lm (Materials and Methods).

C The average migration speed of cells followed for 12 h in complete medium. Error bars represent SEM; the number of samples is between n = 100 and n = 200.D The correlation of predicted model-based EOR and AFR measures to growth and migration rates measured experimentally. Both measures represent a negative

correlation with growth and a positive correlation with migration rates. Significant results (P-value < 0.05) are marked with an asterisk.

Molecular Systems Biology 10: 744 | 2014 ª 2014 The Authors

Molecular Systems Biology Identifying anti-migratory metabolic drug targets Keren Yizhak et al

4

Fig 2D and Supplementary Table S1), it correlates even more

strongly in the positive direction with cancer cell migration

(Spearman correlation of R = 0.88, P-value = 0.03, Fig 2D and

Supplementary Table S1). Controlling for the cell lines’ measured

growth rates, this correlation becomes even more significant (partial

Spearman correlation of R = 0.96, P-value = 7e�3, Supplementary

Table S1). Overall, this finding suggests that glycolytic flux correlates

with migration rather than with growth, while OXPHOS flux exhibits

the opposite behavior. A similar association between lactate secretion

and growth rate has been recently found in an experimental study

by Jain et al (Jain et al, 2012) across the entire NCI-60 collection

(Spearman correlation of R = �0.22, P = 0.09). Furthermore,

previous studies have shown that high concentrations of lactate

correlate with a high incidence of distant metastasis (Hirschhaeuser

et al, 2011). The overall picture portrayed by these correlations is that

while glycolytic carbon diverted to biosynthetic pathways may

support cell proliferation, non-diverted glycolytic carbon supports cell

migration and metastasis (Supplementary Fig S4).

Predicting drug targets that revert the AFR and hence mayinhibit cancer migration

The congruence between AFR levels and disease severity led us to

ask if we could build upon this association to identify potential new

drug targets. We searched for drug targets predicted to reduce the

AFR ratio by simulating the knockout of each metabolic reaction

across the NCI-60 models, and examining the effects of the knock-

outs on biomass production, lactate secretion, and the AFR. As

lactate secretion is a basic indicator of the Warburg effect, we first

identified a set of 113 reactions whose knockout is predicted to

abolish lactate secretion rate in all cancer cell lines under biomass

maximization. Interestingly, the set of enzymes catalyzing these

reactions is significantly more highly expressed in the NCI-60 cell

lines than the background metabolic genes (one-sided Wilcoxon

P-value < 1.6e�8), indicating the potential oncogenic nature of

these genes.

To avoid selecting for drug-resistant clones it would be advanta-

geous to develop drugs that reduce the virulence of cancer cells but

avoid killing them. The knockout of 12 of 113 lactate-reducing reac-

tions reduces the AFR but relatively spares biomass production

(Materials and Methods and Supplementary Table S2). Importantly,

the knockout of these 12 reactions according to models of healthy

lymphoblast cells built by PRIME (Choy et al, 2008) also spares

their biomass production (Materials and Methods). Moreover, we

found that none of the lymphoblast cell lines show the forced lactate

secretion that is observed in cancer cells. While the Warburg effect

is sometimes referred in the literature as occurring in highly prolifer-

ating cells in general, our analysis finds that this phenomenon is

apparently more prominent in cancer cells, at least with regard to

the lymphoblastoid cell population studied here.

The final list of predicted gene targets includes 17 metabolic

enzymes that are associated with the final 12 reactions, spanning

glycolysis, serine, and methionine metabolism (Fig 3A). 10 of the

predicted targets have significantly higher expression levels in meta-

static versus non-metastatic breast cancer patients (Chang et al,

2005) (one-sided Wilcoxon P-value < 0.05, Fig 3B). Moreover, 9 of

the predicted targets exhibit higher expression levels in grade 3

tumors than in grade 1 tumors (Miller et al, 2005) (one-sided

Wilcoxon P-value < 0.05, Fig 3C). Finally, lower expression of nine

of the predicted targets is significantly associated with improved

long-term survival (Curtis et al, 2012) (log-rank P-value < 0.05,

Fig 3D), testifying for their potential role as therapeutic targets. All

P-values are corrected for multiple hypothesis using FDR with a = 0.05.

siRNA-mediated gene knockdown experiments testing thepredicted targets

To experimentally test our predictions we silenced the 17 predicted

AFR-reducing genes and examined their phenotypic effects in the

MDA-MB-231, MDA-MB-435, BT549, and A549 cell lines. Knock-

down experiments were performed with SmartPools from Dharma-

con using a live cell migration and fixed proliferation assays

(Materials and Methods). 8–13 out of the 17 enzymes (8–10 out of

12 metabolic reactions) were found to significantly attenuate migra-

tion speed in each cell line (two-sided t-test P-value < 0.05, FDR

corrected with a = 0.05, Fig 4, Materials and Methods and Supple-

mentary Dataset S4). This result is highly significant as only 17% of

the metabolic genes were found to impair cell migration in a siRNA

screen of 190 metabolic genes (Fokkelman M, Rogkoti VM et al,

unpublished data, Bernoulli P-value in the range of 3.9e�3 and

1.18e�7). Of note, the association between the gene expression of

the predicted targets and the measured migration speed is insignifi-

cant for all targets but one, testifying for the inherent value of our

model-based prediction analysis (Supplementary Table S3). It

should also be noted that the knockdown of the three splices of the

enolase gene have almost no significant effect on these cells’ migra-

tion speed, possibly because of isoenzymes backup mechanisms.

Importantly, most of the gene knockdown experiments do not

manifest any significant effects on cell proliferation (Fig 4). In

accordance with the findings of Simpson et al (Simpson et al,

2008), we found that the correlation between the reduction in

migration speed and reduction in proliferation rate is mostly

insignificant (Supplementary Dataset S4), suggesting that the

reduced migration observed is not simply a consequence of

common mechanisms hindering proliferation, but rather that it

occurs due to the disruption of distinct migratory-associated

metabolic pathways.

ECAR and OCR levels following selected gene silencing

To further study the association between reduced AFR levels and

impaired cell migration we used the Seahorse XF96 extracellular

flux analyzer to measure both ECAR and OCR fluxes in the MDA-

MB-231 cell line, following knockdown of a selected group of targets

(Materials and Methods and Supplementary Fig S6). As the AFR

measure is very difficult to measure experimentally, we tested the

conventionally measured EOR (ECAR/OCR) as its proxy. We

focused on a subset of seven genes (Fig 5) whose knockdown is

predicted to have the highest effect on cell migration and span all

three predicted metabolic pathways. As shown in Fig 5, a significant

EOR reduction versus the control is found for all seven examined

genes (two-sided t-test P-value < 0.05, FDR corrected with a = 0.05,

Materials and Methods and Supplementary Table S4). The silencing

of the four glycolytic genes (HK2, PGAM1, PGK2, and GAPDH)

results in both decreased ECAR and increased OCR levels, while

the silencing of the serine- and methionine-associated genes

ª 2014 The Authors Molecular Systems Biology 10: 744 | 2014

Keren Yizhak et al Identifying anti-migratory metabolic drug targets Molecular Systems Biology

5

growth. The effect of most of these compounds is also negatively

correlated with the cells’ growth rates, suggesting that slowly

proliferating cells are more resistant to treatment (similar results were

previously shown for compounds targeting cell growth (Penault-

Llorca et al, 2009; Vincent-Salomon et al, 2004)). Interestingly, the

response to many compounds in this dataset shows a significant

association with the AFR measure while having no association with

the cells’ growth rate. 133 such compounds were identified (Supple-

mentary Dataset S3), possibly suggesting that their mechanism

might be related to the Warburg level of the cells rather than to their

proliferation. Finally, predicted AFR values correctly separate

between epithelial and mesenchymal breast cancer cell lines (with

the more aggressive mesenchymal cell lines exhibiting larger

Warburg effect (Sarrio et al, 2008), Fig 2A). Once again, the AFR

was more predictive of this experimental observation than the EOR

(Supplementary Dataset S2).

We next turned to our primary objective of examining the rela-

tion between the Warburg effect and tumor proliferation and migra-

tion. To this end, we experimentally measured the migration speed

of six NCI-60 breast cancer cell lines (Fig 2B and C, Materials and

Methods, Supplementary Fig S3, and Supplementary Dataset S2)

and utilized publically available measured growth rates for these

cell lines. While the AFR correlates markedly negatively with cell

growth rate (Spearman correlation of R = �0.55, P-value = 4.53e�6,

* *

*

A

Basa

lLu

min

al

DB

C

Figure 2. Association between AFR levels and cell proliferation and migration.

A The 20 cell lines that are predicted to exhibit the Warburg effect to the greatest/least extent according to the AFR measure. The x-axis and y-axis represent the meanand SD of the normalized ATP flux rate in glycolysis and OXPHOS, respectively (Materials and Methods). The AFR measure correctly separates between mesenchymal(orange) and epithelial cell lines (green), showing that the former (which are known to be more aggressive) have higher AFR levels.

B We analyzed a panel of six breast cancer cell lines for their migration capacity using live cell imaging. Differential Interference Contrast (DIC) images of the six celllines in the order of their respective migration speed (from low to high), scale bar is 100 lm (Materials and Methods).

C The average migration speed of cells followed for 12 h in complete medium. Error bars represent SEM; the number of samples is between n = 100 and n = 200.D The correlation of predicted model-based EOR and AFR measures to growth and migration rates measured experimentally. Both measures represent a negative

correlation with growth and a positive correlation with migration rates. Significant results (P-value < 0.05) are marked with an asterisk.

Molecular Systems Biology 10: 744 | 2014 ª 2014 The Authors

Molecular Systems Biology Identifying anti-migratory metabolic drug targets Keren Yizhak et al

4

and the b-catalytic subunit of ATP synthase forming the BEC index

was found to have a prognostic value in assessing the clinical

outcome of patients with early-stage colorectal carcinomas. The

AFR measure and the BEC index (as computed by its corresponding

RNA levels) are significantly correlated (Spearman R = 0.58,

P-value = 1.6e�6) across the NCI-60 cell lines, and the BEC index is

perfectly correlated with migration speed across the six breast

cancer cell lines (Spearman R = 1, P-value = 2.8e�3). However, the

BEC index has inferior performance in predicting drug response

(Supplementary Table S1).

The finding that enhanced glycolytic activity plays a key role in

cancer cell migration is also in line with a very recent study by

De Bock et al, showing that glycolysis is the major source of ATP

production in endothelial cells and that the silencing of the glyco-

lytic regulator PFKFB3 impairs the cell migration capacity and inter-

feres with vessel sprouting (De Bock et al, 2013). In addition,

silencing of PFKFB3 was shown to suppress cell proliferation in

about 50% (De Bock et al, 2013). Overall, the results presented in

this study, as well as findings reported by others (Simpson et al,

2008), suggest that proliferation and migration are not mutually

exclusive, and the effect of potential targets on both processes

should be carefully examined.

Some of our predicted targets have been previously studied in

the context of cell proliferation as well (Cheong et al, 2012).

Possemato et al (Possemato et al, 2011) have showed that suppres-

sion of PHGDH in cell lines with elevated PHGDH expression, but not

Figure 4. Normalized to control mean speed per SmartPool gene silencing of the predicted targets.

A–D The four different cell lines that were analyzed: MDA-MB-231, MDA-MB-435s, BT549, and A549. Significant results (two-sided t-test, P-value < 0.05 after correctingfor multiple hypothesis using FDR with a = 0.05) are marked with an asterisk. Two different controls are used: (1) non-targeting siRNA (= negative control); and (2)a positive control DNM2 which is known to block both migration and proliferation (Ezratty et al, 2005). Left panel shows migration speed and right panel showsnuclear count. Error bars represent SD; the number of samples is n = 3.

ª 2014 The Authors Molecular Systems Biology 10: 744 | 2014

Keren Yizhak et al Identifying anti-migratory metabolic drug targets Molecular Systems Biology

7

(PSPH, AHCY, and PHGDH) results with decreased ECAR solely

(Fig 5A). Furthermore, a matching significant difference in

experimentally measured EOR levels is found between the lowest and

highest AFR-reducing genes (one-sided Wilcoxon P-value = 0.05).

Overall, taken together our results testify that, as predicted, the

knockdown of the top-ranked genes results in attenuated cell

migration that is accompanied by reduced EOR and AFR levels.

Discussion

In this study we explored the role of the Warburg effect in support-

ing tumor migration, going beyond recent investigations focusing on

its role in assisting cancer proliferation. A model-based investigation

across cancer cell lines shows that the ratio between glycolytic and

oxidative ATP flux rate is significantly associated with cancer migra-

tory behavior. Gene silencing perturbations predicted to reduce this

ratio were indeed found to attenuate cell migration, and result with

a significant reduction in ECAR to OCR levels. Of note, our modeling

approach relies on gene expression differences between the cells

and does not take into account specific uptake rates. It is therefore

more suited for capturing qualitative rather than exact quantitative

differences between the cells, as demonstrated throughout the

paper. Moreover, the lion share of our analysis is focused on the

simulations of perturbations where specific uptake rates are not

available. Nonetheless, utilizing such uptake measurements can

significantly increase the correlation to the measured lactate rates

(Spearman correlation R = 0.67, P-value = 1.5e�8), suggesting that

uptake rates measurements under perturbation states can signifi-

cantly increase the models’ prediction power.

Our AFR measure is conceptually analogous to a bioenergetic

(BEC) index previously introduced by Cuezva et al (Cuezva et al,

2002). In that study, the ratio between the expression of the glyco-

lytic enzyme glyceraldehyde-3-phosphate dehydrogenase (GAPDH)

Glucose

G6P

F6P

FBP

G3PDHAP

1,3BPG

3PG

2PG

PEP

PyruvateLactateGln

Pyruvate

Ac-CoA

To PentosePhosphate Pathway

R5P

TCACycle

Glycolysis

Serine

Glycine

mTHF

THF

HK2

PKM

GAPDHTPI1

PGAM

PGK3PHP

NAD NADH

PHGDHSerineP-Serine

α-kgGlu

PSAT1 PSPH

L-Met adenosyl-Met

MAT1/2

Serine Biosynthesis

Methionine Metabolism

ENO

adenosyl-hcys

adn + HcysAHCY/AHCYL

A

P = 5.34e-4

P = 3.8e-5

P = 1.08e-7

P = 2.4e-3

P = 2.69e-5

P = 4.47e-5

P = 2.2e-6

P = 4.3e-4

P = 1.6e-2

P = 1.4e-2

P = 5.34e-5

P = 9.8e-3

P = 8.52e-5

P = 4.05e-4

P = 1.2-3

P = 2.78e-8

P = 7.08e-13

P = 4.61e-7

P = 1.83e-4

P = 4.61e-8

P = 1.96e-6

P = 6.08e-4

P = 6.12e-5P = 1.86e-5

P = 3.75e-4P = 1.5e-3

P = 1.1e-4P = 2.85e-4

B

C

D

Figure 3. Gene targets that are predicted to reduce the AFR and their association with prognostic markers of breast cancer patients.

A A schematic representation of the 12 predicted gene targets, marked in red.B Ten predicted targets that show a significantly higher expression in metastatic versus non-metastatic tumor samples (n = 295).C Nine predicted targets that show a significantly higher expression in grade 3 versus grade 1 tumor samples (n = 236).D Nine predicted targets whose lower expression is significantly associated with improved long-term survival (n = 1568).

Molecular Systems Biology 10: 744 | 2014 ª 2014 The Authors

Molecular Systems Biology Identifying anti-migratory metabolic drug targets Keren Yizhak et al

6

and the b-catalytic subunit of ATP synthase forming the BEC index

was found to have a prognostic value in assessing the clinical

outcome of patients with early-stage colorectal carcinomas. The

AFR measure and the BEC index (as computed by its corresponding

RNA levels) are significantly correlated (Spearman R = 0.58,

P-value = 1.6e�6) across the NCI-60 cell lines, and the BEC index is

perfectly correlated with migration speed across the six breast

cancer cell lines (Spearman R = 1, P-value = 2.8e�3). However, the

BEC index has inferior performance in predicting drug response

(Supplementary Table S1).

The finding that enhanced glycolytic activity plays a key role in

cancer cell migration is also in line with a very recent study by

De Bock et al, showing that glycolysis is the major source of ATP

production in endothelial cells and that the silencing of the glyco-

lytic regulator PFKFB3 impairs the cell migration capacity and inter-

feres with vessel sprouting (De Bock et al, 2013). In addition,

silencing of PFKFB3 was shown to suppress cell proliferation in

about 50% (De Bock et al, 2013). Overall, the results presented in

this study, as well as findings reported by others (Simpson et al,

2008), suggest that proliferation and migration are not mutually

exclusive, and the effect of potential targets on both processes

should be carefully examined.

Some of our predicted targets have been previously studied in

the context of cell proliferation as well (Cheong et al, 2012).

Possemato et al (Possemato et al, 2011) have showed that suppres-

sion of PHGDH in cell lines with elevated PHGDH expression, but not

Figure 4. Normalized to control mean speed per SmartPool gene silencing of the predicted targets.

A–D The four different cell lines that were analyzed: MDA-MB-231, MDA-MB-435s, BT549, and A549. Significant results (two-sided t-test, P-value < 0.05 after correctingfor multiple hypothesis using FDR with a = 0.05) are marked with an asterisk. Two different controls are used: (1) non-targeting siRNA (= negative control); and (2)a positive control DNM2 which is known to block both migration and proliferation (Ezratty et al, 2005). Left panel shows migration speed and right panel showsnuclear count. Error bars represent SD; the number of samples is n = 3.

ª 2014 The Authors Molecular Systems Biology 10: 744 | 2014

Keren Yizhak et al Identifying anti-migratory metabolic drug targets Molecular Systems Biology

7

(PSPH, AHCY, and PHGDH) results with decreased ECAR solely

(Fig 5A). Furthermore, a matching significant difference in

experimentally measured EOR levels is found between the lowest and

highest AFR-reducing genes (one-sided Wilcoxon P-value = 0.05).

Overall, taken together our results testify that, as predicted, the

knockdown of the top-ranked genes results in attenuated cell

migration that is accompanied by reduced EOR and AFR levels.

Discussion

In this study we explored the role of the Warburg effect in support-

ing tumor migration, going beyond recent investigations focusing on

its role in assisting cancer proliferation. A model-based investigation

across cancer cell lines shows that the ratio between glycolytic and

oxidative ATP flux rate is significantly associated with cancer migra-

tory behavior. Gene silencing perturbations predicted to reduce this

ratio were indeed found to attenuate cell migration, and result with

a significant reduction in ECAR to OCR levels. Of note, our modeling

approach relies on gene expression differences between the cells

and does not take into account specific uptake rates. It is therefore

more suited for capturing qualitative rather than exact quantitative

differences between the cells, as demonstrated throughout the

paper. Moreover, the lion share of our analysis is focused on the

simulations of perturbations where specific uptake rates are not

available. Nonetheless, utilizing such uptake measurements can

significantly increase the correlation to the measured lactate rates

(Spearman correlation R = 0.67, P-value = 1.5e�8), suggesting that

uptake rates measurements under perturbation states can signifi-

cantly increase the models’ prediction power.

Our AFR measure is conceptually analogous to a bioenergetic

(BEC) index previously introduced by Cuezva et al (Cuezva et al,

2002). In that study, the ratio between the expression of the glyco-

lytic enzyme glyceraldehyde-3-phosphate dehydrogenase (GAPDH)

Glucose

G6P

F6P

FBP

G3PDHAP

1,3BPG

3PG

2PG

PEP

PyruvateLactateGln

Pyruvate

Ac-CoA

To PentosePhosphate Pathway

R5P

TCACycle

Glycolysis

Serine

Glycine

mTHF

THF

HK2

PKM

GAPDHTPI1

PGAM

PGK3PHP

NAD NADH

PHGDHSerineP-Serine

α-kgGlu

PSAT1 PSPH

L-Met adenosyl-Met

MAT1/2

Serine Biosynthesis

Methionine Metabolism

ENO

adenosyl-hcys

adn + HcysAHCY/AHCYL

A

P = 5.34e-4

P = 3.8e-5

P = 1.08e-7

P = 2.4e-3

P = 2.69e-5

P = 4.47e-5

P = 2.2e-6

P = 4.3e-4

P = 1.6e-2

P = 1.4e-2

P = 5.34e-5

P = 9.8e-3

P = 8.52e-5

P = 4.05e-4

P = 1.2-3

P = 2.78e-8

P = 7.08e-13

P = 4.61e-7

P = 1.83e-4

P = 4.61e-8

P = 1.96e-6

P = 6.08e-4

P = 6.12e-5P = 1.86e-5

P = 3.75e-4P = 1.5e-3

P = 1.1e-4P = 2.85e-4

B

C

D

Figure 3. Gene targets that are predicted to reduce the AFR and their association with prognostic markers of breast cancer patients.

A A schematic representation of the 12 predicted gene targets, marked in red.B Ten predicted targets that show a significantly higher expression in metastatic versus non-metastatic tumor samples (n = 295).C Nine predicted targets that show a significantly higher expression in grade 3 versus grade 1 tumor samples (n = 236).D Nine predicted targets whose lower expression is significantly associated with improved long-term survival (n = 1568).

Molecular Systems Biology 10: 744 | 2014 ª 2014 The Authors

Molecular Systems Biology Identifying anti-migratory metabolic drug targets Keren Yizhak et al

6

corresponding metabolic reaction to zero. The biomass function

utilized here is taken from (Folger et al, 2011). The media simu-

lated in all the analyses throughout the paper is the RPMI-1640

media that was used to grow the cell lines experimentally (Lee

et al, 2007; Choy et al, 2008).

Building cell-specific metabolic models and computing lactate secretion

Our method to reconstruct the NCI-60 cancer cell lines (see Supple-

mentary Material, based on the yet unpublished methods in Yizhak

et al, submitted) required several key inputs: (a) the generic human

model (Duarte et al, 2007), (b) gene expression data for each cancer

cell line from (Lee et al, 2007), and (c) growth rate measurements.

The algorithm then reconstructs a specific metabolic model for each

sample by modifying the upper bounds of growth-associated reac-

tions in accordance with their gene expression (Note: the growth

rates were used only to determine which reactions should be used

in constraining the models, in order to obtain models that were as

physiologically relevant as possible; they were not used to deter-

mine reaction bounds). A similar procedure was used to reconstruct

the lymphoblast metabolic models (Choy et al, 2008) for compari-

son against normal proliferating cells. A more detailed description is

found in the Supplementary Material.

Simulations of the Warburg effect include the examination of

minimal lactate production rate under different demands for

biomass production, glucose, glutamine, and oxygen uptake rates

(Supplementary Material). We examined the minimal value of

lactate secretion as it testifies whether or not the cell is enforced to

secrete lactate under a given condition (Supplementary Fig S1). All

the correlations reported in the paper are Spearman rank correla-

tions and their associated P-values are computed using the exact

permutation distribution.

Calculating wild-type and perturbed lactate secretion rates and

OCR levels

For simulating lactate secretion under normoxic conditions (when

comparing to Jain et al (Jain et al, 2012), Wu et al (Wu et al, 2007)

and the breast cancer data collected in this paper), oxygen maximal

uptake rate was set to the highest value under which minimal

lactate secretion is positive. Since metabolic models are designed to

maximize growth yield rather than growth rate, using an unlimited

amount of oxygen in GSMM simulations will result in a state where

the minimal lactate secretion rate equals zero. However, it’s impor-

tant to note that even under the limited oxygen levels simulated

here, the generic human model doesn’t show lactate secretion (as

opposed to the NCI-60 cancer cell line models described above). For

simulating the hypoxic conditions measured here for the breast

cancer cell lines, we lowered the oxygen maximal uptake rate by

50% of its normoxic state as described above. Under each of these

conditions, we sampled the solution space under maximal biomass

yield and obtained 1,000 feasible flux distributions (Bordel et al,

2010). The predicted lactate secretion rate is the average lactate

secretion flux over these samples. For emulating the perturbation

experiments in Wu et al we gradually lowered the bound of the

corresponding compound target (from the maximal bound to 0)

and repeated the procedure described above for computing

the ECAR (lactate secretion) and the OCR, which in a similar

manner is defined as the average oxygen consumption flux across

all samples.

Calculating the EOR and AFR measures for assessing the Warburg level

of the cell lines and using them to predict drug response

The EOR and AFR measures were calculated in a similar manner to

that described above. Specifically, the EOR is calculated as the mean

over lactate secretion across all samples divided by the mean over

oxygen consumption across all samples. Similarly, the AFR is calcu-

lated as the mean flux carried by the reactions producing ATP in

glycolysis versus the mean flux carried by the reaction producing

ATP in OXPHOS. To determine an empiric P-value in the drug

response analysis we randomly shuffled the drug response data

1,000 times, each time examining the resulting Wilcoxon P-value

over the original set of cell lines.

Predicting the effect of reaction knockouts

Each metabolic reaction in each cell line model is perturbed by

constraining its flux to zero. Under each perturbation the minimal

lactate secretion (under maximal growth rate) and the maximal

growth rate is calculated. The set of reactions that eliminate forced

lactate secretion while maintaining a level of cell growth that is

> 10% of the wild-type growth prediction is further tested for the

AFR level. The mean AFR level for each cell line under each of these

perturbations is calculated over 1,000 flux distribution samples as

described above. The final set of predicted reactions includes those

whose knockout reduces the AFR to below 60% of its wild-type level.

Datasets

Growth rate measurements and drug response data were down-

loaded from the NCI website.

Growth rate: http://dtp.nci.nih.gov/docs/misc/common_files/

cell_list.html

Drug response: http://discover.nci.nih.gov/nature2000/naturein

tromain.jsp

Experimentally measuring lactate secretion rates of breastcancer cell lines

Cell Culture

The MCF7, T47D, Hs578T and BT549 breast cancer cell lines were

obtained from the American Type Culture Collection and London

Research Institute Cell Services. Cells were cultured in DMEM/F12

(1:1), with 2 mM L-glutamine and penicillin/streptomycin. Medium

was supplemented with 10% FCS (GIBCO) for the cancer cell lines

and 5% horse serum, 20 ng/ml EGF, 5 lg/ml hydrocortisone,

10 lg/ml insulin, and 100 ng/ml cholera toxin for the non-

malignant cell lines.

Lactate secretion measurements

Cells were cultured under normoxic (20% O2) and hypoxic (0.5%

O2) conditions for 72 h. Cells were starved of glucose and glutamine

for 1 h and full medium was added for 1 h. Lactate secretion was

determined from normoxic and hypoxic cells and normalized to

cell growth (increase in total protein during the 72 h incubation in

normoxia). Lactate concentrations in media incubated with or

without cells were determined using lactate assay kits (BioVision).

Total protein content determined by Sulforhodamine B assay was

used for normalization. Two experiments were performed with

three or four biologically independent replicates (total of seven

replicates).

ª 2014 The Authors Molecular Systems Biology 10: 744 | 2014

Keren Yizhak et al Identifying anti-migratory metabolic drug targets Molecular Systems Biology

9

in those without, inhibits cell proliferation. Accordingly, as PHGDH

is not amplified in the cell line MDA-MB-231 which was examined

in both studies, its suppression is indeed non-lethal. However, we

show that its suppression significantly attenuates cell migration,

suggesting that metabolic enzymes can promote different cancerous

phenotypes in different cancer cells.

Remarkably, analyzing the model-predicted flux rates has

successfully uncovered a fundamental association between the AFR

and cancer migration, even given the relatively small set of cell lines

for which migration was measured. Our analysis has also revealed

other potential associations between individual fluxes and cell

migration (Supplementary Fig S4). However, future studies measur-

ing cellular migration data across a much wider array of cell lines

(of the order for which we already have proliferation data) are

needed to determine the actual significance of these potential leads.

As this study has shown, cellular proliferation and migration have

distinct underlying metabolite correlates; understanding the meta-

bolic correlates that are strongly associated with cell migration may

lead to new anti-metastatic treatment opportunities. It is important

to note, however, that while the inhibition of migration alone might

be a good strategy for avoiding the adverse side effects of cytotoxic

treatment, cell migration is a crucial process also in normal physiol-

ogy, for instance, in immune response and tissue repair (Forster

et al, 1999; Ridley et al, 2003). Therefore, future anti-migratory

drugs may pose different drug selectivity challenges that should be

carefully addressed in the future studies. Irrespectively, they may

result in lesser clonal selection, and as a result, their usage may be

accompanied with lesser rate of emergence of drug-resistant clones.

Materials and Methods

Computational methods

Genome-scale metabolic modeling (GSSM)

A metabolic network consisting of m metabolites and n reactions

can be represented by a stoichiometric matrix S, where the entry Sijrepresents the stoichiometric coefficient of metabolite i in reaction j

(Price et al, 2004). A CBM model imposes mass balance, directional-

ity, and flux capacity constraints on the space of possible fluxes in

the metabolic network’s reactions through a set of linear equations:

Sv ¼ 0 (1)

vmin � v� vmax (2)

where v stands for the flux vector for all of the reactions in the

model (i.e. the flux distribution). The exchange of metabolites with

the environment is represented as a set of exchange (transport)

reactions, enabling a pre-defined set of metabolites to be either

taken up or secreted from the growth media. The steady-state

assumption represented in equation (1) constrains the production

rate of each metabolite to be equal to its consumption rate. Enzy-

matic directionality and flux capacity constraints define lower and

upper bounds on the fluxes and are embedded in equation (2).

In the following, flux vectors satisfying these conditions will

be referred to as feasible steady-state flux distributions. Gene

knockouts are simulated by constraining the flux through the

Figure 5. ECAR and OCR levels of top predicted gene targets.

A Mean and SEM (normalized to nuclear count) ECAR and OCR levels after silencing of seven different genes (HK2, PGAM1, PGK2, GAPDH, PSPH, AHCY, and PHGDH)compared to the control. Silencing of the four glycolytic genes results in both a decrease in ECAR levels (x-axis) and an increase in OCR levels (y-axis), while theserine- and methionine-associated genes show only a decrease in ECAR levels. Error bars represent SEM. The number of samples is n = 18.

B Mean and SD of computed ECAR/OCR (EOR) levels for control and selected gene silencing (Materials and Methods). For all genes a significant reduction in EOR levelsis observed. Error bars represent SD. The number of samples is n = 18.

Molecular Systems Biology 10: 744 | 2014 ª 2014 The Authors

Molecular Systems Biology Identifying anti-migratory metabolic drug targets Keren Yizhak et al

8

corresponding metabolic reaction to zero. The biomass function

utilized here is taken from (Folger et al, 2011). The media simu-

lated in all the analyses throughout the paper is the RPMI-1640

media that was used to grow the cell lines experimentally (Lee

et al, 2007; Choy et al, 2008).

Building cell-specific metabolic models and computing lactate secretion

Our method to reconstruct the NCI-60 cancer cell lines (see Supple-

mentary Material, based on the yet unpublished methods in Yizhak

et al, submitted) required several key inputs: (a) the generic human

model (Duarte et al, 2007), (b) gene expression data for each cancer

cell line from (Lee et al, 2007), and (c) growth rate measurements.

The algorithm then reconstructs a specific metabolic model for each

sample by modifying the upper bounds of growth-associated reac-

tions in accordance with their gene expression (Note: the growth

rates were used only to determine which reactions should be used

in constraining the models, in order to obtain models that were as

physiologically relevant as possible; they were not used to deter-

mine reaction bounds). A similar procedure was used to reconstruct

the lymphoblast metabolic models (Choy et al, 2008) for compari-

son against normal proliferating cells. A more detailed description is

found in the Supplementary Material.

Simulations of the Warburg effect include the examination of

minimal lactate production rate under different demands for

biomass production, glucose, glutamine, and oxygen uptake rates

(Supplementary Material). We examined the minimal value of

lactate secretion as it testifies whether or not the cell is enforced to

secrete lactate under a given condition (Supplementary Fig S1). All

the correlations reported in the paper are Spearman rank correla-

tions and their associated P-values are computed using the exact

permutation distribution.

Calculating wild-type and perturbed lactate secretion rates and

OCR levels

For simulating lactate secretion under normoxic conditions (when

comparing to Jain et al (Jain et al, 2012), Wu et al (Wu et al, 2007)

and the breast cancer data collected in this paper), oxygen maximal

uptake rate was set to the highest value under which minimal

lactate secretion is positive. Since metabolic models are designed to

maximize growth yield rather than growth rate, using an unlimited

amount of oxygen in GSMM simulations will result in a state where

the minimal lactate secretion rate equals zero. However, it’s impor-

tant to note that even under the limited oxygen levels simulated

here, the generic human model doesn’t show lactate secretion (as

opposed to the NCI-60 cancer cell line models described above). For

simulating the hypoxic conditions measured here for the breast

cancer cell lines, we lowered the oxygen maximal uptake rate by

50% of its normoxic state as described above. Under each of these

conditions, we sampled the solution space under maximal biomass

yield and obtained 1,000 feasible flux distributions (Bordel et al,

2010). The predicted lactate secretion rate is the average lactate

secretion flux over these samples. For emulating the perturbation

experiments in Wu et al we gradually lowered the bound of the

corresponding compound target (from the maximal bound to 0)

and repeated the procedure described above for computing

the ECAR (lactate secretion) and the OCR, which in a similar

manner is defined as the average oxygen consumption flux across

all samples.

Calculating the EOR and AFR measures for assessing the Warburg level

of the cell lines and using them to predict drug response

The EOR and AFR measures were calculated in a similar manner to

that described above. Specifically, the EOR is calculated as the mean

over lactate secretion across all samples divided by the mean over

oxygen consumption across all samples. Similarly, the AFR is calcu-

lated as the mean flux carried by the reactions producing ATP in

glycolysis versus the mean flux carried by the reaction producing

ATP in OXPHOS. To determine an empiric P-value in the drug

response analysis we randomly shuffled the drug response data

1,000 times, each time examining the resulting Wilcoxon P-value

over the original set of cell lines.

Predicting the effect of reaction knockouts

Each metabolic reaction in each cell line model is perturbed by

constraining its flux to zero. Under each perturbation the minimal

lactate secretion (under maximal growth rate) and the maximal

growth rate is calculated. The set of reactions that eliminate forced

lactate secretion while maintaining a level of cell growth that is

> 10% of the wild-type growth prediction is further tested for the

AFR level. The mean AFR level for each cell line under each of these

perturbations is calculated over 1,000 flux distribution samples as

described above. The final set of predicted reactions includes those

whose knockout reduces the AFR to below 60% of its wild-type level.

Datasets

Growth rate measurements and drug response data were down-

loaded from the NCI website.

Growth rate: http://dtp.nci.nih.gov/docs/misc/common_files/

cell_list.html

Drug response: http://discover.nci.nih.gov/nature2000/naturein

tromain.jsp

Experimentally measuring lactate secretion rates of breastcancer cell lines

Cell Culture

The MCF7, T47D, Hs578T and BT549 breast cancer cell lines were

obtained from the American Type Culture Collection and London

Research Institute Cell Services. Cells were cultured in DMEM/F12

(1:1), with 2 mM L-glutamine and penicillin/streptomycin. Medium

was supplemented with 10% FCS (GIBCO) for the cancer cell lines

and 5% horse serum, 20 ng/ml EGF, 5 lg/ml hydrocortisone,

10 lg/ml insulin, and 100 ng/ml cholera toxin for the non-

malignant cell lines.

Lactate secretion measurements

Cells were cultured under normoxic (20% O2) and hypoxic (0.5%

O2) conditions for 72 h. Cells were starved of glucose and glutamine

for 1 h and full medium was added for 1 h. Lactate secretion was

determined from normoxic and hypoxic cells and normalized to

cell growth (increase in total protein during the 72 h incubation in

normoxia). Lactate concentrations in media incubated with or

without cells were determined using lactate assay kits (BioVision).

Total protein content determined by Sulforhodamine B assay was

used for normalization. Two experiments were performed with

three or four biologically independent replicates (total of seven

replicates).

ª 2014 The Authors Molecular Systems Biology 10: 744 | 2014

Keren Yizhak et al Identifying anti-migratory metabolic drug targets Molecular Systems Biology

9

in those without, inhibits cell proliferation. Accordingly, as PHGDH

is not amplified in the cell line MDA-MB-231 which was examined

in both studies, its suppression is indeed non-lethal. However, we

show that its suppression significantly attenuates cell migration,

suggesting that metabolic enzymes can promote different cancerous

phenotypes in different cancer cells.

Remarkably, analyzing the model-predicted flux rates has

successfully uncovered a fundamental association between the AFR

and cancer migration, even given the relatively small set of cell lines

for which migration was measured. Our analysis has also revealed

other potential associations between individual fluxes and cell

migration (Supplementary Fig S4). However, future studies measur-

ing cellular migration data across a much wider array of cell lines

(of the order for which we already have proliferation data) are

needed to determine the actual significance of these potential leads.

As this study has shown, cellular proliferation and migration have

distinct underlying metabolite correlates; understanding the meta-

bolic correlates that are strongly associated with cell migration may

lead to new anti-metastatic treatment opportunities. It is important

to note, however, that while the inhibition of migration alone might

be a good strategy for avoiding the adverse side effects of cytotoxic

treatment, cell migration is a crucial process also in normal physiol-

ogy, for instance, in immune response and tissue repair (Forster

et al, 1999; Ridley et al, 2003). Therefore, future anti-migratory

drugs may pose different drug selectivity challenges that should be

carefully addressed in the future studies. Irrespectively, they may

result in lesser clonal selection, and as a result, their usage may be

accompanied with lesser rate of emergence of drug-resistant clones.

Materials and Methods

Computational methods

Genome-scale metabolic modeling (GSSM)

A metabolic network consisting of m metabolites and n reactions

can be represented by a stoichiometric matrix S, where the entry Sijrepresents the stoichiometric coefficient of metabolite i in reaction j

(Price et al, 2004). A CBM model imposes mass balance, directional-

ity, and flux capacity constraints on the space of possible fluxes in

the metabolic network’s reactions through a set of linear equations:

Sv ¼ 0 (1)

vmin � v� vmax (2)

where v stands for the flux vector for all of the reactions in the

model (i.e. the flux distribution). The exchange of metabolites with

the environment is represented as a set of exchange (transport)

reactions, enabling a pre-defined set of metabolites to be either

taken up or secreted from the growth media. The steady-state

assumption represented in equation (1) constrains the production

rate of each metabolite to be equal to its consumption rate. Enzy-

matic directionality and flux capacity constraints define lower and

upper bounds on the fluxes and are embedded in equation (2).

In the following, flux vectors satisfying these conditions will

be referred to as feasible steady-state flux distributions. Gene

knockouts are simulated by constraining the flux through the

Figure 5. ECAR and OCR levels of top predicted gene targets.

A Mean and SEM (normalized to nuclear count) ECAR and OCR levels after silencing of seven different genes (HK2, PGAM1, PGK2, GAPDH, PSPH, AHCY, and PHGDH)compared to the control. Silencing of the four glycolytic genes results in both a decrease in ECAR levels (x-axis) and an increase in OCR levels (y-axis), while theserine- and methionine-associated genes show only a decrease in ECAR levels. Error bars represent SEM. The number of samples is n = 18.

B Mean and SD of computed ECAR/OCR (EOR) levels for control and selected gene silencing (Materials and Methods). For all genes a significant reduction in EOR levelsis observed. Error bars represent SD. The number of samples is n = 18.

Molecular Systems Biology 10: 744 | 2014 ª 2014 The Authors

Molecular Systems Biology Identifying anti-migratory metabolic drug targets Keren Yizhak et al

8

computations. SLD, VMR and VCB performed the experimental procedures. KY,

SLD, BvW, and ER wrote the paper.

Conflict of interestThe authors declare that they have no conflict of interest.

References

Agren R, Bordel S, Mardinoglu A, Pornputtapong N, Nookaew I, Nielsen J

(2012) Reconstruction of genome-scale active metabolic networks for 69

human cell types and 16 cancer types using INIT. PLoS Comput Biol 8:

e1002518

Agren R, Mardinoglu A, Asplund A, Kampf C, Uhlen M, Nielsen J (2014)

Identification of anticancer drugs for hepatocellular carcinoma through

personalized genome-scale metabolic modeling. Mol Syst Biol 10: 721

Bordel S, Agren R, Nielsen J (2010) Sampling the solution space in

genome-scale metabolic networks reveals transcriptional regulation in key

enzymes. PLoS Comput Biol 6: e1000859

Burgard AP, Pharkya P, Maranas CD (2003) Optknock: a bilevel programming

framework for identifying gene knockout strategies for microbial strain

optimization. Biotechnol Bioeng 84: 647 – 657

Chandrasekaran S, Price ND (2010) Probabilistic integrative modeling of

genome-scale metabolic and regulatory networks in Escherichia coli and

Mycobacterium tuberculosis. Proc Natl Acad Sci 107: 17845 – 17850

Chang HY, Nuyten DSA, Sneddon JB, Hastie T, Tibshirani R, Sørlie T, Dai H, He

YD, van’t Veer LJ, Bartelink H, van de Rijn M, Brown PO, van de Vijver MJ

(2005) Robustness, scalability, and integration of a wound-response gene

expression signature in predicting breast cancer survival. Proc Natl Acad

Sci U S A 102: 3738 – 3743

Cheong H, Lu C, Lindsten T, Thompson CB (2012) Therapeutic targets in

cancer cell metabolism and autophagy. Nat Biotechnol 30: 671 – 678

Choy E, Yelensky R, Bonakdar S, Plenge RM, Saxena R, De Jager PL, Shaw SY,

Wolfish CS, Slavik JM, Cotsapas C, Rivas M, Dermitzakis ET,

Cahir-McFarland E, Kieff E, Hafler D, Daly MJ, Altshuler D (2008) Genetic

analysis of human traits in vitro: drug response and gene expression in

lymphoblastoid cell lines. PLoS Genet 4: e1000287

Covert MW, Knight EM, Reed JL, Herrgard MJ, Palsson BO (2004) Integrating

high-throughput and computational data elucidates bacterial networks.

Nature 429: 92 – 96

Cuezva JM, Krajewska M, de Heredia ML, Krajewski S, Santamaría G, Kim H,

Zapata JM, Marusawa H, Chamorro M, Reed JC (2002) The bioenergetic

signature of cancer: a marker of tumor progression. Cancer Res 62:

6674 – 6681

Curtis C, Shah SP, Chin S-F, Turashvili G, Rueda OM, Dunning MJ, Speed D,

Lynch AG, Samarajiwa S, Yuan Y, Graf S, Ha G, Haffari G, Bashashati A,

Russell R, McKinney S, Langerod A, Green A, Provenzano E, Wishart G et al

(2012) The genomic and transcriptomic architecture of 2,000 breast

tumours reveals novel subgroups. Nature 486: 346 – 352

De Bock K, Georgiadou M, Schoors S, Kuchnio A, Wong Brian W, Cantelmo

Anna R, Quaegebeur A, Ghesquière B, Cauwenberghs S, Eelen G, Phng L-K,

Betz I, Tembuyser B, Brepoels K, Welti J, Geudens I, Segura I, Cruys B,

Bifari F, Decimo I et al (2013) Role of PFKFB3-driven glycolysis in vessel

sprouting. Cell 154: 651 – 663

Duarte NC, Becker SA, Jamshidi N, Thiele I, Mo ML, Vo TD, Srivas R,

Palsson BO (2007) Global reconstruction of the human metabolic

network based on genomic and bibliomic data. Proc Natl Acad Sci

U S A 104: 1777 – 1782

Ezratty EJ, Partridge MA, Gundersen GG (2005) Microtubule-induced focal

adhesion disassembly is mediated by dynamin and focal adhesion kinase.

Nat Cell Biol 7: 581 – 590

Folger O, Jerby L, Frezza C, Gottlieb E, Ruppin E, Shlomi T (2011) Predicting

selective drug targets in cancer through metabolic networks. Mol Syst Biol

7: 501

Förster R, Schubel A, Breitfeld D, Kremmer E, Renner-Müller I, Wolf E, Lipp M

(1999) CCR7 coordinates the primary immune response by establishing

functional microenvironments in secondary lymphoid organs. Cell 99:

23 – 33

Frezza C, Zheng L, Folger O, Rajagopalan KN, MacKenzie ED, Jerby L, Micaroni

M, Chaneton B, Adam J, Hedley A, Kalna G, Tomlinson IPM, Pollard PJ,

Watson DG, Deberardinis RJ, Shlomi T, Ruppin E, Gottlieb E (2011) Haem

oxygenase is synthetically lethal with the tumour suppressor fumarate

hydratase. Nature 477: 225 – 228

Gatto F, Nookaew I, Nielsen J (2014) Chromosome 3p loss of heterozygosity is

associated with a unique metabolic network in clear cell renal carcinoma.

Proc Natl Acad Sci 111: E866 – E875

Gille C, Bolling C, Hoppe A, Bulik S, Hoffmann S, Hubner K, Karlstadt A,

Ganeshan R, Konig M, Rother K, Weidlich M, Behre J, Holzhutter H-G

(2010) HepatoNet1: a comprehensive metabolic reconstruction of the

human hepatocyte for the analysis of liver physiology. Mol Syst Biol 6: 411

Goldstein I, Yizhak K, Madar S, Goldfinger N, Ruppin E, Rotter V (2013) p53

promotes the expression of gluconeogenesis-related genes and enhances

hepatic glucose production. Cancer Metab 1: 9

Hanahan D, Weinberg RA (2011) Hallmarks of cancer: the next generation.

Cell 144: 646 – 674

Hirschhaeuser F, Sattler UGA, Mueller-Klieser W (2011) Lactate: a metabolic

key player in cancer. Cancer Res 71: 6921 – 6925

Jain M, Nilsson R, Sharma S, Madhusudhan N, Kitami T, Souza AL, Kafri R,

Kirschner MW, Clish CB, Mootha VK (2012) Metabolite profiling identifies a

key role for glycine in rapid cancer cell proliferation. Science 336:

1040 – 1044

Jensen PA, Papin JA (2010) Functional integration of a metabolic network

model and expression data without arbitrary thresholding. Bioinformatics

27: 541 – 547

Jerby L, Wolf L, Denkert C, Stein GY, Hilvo M, Oresic M, Geiger T, Ruppin E

(2012) Metabolic associations of reduced proliferation and oxidative stress

in advanced breast cancer. Cancer Res 72: 5712 – 5720

Lee JK, Havaleshko DM, Cho H, Weinstein JN, Kaldjian EP, Karpovich J,

Grimshaw A, Theodorescu D (2007) A strategy for predicting the

chemosensitivity of human cancers and its application to drug discovery.

Proc Natl Acad Sci 104: 13086 – 13091

Lerman JA, Hyduke DR, Latif H, Portnoy VA, Lewis NE, Orth JD,

Schrimpe-Rutledge AC, Smith RD, Adkins JN, Zengler K, Palsson BO (2012)

In silico method for modelling metabolism and gene product expression

at genome scale. Nat Commun 3: 929

Lewis NE, Schramm G, Bordbar A, Schellenberger J, Andersen MP, Cheng JK,

Patel N, Yee A, Lewis RA, Eils R, Konig R, Palsson BO (2010) Large-scale in

silico modeling of metabolic interactions between cell types in the human

brain. Nat Biotechnol 28: 1279 – 1285

Ma H, Sorokin A, Mazein A, Selkov A, Selkov E, Demin O, Goryanin I (2007)

The Edinburgh human metabolic network reconstruction and its

functional analysis. Mol Syst Biol 3: 135

Mardinoglu A, Agren R, Kampf C, Asplund A, Nookaew I, Jacobson P, Walley

AJ, Froguel P, Carlsson LM, Uhlen M, Nielsen J (2013) Integration of clinical

data with a genome-scale metabolic model of the human adipocyte. Mol

Syst Biol 9: 649

ª 2014 The Authors Molecular Systems Biology 10: 744 | 2014

Keren Yizhak et al Identifying anti-migratory metabolic drug targets Molecular Systems Biology

11

Cell culture for live cell imaging and cell migration assays

T47D, MCF-7, MDA-MB-435, BT549, MDA-MB-231 and Hs578t were

cultured in RPMI (GIBCO, Life Technologies, Carlsbad, CA, USA)

supplemented with 10% FBS (PAA, Pashing Austria) and 100

International Units/ml penicillin and 100 lg/ml streptomycin

(Invitrogen, Carlsbad, CA, USA).

Gene silencing

Human siRNA SmartPools (a combination of four individual singles)

for the 17 predicted genes were purchased in siGENOME format

from Dharmacon (Lafayette, CO, USA). Plates were diluted to 1 lMworking concentration in complementary 1× siRNA buffer in a

96-well plate format. A non-targeting siRNA was used as negative

control. A 50 nM reverse transfection was performed according to

manufacturer’s guidelines. Complex time was 20 min and 5,000

cells were added. The plate was placed in the incubator overnight

and the medium was refreshed the following morning. After

48–72 h cells were used for various assays. Cell migration and meta-

bolic flux assay experiments were performed in duplicate while the

cell proliferation assay was performed in triplicate.

Live cell imaging random cell migration assay

Glass bottom 96-well plates (Greiner Bio-one, Monroe, NC, USA)

were coated with 20 lg/ll collagen type I (isolated from rat tails)

for 1 h at 37°C. 48 h after silencing, the MDA-MB-231 cells were re-

plated onto the collagen-coated glass bottom plate. 24 h after seed-

ing, cells were pre-exposed for 45 min to 0.1 lg/µl Hoechst 33342(Fisher Scientific, Hampton, NH, USA) to visualize nuclei. After

refreshing the medium, cells were placed on a Nikon Eclipse

TE2000-E microscope fitted with a 37°C incubation chamber, 20×

objective (0.75 NA, 1.00 WD) automated stage and perfect focus

system. Three positions per well were automatically defined, and

the Differential Interference Contrast (DIC) and Hoechst signals

were acquired with a CCD camera (Pixel size: 0.64 lm) every

20 min for a total imaging period of 12 h using NIS software

(Nikon). All data were converted and analyzed using custom-made

ImagePro Plus macros (Roosmalen et al, 2011). Cell migration was

quantified by tracking nuclei in time. Changes in migration speed

per knockdown were evaluated via a two-sided t-test comparing the

speed for every individual cell followed overtime for 16 h and the

corresponding control values. Data shown are normalized to control

and represent only one replicate. Of note, for all four cell lines both

replicates showed a R2 of reproducibility above 0.75. Genes achiev-

ing P-value < 0.05 after correcting for multiple hypothesis using

FDR with a = 0.05 are considered as hits.

Proliferation assay

Cells were directly transfected and plated onto micro-clear 96-well

plates (Greiner Bio-one). After 5 days of incubation, the cells were

stained with Hoechst 33342 and fixed with TCA (Trichloroacetic

acid) allowing both a nuclear counting and/or Sulforodamine B

(SRB) readout. Whole wells were imaged using epi-fluorescence

and the number of nuclei was determined using a custom-made

ImagePro macro. Plates were further processed for SRB staining as

described earlier (Zhang et al, 2011). SRB data showed a complete

overlap with the nuclear count so this measure is used in all

figures. Changes in proliferation rates upon knockdown when

compared to control were evaluated in triplicate via a two-sided

t-test. The mean proliferation rate after knockdown between all

three replicates was calculated and normalized to the non-targeting

siRNA (= control). Genes achieving P-value < 0.05 after correcting

for multiple hypothesis using FDR with a = 0.05 are considered as

hits.

Metabolic flux assay

The bioenergetics flux of cells in response to gene silencing was

assessed using the Seahorse XF96 extracellular flux analyzer

(Seahorse Bioscience). About 8,000 MDA-MB-231 cells per well

(Seahorse plate) were treated with siRNAs or control for 72 h. Each

gene (in total 7) was knockdown in six different wells and the

experiment was performed twice (so a total of six replicates per

plate and two plates). Prior to measurement, the medium was

replaced with unbuffered DMEM XF assay medium. The basal

oxygen consumption rate (OCR) and extracellular acidification rate

(ECAR) were then determined using the XP96 plate reader with the

standard program as recommended by the manufacturer: three

measurements per well were done (so for each gene 18 measure-

ments were obtained for both OCR and ECAR). After the measure-

ments were completed, the plates were live stained with Hoechst

33342 for 1 h and fixed with TCA allowing both a nuclear counting

and/or SRB readout. Whole wells were imaged using epi-fluores-

cence and the number of nuclei was determined using a custom-

made ImagePro macro. Plates were further processed for SRB stain-

ing as described earlier (Zhang et al, 2011). SRB data showed a

complete overlap with the nuclear count so this measure was used

for normalization. All values are normalized to nuclear count. EOR

for control and each gene knockdown is computed by dividing the

corresponding ECAR and OCR values. A two-sided t-test is applied

to examine significant changes between control and knockdown-

induced EOR.

Supplementary information for this article is available online:

http://msb.embopress.org

AcknowledgementsWe would like to thank Hans de Bont and Michiel Fokkelman for their technical

support, Yoav Teboulle, Matthew Oberhardt, Edoardo Gaude, Gideon Y. Stein

and Tami Geiger for their helpful comments on the manuscript. KY is partially

supported by a fellowship from the Edmond J. Safra Bioinformatics center at

Tel-Aviv University and is grateful to the Azrieli Foundation for the award of an

Azrieli Fellowship; SLD is supported by the Netherlands Consortium for Systems

Biology and the EU FP7 Systems Microscopy NoE project (258068) and BvdW

from the Netherlands Genomics Initiative. ER acknowledges the generous

support of grants from the Israeli Science Foundation (ISF), the Israeli Cancer

Research Fund (ICRF) and the I-CORE Program of the Planning and Budgeting

Committee and The Israel Science Foundation (grant No 41/11).

Author contributionsKY and ER conceived and designed the research. SLD, VCB, CF, and BvW

designed the experimental procedures. FB and AS contributed the lactate

secretion data. KY performed the computational analysis and the statistical

Molecular Systems Biology 10: 744 | 2014 ª 2014 The Authors

Molecular Systems Biology Identifying anti-migratory metabolic drug targets Keren Yizhak et al

10

computations. SLD, VMR and VCB performed the experimental procedures. KY,

SLD, BvW, and ER wrote the paper.

Conflict of interestThe authors declare that they have no conflict of interest.

References

Agren R, Bordel S, Mardinoglu A, Pornputtapong N, Nookaew I, Nielsen J

(2012) Reconstruction of genome-scale active metabolic networks for 69

human cell types and 16 cancer types using INIT. PLoS Comput Biol 8:

e1002518

Agren R, Mardinoglu A, Asplund A, Kampf C, Uhlen M, Nielsen J (2014)

Identification of anticancer drugs for hepatocellular carcinoma through

personalized genome-scale metabolic modeling. Mol Syst Biol 10: 721

Bordel S, Agren R, Nielsen J (2010) Sampling the solution space in

genome-scale metabolic networks reveals transcriptional regulation in key

enzymes. PLoS Comput Biol 6: e1000859

Burgard AP, Pharkya P, Maranas CD (2003) Optknock: a bilevel programming

framework for identifying gene knockout strategies for microbial strain

optimization. Biotechnol Bioeng 84: 647 – 657

Chandrasekaran S, Price ND (2010) Probabilistic integrative modeling of

genome-scale metabolic and regulatory networks in Escherichia coli and

Mycobacterium tuberculosis. Proc Natl Acad Sci 107: 17845 – 17850

Chang HY, Nuyten DSA, Sneddon JB, Hastie T, Tibshirani R, Sørlie T, Dai H, He

YD, van’t Veer LJ, Bartelink H, van de Rijn M, Brown PO, van de Vijver MJ

(2005) Robustness, scalability, and integration of a wound-response gene

expression signature in predicting breast cancer survival. Proc Natl Acad

Sci U S A 102: 3738 – 3743

Cheong H, Lu C, Lindsten T, Thompson CB (2012) Therapeutic targets in

cancer cell metabolism and autophagy. Nat Biotechnol 30: 671 – 678

Choy E, Yelensky R, Bonakdar S, Plenge RM, Saxena R, De Jager PL, Shaw SY,

Wolfish CS, Slavik JM, Cotsapas C, Rivas M, Dermitzakis ET,

Cahir-McFarland E, Kieff E, Hafler D, Daly MJ, Altshuler D (2008) Genetic

analysis of human traits in vitro: drug response and gene expression in

lymphoblastoid cell lines. PLoS Genet 4: e1000287

Covert MW, Knight EM, Reed JL, Herrgard MJ, Palsson BO (2004) Integrating

high-throughput and computational data elucidates bacterial networks.

Nature 429: 92 – 96

Cuezva JM, Krajewska M, de Heredia ML, Krajewski S, Santamaría G, Kim H,

Zapata JM, Marusawa H, Chamorro M, Reed JC (2002) The bioenergetic

signature of cancer: a marker of tumor progression. Cancer Res 62:

6674 – 6681

Curtis C, Shah SP, Chin S-F, Turashvili G, Rueda OM, Dunning MJ, Speed D,

Lynch AG, Samarajiwa S, Yuan Y, Graf S, Ha G, Haffari G, Bashashati A,

Russell R, McKinney S, Langerod A, Green A, Provenzano E, Wishart G et al

(2012) The genomic and transcriptomic architecture of 2,000 breast

tumours reveals novel subgroups. Nature 486: 346 – 352

De Bock K, Georgiadou M, Schoors S, Kuchnio A, Wong Brian W, Cantelmo

Anna R, Quaegebeur A, Ghesquière B, Cauwenberghs S, Eelen G, Phng L-K,

Betz I, Tembuyser B, Brepoels K, Welti J, Geudens I, Segura I, Cruys B,

Bifari F, Decimo I et al (2013) Role of PFKFB3-driven glycolysis in vessel

sprouting. Cell 154: 651 – 663

Duarte NC, Becker SA, Jamshidi N, Thiele I, Mo ML, Vo TD, Srivas R,

Palsson BO (2007) Global reconstruction of the human metabolic

network based on genomic and bibliomic data. Proc Natl Acad Sci

U S A 104: 1777 – 1782

Ezratty EJ, Partridge MA, Gundersen GG (2005) Microtubule-induced focal

adhesion disassembly is mediated by dynamin and focal adhesion kinase.

Nat Cell Biol 7: 581 – 590

Folger O, Jerby L, Frezza C, Gottlieb E, Ruppin E, Shlomi T (2011) Predicting

selective drug targets in cancer through metabolic networks. Mol Syst Biol

7: 501

Förster R, Schubel A, Breitfeld D, Kremmer E, Renner-Müller I, Wolf E, Lipp M

(1999) CCR7 coordinates the primary immune response by establishing

functional microenvironments in secondary lymphoid organs. Cell 99:

23 – 33

Frezza C, Zheng L, Folger O, Rajagopalan KN, MacKenzie ED, Jerby L, Micaroni

M, Chaneton B, Adam J, Hedley A, Kalna G, Tomlinson IPM, Pollard PJ,

Watson DG, Deberardinis RJ, Shlomi T, Ruppin E, Gottlieb E (2011) Haem

oxygenase is synthetically lethal with the tumour suppressor fumarate

hydratase. Nature 477: 225 – 228

Gatto F, Nookaew I, Nielsen J (2014) Chromosome 3p loss of heterozygosity is

associated with a unique metabolic network in clear cell renal carcinoma.

Proc Natl Acad Sci 111: E866 – E875

Gille C, Bolling C, Hoppe A, Bulik S, Hoffmann S, Hubner K, Karlstadt A,

Ganeshan R, Konig M, Rother K, Weidlich M, Behre J, Holzhutter H-G

(2010) HepatoNet1: a comprehensive metabolic reconstruction of the

human hepatocyte for the analysis of liver physiology. Mol Syst Biol 6: 411

Goldstein I, Yizhak K, Madar S, Goldfinger N, Ruppin E, Rotter V (2013) p53

promotes the expression of gluconeogenesis-related genes and enhances

hepatic glucose production. Cancer Metab 1: 9

Hanahan D, Weinberg RA (2011) Hallmarks of cancer: the next generation.

Cell 144: 646 – 674

Hirschhaeuser F, Sattler UGA, Mueller-Klieser W (2011) Lactate: a metabolic

key player in cancer. Cancer Res 71: 6921 – 6925

Jain M, Nilsson R, Sharma S, Madhusudhan N, Kitami T, Souza AL, Kafri R,

Kirschner MW, Clish CB, Mootha VK (2012) Metabolite profiling identifies a

key role for glycine in rapid cancer cell proliferation. Science 336:

1040 – 1044

Jensen PA, Papin JA (2010) Functional integration of a metabolic network

model and expression data without arbitrary thresholding. Bioinformatics

27: 541 – 547

Jerby L, Wolf L, Denkert C, Stein GY, Hilvo M, Oresic M, Geiger T, Ruppin E

(2012) Metabolic associations of reduced proliferation and oxidative stress

in advanced breast cancer. Cancer Res 72: 5712 – 5720

Lee JK, Havaleshko DM, Cho H, Weinstein JN, Kaldjian EP, Karpovich J,

Grimshaw A, Theodorescu D (2007) A strategy for predicting the

chemosensitivity of human cancers and its application to drug discovery.

Proc Natl Acad Sci 104: 13086 – 13091

Lerman JA, Hyduke DR, Latif H, Portnoy VA, Lewis NE, Orth JD,

Schrimpe-Rutledge AC, Smith RD, Adkins JN, Zengler K, Palsson BO (2012)

In silico method for modelling metabolism and gene product expression

at genome scale. Nat Commun 3: 929

Lewis NE, Schramm G, Bordbar A, Schellenberger J, Andersen MP, Cheng JK,

Patel N, Yee A, Lewis RA, Eils R, Konig R, Palsson BO (2010) Large-scale in

silico modeling of metabolic interactions between cell types in the human

brain. Nat Biotechnol 28: 1279 – 1285

Ma H, Sorokin A, Mazein A, Selkov A, Selkov E, Demin O, Goryanin I (2007)

The Edinburgh human metabolic network reconstruction and its

functional analysis. Mol Syst Biol 3: 135

Mardinoglu A, Agren R, Kampf C, Asplund A, Nookaew I, Jacobson P, Walley

AJ, Froguel P, Carlsson LM, Uhlen M, Nielsen J (2013) Integration of clinical

data with a genome-scale metabolic model of the human adipocyte. Mol

Syst Biol 9: 649

ª 2014 The Authors Molecular Systems Biology 10: 744 | 2014

Keren Yizhak et al Identifying anti-migratory metabolic drug targets Molecular Systems Biology

11

Cell culture for live cell imaging and cell migration assays

T47D, MCF-7, MDA-MB-435, BT549, MDA-MB-231 and Hs578t were

cultured in RPMI (GIBCO, Life Technologies, Carlsbad, CA, USA)

supplemented with 10% FBS (PAA, Pashing Austria) and 100

International Units/ml penicillin and 100 lg/ml streptomycin

(Invitrogen, Carlsbad, CA, USA).

Gene silencing

Human siRNA SmartPools (a combination of four individual singles)

for the 17 predicted genes were purchased in siGENOME format

from Dharmacon (Lafayette, CO, USA). Plates were diluted to 1 lMworking concentration in complementary 1× siRNA buffer in a

96-well plate format. A non-targeting siRNA was used as negative

control. A 50 nM reverse transfection was performed according to

manufacturer’s guidelines. Complex time was 20 min and 5,000

cells were added. The plate was placed in the incubator overnight

and the medium was refreshed the following morning. After

48–72 h cells were used for various assays. Cell migration and meta-

bolic flux assay experiments were performed in duplicate while the

cell proliferation assay was performed in triplicate.

Live cell imaging random cell migration assay

Glass bottom 96-well plates (Greiner Bio-one, Monroe, NC, USA)

were coated with 20 lg/ll collagen type I (isolated from rat tails)

for 1 h at 37°C. 48 h after silencing, the MDA-MB-231 cells were re-

plated onto the collagen-coated glass bottom plate. 24 h after seed-

ing, cells were pre-exposed for 45 min to 0.1 lg/µl Hoechst 33342(Fisher Scientific, Hampton, NH, USA) to visualize nuclei. After

refreshing the medium, cells were placed on a Nikon Eclipse

TE2000-E microscope fitted with a 37°C incubation chamber, 20×

objective (0.75 NA, 1.00 WD) automated stage and perfect focus

system. Three positions per well were automatically defined, and

the Differential Interference Contrast (DIC) and Hoechst signals

were acquired with a CCD camera (Pixel size: 0.64 lm) every

20 min for a total imaging period of 12 h using NIS software

(Nikon). All data were converted and analyzed using custom-made

ImagePro Plus macros (Roosmalen et al, 2011). Cell migration was

quantified by tracking nuclei in time. Changes in migration speed

per knockdown were evaluated via a two-sided t-test comparing the

speed for every individual cell followed overtime for 16 h and the

corresponding control values. Data shown are normalized to control

and represent only one replicate. Of note, for all four cell lines both

replicates showed a R2 of reproducibility above 0.75. Genes achiev-

ing P-value < 0.05 after correcting for multiple hypothesis using

FDR with a = 0.05 are considered as hits.

Proliferation assay

Cells were directly transfected and plated onto micro-clear 96-well

plates (Greiner Bio-one). After 5 days of incubation, the cells were

stained with Hoechst 33342 and fixed with TCA (Trichloroacetic

acid) allowing both a nuclear counting and/or Sulforodamine B

(SRB) readout. Whole wells were imaged using epi-fluorescence

and the number of nuclei was determined using a custom-made

ImagePro macro. Plates were further processed for SRB staining as

described earlier (Zhang et al, 2011). SRB data showed a complete

overlap with the nuclear count so this measure is used in all

figures. Changes in proliferation rates upon knockdown when

compared to control were evaluated in triplicate via a two-sided

t-test. The mean proliferation rate after knockdown between all

three replicates was calculated and normalized to the non-targeting

siRNA (= control). Genes achieving P-value < 0.05 after correcting

for multiple hypothesis using FDR with a = 0.05 are considered as

hits.

Metabolic flux assay

The bioenergetics flux of cells in response to gene silencing was

assessed using the Seahorse XF96 extracellular flux analyzer

(Seahorse Bioscience). About 8,000 MDA-MB-231 cells per well

(Seahorse plate) were treated with siRNAs or control for 72 h. Each

gene (in total 7) was knockdown in six different wells and the

experiment was performed twice (so a total of six replicates per

plate and two plates). Prior to measurement, the medium was

replaced with unbuffered DMEM XF assay medium. The basal

oxygen consumption rate (OCR) and extracellular acidification rate

(ECAR) were then determined using the XP96 plate reader with the

standard program as recommended by the manufacturer: three

measurements per well were done (so for each gene 18 measure-

ments were obtained for both OCR and ECAR). After the measure-

ments were completed, the plates were live stained with Hoechst

33342 for 1 h and fixed with TCA allowing both a nuclear counting

and/or SRB readout. Whole wells were imaged using epi-fluores-

cence and the number of nuclei was determined using a custom-

made ImagePro macro. Plates were further processed for SRB stain-

ing as described earlier (Zhang et al, 2011). SRB data showed a

complete overlap with the nuclear count so this measure was used

for normalization. All values are normalized to nuclear count. EOR

for control and each gene knockdown is computed by dividing the

corresponding ECAR and OCR values. A two-sided t-test is applied

to examine significant changes between control and knockdown-

induced EOR.

Supplementary information for this article is available online:

http://msb.embopress.org

AcknowledgementsWe would like to thank Hans de Bont and Michiel Fokkelman for their technical

support, Yoav Teboulle, Matthew Oberhardt, Edoardo Gaude, Gideon Y. Stein

and Tami Geiger for their helpful comments on the manuscript. KY is partially

supported by a fellowship from the Edmond J. Safra Bioinformatics center at

Tel-Aviv University and is grateful to the Azrieli Foundation for the award of an

Azrieli Fellowship; SLD is supported by the Netherlands Consortium for Systems

Biology and the EU FP7 Systems Microscopy NoE project (258068) and BvdW

from the Netherlands Genomics Initiative. ER acknowledges the generous

support of grants from the Israeli Science Foundation (ISF), the Israeli Cancer

Research Fund (ICRF) and the I-CORE Program of the Planning and Budgeting

Committee and The Israel Science Foundation (grant No 41/11).

Author contributionsKY and ER conceived and designed the research. SLD, VCB, CF, and BvW

designed the experimental procedures. FB and AS contributed the lactate

secretion data. KY performed the computational analysis and the statistical

Molecular Systems Biology 10: 744 | 2014 ª 2014 The Authors

Molecular Systems Biology Identifying anti-migratory metabolic drug targets Keren Yizhak et al

10

Miller LD, Smeds J, George J, Vega VB, Vergara L, Ploner A, Pawitan Y, Hall P,

Klaar S, Liu ET, Bergh J (2005) An expression signature for p53 status in

human breast cancer predicts mutation status, transcriptional effects, and

patient survival. Proc Natl Acad Sci U S A 102: 13550 – 13555

Nogales J, Gudmundsson S, Knight EM, Palsson BO, Thiele I (2012) Detailing

the optimality of photosynthesis in cyanobacteria through systems

biology analysis. Proc Natl Acad Sci 109: 2678 – 2683

Oberhardt MA, Palsson BO, Papin JA (2009) Applications of genome-scale

metabolic reconstructions. Mol Syst Biol 5: 320

Oberhardt MA, Yizhak K, Ruppin E (2013) Metabolically re-modeling the drug

pipeline. Curr Opin Pharmacol 13: 778 – 785

Penault-Llorca F, André F, Sagan C, Lacroix-Triki M, Denoux Y, Verriele V,

Jacquemier J, Baranzelli MC, Bibeau F, Antoine M, Lagarde N, Martin A-L,

Asselain B, Roché H (2009) Ki67 expression and docetaxel efficacy in

patients with estrogen receptor-positive breast cancer. J Clin Oncol 27:

2809 – 2815

Pharkya P, Burgard AP, Maranas CD (2004) OptStrain: a computational

framework for redesign of microbial production systems. Genome Res 14:

2367 – 2376

Possemato R, Marks KM, Shaul YD, Pacold ME, Kim D, Birsoy K,

Sethumadhavan S, Woo H-K, Jang HG, Jha AK, Chen WW, Barrett FG,

Stransky N, Tsun Z-Y, Cowley GS, Barretina J, Kalaany NY, Hsu PP, Ottina

K, Chan AM et al (2011) Functional genomics reveal that the serine

synthesis pathway is essential in breast cancer. Nature 476: 346 – 350

Price ND, Reed JL, Palsson BO (2004) Genome-scale models of microbial cells:

evaluating the consequences of constraints. Nat Rev Microbiol 2: 886 – 897

Ridley AJ, Schwartz MA, Burridge K, Firtel RA, Ginsberg MH, Borisy G, Parsons

JT, Horwitz AR (2003) Cell migration: integrating signals from front to

back. Science 302: 1704 – 1709

van Roosmalen W, Le Dévédec SE, Zovko S, de Bont H, van de Water B (2011)

Functional screening with a live cell imaging-based random cell migration

assay. Methods Mol Biol 769: 435 – 448

Sarrio D, Rodriguez-Pinilla SM, Hardisson D, Cano A, Moreno-Bueno G,

Palacios J (2008) Epithelial-mesenchymal transition in breast cancer

relates to the basal-like phenotype. Cancer Res 68: 989 – 997

Scherf U, Ross DT, Waltham M, Smith LH, Lee JK, Tanabe L, Kohn KW, Reinhold

WC, Myers TG, Andrews DT, Scudiero DA, Eisen MB, Sausville EA, Pommier Y,

Botstein D, Brown PO, Weinstein JN (2000) A gene expression database for

the molecular pharmacology of cancer. Nat Genet 24: 236 – 244

Schuetz R, Zamboni N, Zampieri M, Heinemann M, Sauer U (2012)

Multidimensional optimality of microbial metabolism. Science 336:

601 – 604

Shlomi T, Benyamini T, Gottlieb E, Sharan R, Ruppin E (2011) Genome-scale

metabolic modeling elucidates the role of proliferative adaptation in

causing the Warburg effect. PLoS Comput Biol 7: e1002018

Shlomi T, Cabili MN, Herrgård MJ, Palsson BØ, Ruppin E (2008)

Network-based prediction of human tissue-specific metabolism. Nat

Biotechnol 26: 1003 – 1010

Simpson KJ, Selfors LM, Bui J, Reynolds A, Leake D, Khvorova A, Brugge JS

(2008) Identification of genes that regulate epithelial cell migration using

an siRNA screening approach. Nat Cell Biol 10: 1027 – 1038

Szappanos B, Kovacs K, Szamecz B, Honti F, Costanzo M, Baryshnikova A,

Gelius-Dietrich G, Lercher MJ, Jelasity M, Myers CL, Andrews BJ, Boone C,

Oliver SG, Pal C, Papp B (2011) An integrated approach to characterize

genetic interaction networks in yeast metabolism. Nat Genet 43:

656 – 662

Trawick JD, Schilling CH (2006) Use of constraint-based modeling for the

prediction and validation of antimicrobial targets. Biochem Pharmacol 71:

1026 – 1035

Vander Heiden MG, Cantley LC, Thompson CB (2009) Understanding the

Warburg effect: the metabolic requirements of cell proliferation. Science

324: 1029 – 1033

Vincent-Salomon A, Rousseau A, Jouve M, Beuzeboc P, Sigal-Zafrani B,

Fréneaux P, Rosty C, Nos C, Campana F, Klijanienko J, Al Ghuzlan A,

Sastre-Garau X (2004) Proliferation markers predictive of the pathological

response and disease outcome of patients with breast carcinomas treated

by anthracycline-based preoperative chemotherapy. Eur J Cancer 40:

1502 – 1508

Warburg O (1956) On the origin of cancer cells. Science 123: 309 – 314

Wessely F, Bartl M, Guthke R, Li P, Schuster S, Kaleta C (2011) Optimal

regulatory strategies for metabolic pathways in Escherichia coli depending

on protein costs. Mol Syst Biol 7: 515

Wu M, Neilson A, Swift AL, Moran R, Tamagnine J, Parslow D, Armistead S,

Lemire K, Orrell J, Teich J, Chomicz S, Ferrick DA (2007) Multiparameter

metabolic analysis reveals a close link between attenuated mitochondrial

bioenergetic function and enhanced glycolysis dependency in human

tumor cells. Am J Physiol Cell Physiol 292: C125 –C136

Yizhak K, Gabay O, Cohen H, Ruppin E (2013) Model-based identification of

drug targets that revert disrupted metabolism and its application to

aging. Nat Commun 4: 2632

Zhang Y, Moerkens M, Ramaiahgari S, de Bont H, Price L, Meerman J, van de

Water B (2011) Elevated insulin-like growth factor 1 receptor signaling

induces antiestrogen resistance through the MAPK/ERK and PI3K/Akt

signaling routes. Breast Cancer Res 13: R52

License: This is an open access article under the

terms of the Creative Commons Attribution 4.0

License, which permits use, distribution and reproduc-

tion in any medium, provided the original work is

properly cited.

Molecular Systems Biology 10: 744 | 2014 ª 2014 The Authors

Molecular Systems Biology Identifying anti-migratory metabolic drug targets Keren Yizhak et al

12

Review

Principles of targeting endothelial cell metabolismto treat angiogenesis and endothelial celldysfunction in diseaseJermaine Goveia1,2, Peter Stapor1,2 & Peter Carmeliet1,2,*

Abstract

The endothelium is the orchestral conductor of blood vessel func-tion. Pathological blood vessel formation (a process termed patho-logical angiogenesis) or the inability of endothelial cells (ECs) toperform their physiological function (a condition known as ECdysfunction) are defining features of various diseases. Therapeuticintervention to inhibit aberrant angiogenesis or ameliorate ECdysfunction could be beneficial in diseases such as cancer andcardiovascular disease, respectively, but current strategies havelimited efficacy. Based on recent findings that pathological angio-genesis and EC dysfunction are accompanied by EC-specific meta-bolic alterations, targeting EC metabolism is emerging as a noveltherapeutic strategy. Here, we review recent progress in ourunderstanding of how EC metabolism is altered in disease anddiscuss potential metabolic targets and strategies to reverse ECdysfunction and inhibit pathological angiogenesis.

Keywords angiogenesis; endothelial cell dysfunction; metabolism

DOI 10.15252/emmm.201404156 | Received 8 April 2014 | Revised 14 June

2014 | Accepted 3 July 2014 | Published online 25 July 2014

EMBO Mol Med (2014) 6: 1105–1120

See also Glossary for abbreviations used in this article.

Introduction

Blood vessels perform many functions that are critical for tissue

homeostasis (Carmeliet, 2003). The endothelium, a single layer of

endothelial cells (ECs) that lines the blood vessel lumen, controls

vessel function. EC functions include the regulation of vascular tone

and barrier, leukocyte trafficking, blood coagulation, nutrient and

electrolyte uptake and neovascularization of hypoxic tissue, to name

only a few (Cines et al, 1998; Pober et al, 2009; Potente et al, 2011).

Many diseases are characterized by pathological blood vessel

responses or formation. The inability of ECs to perform their physio-

logical function (a condition termed EC dysfunction) contributes to

cardiovascular disease and diabetes (Davignon & Ganz, 2004),

whereas diseases such as cancer and age-related macula degenera-

tion are characterized by new blood vessel formation (a process

termed angiogenesis) (Carmeliet & Jain, 2011). Targeting ECs to

prevent dysfunction or inhibit pathological angiogenesis is poten-

tially beneficial for a wide variety of diseases, but current treatment

modalities, focusing primarily on growth factors, receptors, signal-

ing molecules and others have limited efficacy or specificity (Bergers

& Hanahan, 2008; Versari et al, 2009; Lee et al, 2012).

An emerging but understudied therapeutic target is EC metabo-

lism. It has been long known that risk factors for cardiovascular

disease (hypercholesterolemia, hypertension, dyslipidemia, diabe-

tes, obesity and aging) cause EC-specific metabolic perturbations

leading to EC dysfunction (Davignon & Ganz, 2004; Pober et al,

2009). Similarly, the links between EC metabolism and angiogene-

sis are apparent as angiogenic ECs migrate and proliferate in

metabolically challenging environments such as hypoxic and

nutrient-deprived tissue (Harjes et al, 2012). Moreover, the

growth factor-induced switch from a quiescent to an angiogenic

phenotype is mediated by important adaptations in EC energy

metabolism (De Bock et al, 2013a,b; Schoors et al, 2014a,b). EC

metabolic alterations are therefore not just innocent bystanders

but mediate pathogenesis. In this review, we summarize existing

data on the role of EC metabolism in mediating vascular disease

and discuss how metabolism may be targeted for therapeutic

benefit.

General endothelial metabolism

Despite their close proximity to oxygenated blood, ECs rely on

glycolysis instead of oxidative metabolism for adenosine triphos-

phate (ATP) production (Parra-Bonilla et al, 2010; De Bock et al,

2013b). In fact, under physiological conditions, over 80% of ATP is

produced by converting glucose into lactate (Fig 1). Less than 1% of

glucose-derived pyruvate enters the mitochondria for oxidative

metabolism through the tricarboxylic acid cycle (TCA) and subse-

quent ATP production via the electron transport chain (ETC) (Fig 1)

(Culic et al, 1997; De Bock et al, 2013b). However, ECs retain the

ability to switch to oxidative metabolism of glucose, amino acids

1 Laboratory of Angiogenesis and Neurovascular Link, Vesalius Research Center, Department of Oncology, University of Leuven, Leuven, Belgium2 Laboratory of Angiogenesis and Neurovascular Link, Vesalius Research Center, VIB, Leuven, Belgium

*Corresponding author. Tel: +32 16 37 32 02; Fax: +32 16 37 25 85; E-mail: peter.carmeliet@vib-kuleuven.be

ª 2014 The Authors. Published under the terms of the CC BY 4.0 license EMBO Molecular Medicine Vol 6 | No 9 | 2014 1105

ETC

MethylglyoxalAGEsMETHYLGLYOXAL PATHWAY

Glucose

GLUT1Glucose

G6P

F6P

F1,6P2

G3PDHAP

F2,6P2

PFKFB3PFK

TKT R5P

RPIPENTOSEPHOSPHATEPATHWAY

NUCLEOTIDESYNTHESIS

GFAT

3PG

G6PDRu5P

6PGDPOLYOL PATHWAY

Sorbitol

NADPH

AR

NADP+

3DG

NAD+NADH

Fructose

AGEs

ATP

ADPPGK

GAPDHNAD+

NADH

GlucN6PUDP-GlcNAcGlycosylation

HEXOSAMINE BIOSYNTHESIS PATHWAY

NADPHNADP+ NADPHNADP+OXIDATIVE

NON-OXIDATIVE

GLYCOLYSIS

NADPH REDOXREGULATION

Cysteine

Folate

PyruvateLactateLDHMCT

GLUTAMINE METABOLISM

ORNITHINE CYCLE

Pyruvate

GS

GLS

α-ketoglutarate

NADPH

NADP+IDH1

NADPH

NADP+

Isocitrate

Citrate

Aspartate

L-ornithine

ARG

Citrulline

Arginine

Aspartate

Oxaloacetate

Malate

NO

ODC

POLYAMINESYNTHESIS

PENTOSE PHOSPHATE PATHWAYONE-CARBON METABOLISM ORNITHINE CYCLE POLYAMINE SYNTHESIS

Fumarate

Ornithine

FA FACPT

Oxaloacetate

Malate

Acetyl-CoA

ME

TCA CYCLE

POLYOL PATHWAY TCA CYCLE

Citrate

Isocitrate

α-ketoglutarate

Succinate

Fumarate

Malate

Oxaloacetate

IDH2

NADPH NADP+

Glutamate-γ-semialdehydeProline

ATP

ADP

Lactate

FATTY ACIDβ-OXIDATION

HEXOSAMINE BIOSYNTHESIS PATHWAYGLYCOLYSISGLUTAMINE METABOLISMFATTY ACID β-OXIDATION

HMG-CoA

Cholesterol

MEVALONATEPATHWAY

METHYLGLYOXAL PATHWAY NUCLEOTIDE SYNTHESISNADPH REDOX REGULATIONMEVALONATE PATHWAY

eNOS

ATP

NADH

GSH

Acetyl-CoA

Glutamine

SLC1A5

HK

Serine

Glycine

Methylation

METHIONINEMETABOLISM

FOLATEMETABOLISM

ONE-CARBONMETABOLISM

SERINEMETABOLISM

meTHF

THF

hCYS

METSAM

SAH

MS

mTHF

GlutamineGlutamate Glutamate

Mitochondria

Figure 1. Overview of general EC metabolism.For clarity, not all metabolites and enzymes of the depicted pathways are shown. Abbreviations: 3DG: 3-deoxyglucosone; 3PG: 3-phosphoglycerate; 6PGD: 6-phosphogluconatedehydrogenase; AGE: advanced glycation end-product; AR: aldose reductase; ARG: arginase; ATP: adenosine triphosphate; CPT: carnitine palmitoyltransferase; DHAP:dihydroxyacetone phosphate; eNOS: endothelial nitric oxide synthase; ETC: electron transport chain; F6P: fructose 6-phosphate; F1,6P2: fructose 1,6-bisphosphate; F2,6P2:fructose 2,6 bisphosphate; FA: fatty acid; G6P: glucose 6-phosphate; G6PD: glucose 6-phosphate dehydrogenase; GAPDH: glyceraldehyde 3-phosphate dehydrogenase; GFAT:glutamine-6-phosphate amidotransferase; GlucN6P: glucosamine-6-phosphate; GLS: glutaminase; GLUT: glucose transporter; GS: glutamine synthetase; GSH: glutathione:hCYS: homocysteine; HMG-CoA: hydroxymethylglutaryl coenzyme A; IDH; isocitrate dehydrogenase; LDH: lactate dehydrogenase; MCT: monocarboxylate transporter; ME:malic enzyme; MET: methionine; meTHF: 5.10-methylene-tetrahydrofolate; mTHF: 5-methyltetrahydrofolate; MS: methionine synthetase; NAD: nicotinamide adeninedinucleotide; NADPH: nicotinamide adenine dinucleotide phosphate; NO: nitric oxide; ODC: ornithine decarboxylase; PFK1: phosphofructokinase-1 PFKFB3: 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase-3; PGK: phosphoglycerate kinase; ROS: reactive oxygen species; RPI: ribose-5-phosphate isomerase; SAH: S-adenosylhomocysteine: SAM:S-adenosylmethionine; TCA cycle: tricarboxylic acid cycle; THF: tetrahydrofolate; TKT: transketolase; UDP-GlcNAc: uridine diphosphate N-acetylglucosamine.

ª 2014 The Authors EMBO Molecular Medicine Vol 6 | No 9 | 2014

Jermaine Goveia et al Targeting EC metabolism in disease EMBO Molecular Medicine

1107

and fatty acids in case of reduced glycolytic rates (Krutzfeldt et al,

1990; Dranka et al, 2010).

ECs lining peripheral tissue vessels or the blood brain barrier

(BBB) express multiple members of the two major families of sugar

transporters, that is, glucose transporters (GLUT) and sodium/

glucose co-transporters (SGLTs), but the high-affinity GLUT1 is

considered to be the main route of glucose uptake in ECs (Fig 1)

(Mann et al, 2003; Gaudreault et al, 2004, 2008; Sahoo et al, 2014).

Phosphorylation of intracellular glucose by hexokinase (HK)

destines it for metabolic utilization, predominately by conversion to

lactate via glycolysis (Fig 1) (Paik et al, 2005; De Bock et al,

2013b). Glycolytic intermediates also serve as precursors for biosyn-

thetic pathways including the pentose phosphate pathway (PPP),

hexosamine biosynthesis and glycogenesis (Fig 1, for an extensive

review see (De Bock et al, 2013a,b)).

The PPP consists of oxidative and non-oxidative branches, and

its overall flux is determined by the rate-limiting enzyme glucose-6-

phosphate dehydrogenase (G6PD) (Fig 1). Partially regulated by

VEGF signaling, G6PD destines glucose-6-phosphate (G6P) for

utilization in the PPP (Pan et al, 2009). The oxidative branch of the

PPP converts G6P into ribulose-5-phosphate (Ru5P) and produces

NADPH from NADP+, thereby generating reducing power to main-

tain EC redox balance and biosynthetic reactions (Dobrina & Rossi,

1983; Jongkind et al, 1989; Spolarics & Spitzer, 1993; Spolarics &

Wu, 1997; Vizan et al, 2009). The non-oxidative branch converts

Ru5P into xylulose-5-phosphate (Xu5P) and ribose-5-phosphate

(R5P), the latter is necessary for nucleotide biosynthesis (Pandolfi

et al, 1995). However, PPP intermediates may also be converted

back into glycolytic intermediates via the action of transketolase

(TKT) and transaldolase. These reactions are reversible, allowing

biosynthesis of macromolecules from glycolytic metabolites via the

non-oxidative arm.

The hexosamine biosynthesis pathway starts with the conver-

sion of the glycolytic intermediate fructose-6-phosphate (F6P) into

glucosamine-6-phosphate (GlucN6P) (Fig 1). GlucN6P is then

metabolized to uridine diphosphate N-acetylglucosamine (UDP-

GlcNAc), a key substrate for glycosylation reactions that control

many aspects of EC function (Benedito et al, 2009; Laczy et al,

Glossary

1C metabolismA complex metabolic network characterized by the transfer of carbonfrom serine/glycine for folate compound chemical reactions andinvolved in nucleotide, lipid and protein biosynthesis, redoxhomeostasis and production of methylation substrates.Advanced glycation end products (AGEs)Proteins or lipids that have been non-enzymatically glycated, often asa result of hyperglycemia and/or oxidative stress, that causedamaging intracellular and extracellular dysfunction.AngiogenesisGrowth of new blood vessels from existing microvasculature.EndotheliumContinuous inner lining of all vasculature composed of endothelialcells (ECs), which regulates physiological vascular function andangiogenesis.EC dysfunctionInability of endothelial cells to fulfill their physiological role asmediators of the blood barrier and vasotone.Fatty acid oxidationMetabolism of fatty acids in mitochondria into acetyl-CoA to fuel theTCA cycle.GlycolysisAnaerobic metabolism of glucose producing ATP and pyruvateGlycosylationA post-translational modification that enzymatically adds glycans, oroligosaccharides, to proteins and lipids.Hexosamine biosynthesis pathwaySide pathway from glycolytic intermediate fructose 6-phosphate (F6P)that produces substrates for glycosylation.IsoprenoidMevalonate pathway intermediates used for the production ofcholesterol and as substrates for prenylation.Metabolic fluxFlow of metabolites through a given metabolic pathway.Metabolic flux analysisQuantification of metabolic flux by tracing the fate of Isotope-labeledsubstrates.MetabolismThe spectrum of organic and chemical cellular reactions dedicated tothe production of energy and building blocks for general maintenanceand functionality.

Methylglyoxal pathwayGlycolytic side pathway from dihydroxyacetone phosphate (DHAP)that results in production of methylglyoxal and/or AGEs.Oxidative metabolismAerobic metabolic pathways that break down substrates throughoxidation for energy production and biosynthesis.Pentose phosphate pathwayMetabolic pathway important for redox homeostasis and biosynthesiswhich utilizes glucose-derived glucose-6-phosphate (G6P) forproduction of NADPH through its oxidative branch, and fructose 6-phosphate (F6P) and 3-phosphoglycerate (3PG) for nucleotideproduction in its non-oxidative branch.Polyol pathwayPathway implicated in diabetic endothelial dysfunction by reductionof glucose into sorbitol and then fructose to fuel production of AGEs.PrenylationPost-translational addition of isoprenoids such as farnesyl or geranyl–geranyl to a protein.QuiescenceCell state defined by a lack of activity.Reactive nitrogen speciesHighly reactive nitrogen-containing molecules that often interact withROS, promote oxidative stress and reduce bioavailability of nitricoxide.Reactive oxygen species (ROS)Highly reactive molecules that contain oxygen (produced by aerobicmetabolic processes) and are involved in normal cell homeostasis andsignaling, but whose accumulation, termed oxidative stress, leads tocell damage.Stalk cellEndothelial cells that trail migratory tip cells and proliferate to extendgrowth of a new blood vessel during sprouting angiogenesis.Tip cellMigratory endothelial cells that lead spouting microvessels up achemokine gradient during angiogenesis.

EMBO Molecular Medicine Vol 6 | No 9 | 2014 ª 2014 The Authors

EMBO Molecular Medicine Targeting EC metabolism in disease Jermaine Goveia et al

1106

ETC

MethylglyoxalAGEsMETHYLGLYOXAL PATHWAY

Glucose

GLUT1Glucose

G6P

F6P

F1,6P2

G3PDHAP

F2,6P2

PFKFB3PFK

TKT R5P

RPIPENTOSEPHOSPHATEPATHWAY

NUCLEOTIDESYNTHESIS

GFAT

3PG

G6PDRu5P

6PGDPOLYOL PATHWAY

Sorbitol

NADPH

AR

NADP+

3DG

NAD+NADH

Fructose

AGEs

ATP

ADPPGK

GAPDHNAD+

NADH

GlucN6PUDP-GlcNAcGlycosylation

HEXOSAMINE BIOSYNTHESIS PATHWAY

NADPHNADP+ NADPHNADP+OXIDATIVE

NON-OXIDATIVE

GLYCOLYSIS

NADPH REDOXREGULATION

Cysteine

Folate

PyruvateLactateLDHMCT

GLUTAMINE METABOLISM

ORNITHINE CYCLE

Pyruvate

GS

GLS

α-ketoglutarate

NADPH

NADP+IDH1

NADPH

NADP+

Isocitrate

Citrate

Aspartate

L-ornithine

ARG

Citrulline

Arginine

Aspartate

Oxaloacetate

Malate

NO

ODC

POLYAMINESYNTHESIS

PENTOSE PHOSPHATE PATHWAYONE-CARBON METABOLISM ORNITHINE CYCLE POLYAMINE SYNTHESIS

Fumarate

Ornithine

FA FACPT

Oxaloacetate

Malate

Acetyl-CoA

ME

TCA CYCLE

POLYOL PATHWAY TCA CYCLE

Citrate

Isocitrate

α-ketoglutarate

Succinate

Fumarate

Malate

Oxaloacetate

IDH2

NADPH NADP+

Glutamate-γ-semialdehydeProline

ATP

ADP

Lactate

FATTY ACIDβ-OXIDATION

HEXOSAMINE BIOSYNTHESIS PATHWAYGLYCOLYSISGLUTAMINE METABOLISMFATTY ACID β-OXIDATION

HMG-CoA

Cholesterol

MEVALONATEPATHWAY

METHYLGLYOXAL PATHWAY NUCLEOTIDE SYNTHESISNADPH REDOX REGULATIONMEVALONATE PATHWAY

eNOS

ATP

NADH

GSH

Acetyl-CoA

Glutamine

SLC1A5

HK

Serine

Glycine

Methylation

METHIONINEMETABOLISM

FOLATEMETABOLISM

ONE-CARBONMETABOLISM

SERINEMETABOLISM

meTHF

THF

hCYS

METSAM

SAH

MS

mTHF

GlutamineGlutamate Glutamate

Mitochondria

Figure 1. Overview of general EC metabolism.For clarity, not all metabolites and enzymes of the depicted pathways are shown. Abbreviations: 3DG: 3-deoxyglucosone; 3PG: 3-phosphoglycerate; 6PGD: 6-phosphogluconatedehydrogenase; AGE: advanced glycation end-product; AR: aldose reductase; ARG: arginase; ATP: adenosine triphosphate; CPT: carnitine palmitoyltransferase; DHAP:dihydroxyacetone phosphate; eNOS: endothelial nitric oxide synthase; ETC: electron transport chain; F6P: fructose 6-phosphate; F1,6P2: fructose 1,6-bisphosphate; F2,6P2:fructose 2,6 bisphosphate; FA: fatty acid; G6P: glucose 6-phosphate; G6PD: glucose 6-phosphate dehydrogenase; GAPDH: glyceraldehyde 3-phosphate dehydrogenase; GFAT:glutamine-6-phosphate amidotransferase; GlucN6P: glucosamine-6-phosphate; GLS: glutaminase; GLUT: glucose transporter; GS: glutamine synthetase; GSH: glutathione:hCYS: homocysteine; HMG-CoA: hydroxymethylglutaryl coenzyme A; IDH; isocitrate dehydrogenase; LDH: lactate dehydrogenase; MCT: monocarboxylate transporter; ME:malic enzyme; MET: methionine; meTHF: 5.10-methylene-tetrahydrofolate; mTHF: 5-methyltetrahydrofolate; MS: methionine synthetase; NAD: nicotinamide adeninedinucleotide; NADPH: nicotinamide adenine dinucleotide phosphate; NO: nitric oxide; ODC: ornithine decarboxylase; PFK1: phosphofructokinase-1 PFKFB3: 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase-3; PGK: phosphoglycerate kinase; ROS: reactive oxygen species; RPI: ribose-5-phosphate isomerase; SAH: S-adenosylhomocysteine: SAM:S-adenosylmethionine; TCA cycle: tricarboxylic acid cycle; THF: tetrahydrofolate; TKT: transketolase; UDP-GlcNAc: uridine diphosphate N-acetylglucosamine.

ª 2014 The Authors EMBO Molecular Medicine Vol 6 | No 9 | 2014

Jermaine Goveia et al Targeting EC metabolism in disease EMBO Molecular Medicine

1107

and fatty acids in case of reduced glycolytic rates (Krutzfeldt et al,

1990; Dranka et al, 2010).

ECs lining peripheral tissue vessels or the blood brain barrier

(BBB) express multiple members of the two major families of sugar

transporters, that is, glucose transporters (GLUT) and sodium/

glucose co-transporters (SGLTs), but the high-affinity GLUT1 is

considered to be the main route of glucose uptake in ECs (Fig 1)

(Mann et al, 2003; Gaudreault et al, 2004, 2008; Sahoo et al, 2014).

Phosphorylation of intracellular glucose by hexokinase (HK)

destines it for metabolic utilization, predominately by conversion to

lactate via glycolysis (Fig 1) (Paik et al, 2005; De Bock et al,

2013b). Glycolytic intermediates also serve as precursors for biosyn-

thetic pathways including the pentose phosphate pathway (PPP),

hexosamine biosynthesis and glycogenesis (Fig 1, for an extensive

review see (De Bock et al, 2013a,b)).

The PPP consists of oxidative and non-oxidative branches, and

its overall flux is determined by the rate-limiting enzyme glucose-6-

phosphate dehydrogenase (G6PD) (Fig 1). Partially regulated by

VEGF signaling, G6PD destines glucose-6-phosphate (G6P) for

utilization in the PPP (Pan et al, 2009). The oxidative branch of the

PPP converts G6P into ribulose-5-phosphate (Ru5P) and produces

NADPH from NADP+, thereby generating reducing power to main-

tain EC redox balance and biosynthetic reactions (Dobrina & Rossi,

1983; Jongkind et al, 1989; Spolarics & Spitzer, 1993; Spolarics &

Wu, 1997; Vizan et al, 2009). The non-oxidative branch converts

Ru5P into xylulose-5-phosphate (Xu5P) and ribose-5-phosphate

(R5P), the latter is necessary for nucleotide biosynthesis (Pandolfi

et al, 1995). However, PPP intermediates may also be converted

back into glycolytic intermediates via the action of transketolase

(TKT) and transaldolase. These reactions are reversible, allowing

biosynthesis of macromolecules from glycolytic metabolites via the

non-oxidative arm.

The hexosamine biosynthesis pathway starts with the conver-

sion of the glycolytic intermediate fructose-6-phosphate (F6P) into

glucosamine-6-phosphate (GlucN6P) (Fig 1). GlucN6P is then

metabolized to uridine diphosphate N-acetylglucosamine (UDP-

GlcNAc), a key substrate for glycosylation reactions that control

many aspects of EC function (Benedito et al, 2009; Laczy et al,

Glossary

1C metabolismA complex metabolic network characterized by the transfer of carbonfrom serine/glycine for folate compound chemical reactions andinvolved in nucleotide, lipid and protein biosynthesis, redoxhomeostasis and production of methylation substrates.Advanced glycation end products (AGEs)Proteins or lipids that have been non-enzymatically glycated, often asa result of hyperglycemia and/or oxidative stress, that causedamaging intracellular and extracellular dysfunction.AngiogenesisGrowth of new blood vessels from existing microvasculature.EndotheliumContinuous inner lining of all vasculature composed of endothelialcells (ECs), which regulates physiological vascular function andangiogenesis.EC dysfunctionInability of endothelial cells to fulfill their physiological role asmediators of the blood barrier and vasotone.Fatty acid oxidationMetabolism of fatty acids in mitochondria into acetyl-CoA to fuel theTCA cycle.GlycolysisAnaerobic metabolism of glucose producing ATP and pyruvateGlycosylationA post-translational modification that enzymatically adds glycans, oroligosaccharides, to proteins and lipids.Hexosamine biosynthesis pathwaySide pathway from glycolytic intermediate fructose 6-phosphate (F6P)that produces substrates for glycosylation.IsoprenoidMevalonate pathway intermediates used for the production ofcholesterol and as substrates for prenylation.Metabolic fluxFlow of metabolites through a given metabolic pathway.Metabolic flux analysisQuantification of metabolic flux by tracing the fate of Isotope-labeledsubstrates.MetabolismThe spectrum of organic and chemical cellular reactions dedicated tothe production of energy and building blocks for general maintenanceand functionality.

Methylglyoxal pathwayGlycolytic side pathway from dihydroxyacetone phosphate (DHAP)that results in production of methylglyoxal and/or AGEs.Oxidative metabolismAerobic metabolic pathways that break down substrates throughoxidation for energy production and biosynthesis.Pentose phosphate pathwayMetabolic pathway important for redox homeostasis and biosynthesiswhich utilizes glucose-derived glucose-6-phosphate (G6P) forproduction of NADPH through its oxidative branch, and fructose 6-phosphate (F6P) and 3-phosphoglycerate (3PG) for nucleotideproduction in its non-oxidative branch.Polyol pathwayPathway implicated in diabetic endothelial dysfunction by reductionof glucose into sorbitol and then fructose to fuel production of AGEs.PrenylationPost-translational addition of isoprenoids such as farnesyl or geranyl–geranyl to a protein.QuiescenceCell state defined by a lack of activity.Reactive nitrogen speciesHighly reactive nitrogen-containing molecules that often interact withROS, promote oxidative stress and reduce bioavailability of nitricoxide.Reactive oxygen species (ROS)Highly reactive molecules that contain oxygen (produced by aerobicmetabolic processes) and are involved in normal cell homeostasis andsignaling, but whose accumulation, termed oxidative stress, leads tocell damage.Stalk cellEndothelial cells that trail migratory tip cells and proliferate to extendgrowth of a new blood vessel during sprouting angiogenesis.Tip cellMigratory endothelial cells that lead spouting microvessels up achemokine gradient during angiogenesis.

EMBO Molecular Medicine Vol 6 | No 9 | 2014 ª 2014 The Authors

EMBO Molecular Medicine Targeting EC metabolism in disease Jermaine Goveia et al

1106

retinal and hindbrain vascularization in mice, showing that

increased glycolytic flux is required for growth factor-induced angio-

genesis (De Bock et al, 2013b). Moreover, PFKFB3 overexpression

in zebrafish drives EC specification into sprout forming tip cells,

even in the presence of tip cell-inhibitory Notch signals that

promote proliferating stalk elongating cells (De Bock et al, 2013b).

Increased glycolysis not only provides energy for proliferation and

biosynthesis, but also for locomotion. Specifically, PFKFB3 and

other glycolytic enzymes co-localize with F-actin bundles in filopodia

and lamellipodia to produce ATP needed for rapid actin remodeling,

underlying locomotion and tip cell formation (De Bock et al,

2013b).

The important role of glycolysis in angiogenesis provides

opportunities for therapeutic targeting. Indeed, pharmacological

blockade with 3-(3-pyridinyl)-1-(4-pyridinyl)-2-propen-1-one (3PO)

or EC-specific genetic silencing of PFKFB3 inhibits tumor growth in

vivo (Xu et al, 2014). In addition, 3PO inhibits glycolytic flux

partially and transiently and has recently shown efficacy in reducing

pathological angiogenesis in a variety of disease models (Schoors

et al, 2014b; Xu et al, 2014). The systemic harm caused by inhibit-

ing glycolysis is minimal, however, showing that even moderate,

short-term impairment of glycolysis renders ECs more quiescent

without overt detrimental side effects (Schoors et al, 2014b). The

finding that partial and transient reduction of glycolysis may be

sufficient to inhibit pathological angiogenesis provides a paradigm

shift in our thinking about anti-glycolytic therapies, away from

complete and permanent blockade of glycolysis, which can induce

undesired adverse systemic effects.

Aside from serving as an energy source or building blocks for

biosynthesis, glycolytic metabolites can also modulate angiogenesis

by acting as bona fide signaling molecules. This is evidenced by the

observation that glycolytic tumor cells secrete lactate, which is

taken up by ECs through the monocarboxylate transporter 1 (MCT1)

(Fig 2A) (Sonveaux et al, 2012). Instead of being metabolized,

A The glycolytic pathway drives pathological angiogenesis

TCA CYCLE

HEXOSAMINE BIOSYNTHESIS PATHA WAYAAGLYCOLL LYSISLL

PHD REGULATION

Pyruvate

Acetyl-CoA

Lactate Lactate

PHD2

PFKFB3

HIF1α

IL-8FGFVEGFR-2

MCT

Glucose

Glucose

GLUT1

F6P

F1,6P2

GlucN6P UDP-GlcNAc

VEGFR-2glycosylationPFK

LDH

Galectin-1 VEGFindependentsignaling

ReducedIncreased

αα-ketoglutarate--ketoglutarate

FATGFAT

Acetyl-CoAFatty acidoxidation

B Altered glucose metabolism and low NO levels are associatedwith pulmonary artery hypertension

PENTOSE PHOSPHATE PATHWAY

ORNITHINE CYCLE

POLYAMINEYY SYNTHESIS

TCA CYCLE

GLYCOLL LYSISLL

FATTYA ACID β-OXIDATIONA

Pyruvate

GLUT1

G6P

F6P

Lactate

Glucose

Citrate

Isocitrate

α-ketoglutarateIDH

MnSOD

ROS

Lactate

L-arginine

eNOS

Ornithine

ARG

NO

6PGXu5P

R5P

F6P

G3PG3P

Glucose

6PDG6PDRPIA

RPETKT

PutrescineSpermidineSpermineSRMSMS

ODC

PGK

ReducedIncreased

Ru5P

Figure 2. Metabolic pathways implicated in diseases characterized by pathological angiogenesis or hyperproliferative ECs.(A) Angiogenic ECs rely on glycolysis, instead of oxidative metabolism, for ATP production and upregulate PFKFB3 to increase the conversion of glucose into lactate throughglycolysis. Lactate is secreted and taken up through lactate transporters. High Lactate influx through MCT1 results in increased intracellular lactate levels that compete witha-ketoglutarate for PHD-2 binding, leading to HIF-1a stabilization and upregulation of pro-angiogenic genes. VEGFR-2 glycosylation is required for galectin-1-induced VEGF-independent signaling. (B) PAH ECs are metabolically characterized by high aerobic glycolysis and low oxidative metabolism. NO production through eNOS is impaired due toupregulation of arginase II and increased oxidative stress due to limited availability of MnSOD. In addition, several enzymes in the pentose phosphate pathway and polyaminebiosynthesis pathway are differentially expressed in PAH ECs, but the importance of these findings remains to be determined (B). Green font / bold line: upregulated, red font /broken line: downregulated. For clarity, not all metabolites and enzymes of the depicted pathways are shown. Abbreviations: as in Fig 1. FGF: fibroblast growth factor; HIF:hypoxia-inducible factor; IL: interleukin; PHD: prolyl hydroxylase domain; R5P: ribose-5-phosphate; RPE: ribulose-5-phosphate 3-epimerase; RPIA: ribose-5-phosphateisomerase; Ru5P: ribulose-5-phosphate; SRM: spermidine synthase; VEGFR: vascular endothelial growth factor receptor; Xu5P: xylulose-5-phosphate.

ª 2014 The Authors EMBO Molecular Medicine Vol 6 | No 9 | 2014

Jermaine Goveia et al Targeting EC metabolism in disease EMBO Molecular Medicine

1109

2009; Croci et al, 2014). The polyol pathway and methylglyoxal

pathways are glycolysis side-pathways that are mostly known for

their role in cardiovascular disease (Fig 1; see below) (Goldin

et al, 2006).

Other metabolic pathways are less well characterized in ECs.

Fatty acid (FA) oxidation (FAO) and glutamine oxidation have been

implicated in replenishing the TCA cycle to produce ATP via oxida-

tive phosphorylation (Fig 1) (Leighton et al, 1987; Hinshaw &

Burger, 1990; Dagher et al, 1999, 2001; De Bock et al, 2013b).

However, since ECs predominately rely on glucose metabolism to

provide ATP, the energetic function of FAO and glutamine oxidation

is not clear (De Bock et al, 2013b). FAs and amino acids can serve

as precursors for biomass production, but such a role in ECs has not

been demonstrated using isotope tracer labeling studies. FAO

produces significant amounts of nicotinamide adenine dinucleotide

phosphate (NADPH), which is an important co-factor in many

biosynthetic reactions and essential to maintain redox balance. In

addition, FAO generates acetyl-coA which is another important

precursor for biomolecule production.

For example, acetyl-CoA is used, among other things, for the

synthesis of cholesterol via the mevalonate pathway (Fig 1).

Although endothelial cholesterol metabolism has been poorly stud-

ied, perturbations in cholesterol homeostasis are known to affect

key EC functions such as intracellular signaling, inflammatory acti-

vation, nitric oxide synthesis and angiogenesis (Boger et al, 2000;

Ivashchenko et al, 2010; Whetzel et al, 2010; Xu et al, 2010; Fang

et al, 2013). ECs express all the cholesterol biosynthesis enzymes

and the LDL receptor for extracellular uptake (Fig 1). These proteins

are under transcriptional control of the sterol regulatory element

binding protein (SREBP1 and -2) and liver X receptors (LXR)

(Noghero et al, 2012). SREBP1 and LXRs inhibit cholesterol synthe-

sis and absorption, whereas SREBP2 induces synthesis and inhibits

cholesterol efflux via transcriptional repression of the ATP-binding

cassette (ABC) transporter 1 ABCA1, which together with ABCG1

mediates cholesterol efflux from ECs (Hassan et al, 2006). Notably,

endothelial SREBP2 also controls expression of arginine metabolism

enzymes, although the physiological significance of this interaction

between cholesterol and arginine metabolism remains to be deter-

mined (Zeng et al, 2004).

Arginine and glutamine are the best studied amino acids

(AAs) in ECs. Arginine is a metabolite in the ornithine cycle and

converted into citruline and nitric oxide (NO) by endothelial

nitric oxide synthase (eNOS) (Fig 1) (Sessa et al, 1990). Altera-

tions in arginine and eNOS metabolism are among the best-

characterized causes of EC dysfunction and a prime therapeutic

target (Leiper & Nandi, 2011). Glutamine is the most abundant

AA in the peripheral blood and preferentially taken up by ECs

via the solute carrier family 1 member 5 (SLC1A5) trans-

porter (Fig 1) (Herskowitz et al, 1991; Pan et al, 1995).

Glutamine-utilizing pathways are mainly biosynthetic and can be

divided into those that utilize the c-nitrogen (nucleotide biosyn-

thesis, hexosamine biosynthesis, asparagine synthesis) and those

that use the a-nitrogen or carbon backbone (DeBerardinis &

Cheng, 2010). The latter reactions use glutamine-derived gluta-

mate rather than glutamine itself and include glutathione (GSH)

synthesis, anaplerotic refueling of the TCA cycle and biosynthesis

of polyamines, proline and other non-essential AAs (NEAAs)

(Fig 1) (DeBerardinis & Cheng, 2010).

Serine and glycine are especially interesting examples of gluta-

mine / glutamate-derived NEAAs, not only because of their direct

effects on ECs (Weinberg et al, 1992; Rose et al, 1999; Yamashina

et al, 2001; Mishra et al, 2008; den Eynden et al, 2009; McCarty

et al, 2009; Stobart et al, 2013), but also since their synthesis

requires both the glutamate a-nitrogen and the glycolytic intermedi-

ate 3-phosphoglycerate (3PG) (Fig 1) (Locasale, 2013). Hence,

serine and glycine metabolism integrates metabolic input from

central carbon (glycolysis) and nitrogen (glutamine) metabolism.

Moreover, the reversible interconversion of serine and glycine is

directly coupled to one-carbon metabolism, intermediates of which

are considered important targets to treat cardiovascular disease

(Fig 1; see below) (Locasale, 2013). In fact, while EC metabolism is

largely understudied, several of the above-mentioned metabolic

pathways have been implicated as mediators of pathological angio-

genesis or EC dysfunction.

EC metabolism in diseases characterized by angiogenesisand EC hyperproliferation

Cancer

Tumors need blood vessels to supply oxygen and detoxify waste

products (Jain, 1987; Papetti & Herman, 2002; Welti et al, 2013).

When tumors become too large to allow adequate diffusion of

oxygen and nutrients from local vasculature they secrete pro-

angiogenic growth factors to induce angiogenesis (Bergers &

Benjamin, 2003). Pharmacological inhibition of growth factor

signaling (primarily vascular endothelial growth factor (VEGF)

signaling) is the only clinically approved anti-angiogenic strategy,

but the benefits are limited as tumors acquire resistance within

months after treatment initiation (Bergers & Hanahan, 2008; Carme-

liet & Jain, 2011; Ebos & Kerbel, 2011; Welti et al, 2013). Escape

from anti-angiogenic therapy is mediated by increased secretion of

pro-angiogenic factors, activation of alternative angiogenic signaling

pathways, recruitment of pro-angiogenic accessory cells and other

mechanisms (Loges et al, 2010; Sennino & McDonald, 2012). A

recent report indicated that glycosylation-dependent interactions of

galectin-1 with VEGF receptor 2 (VEGFR2) could activate pro-angio-

genic signaling even when the VEGF ligand is blocked (Fig 2A)

(Croci et al, 2014). Hence, angiogenic signaling is robust and redun-

dant, and inhibition of individual signaling molecules and growth

factors can be overcome by escape mechanisms.

The switch from a quiescent to an angiogenic phenotype (as

occurs in cancer) is metabolically demanding and mediated by

adaptations in EC metabolism (Fig 2A). While the changes in meta-

bolic fluxes of ECs, freshly isolated from tumors, have not been

characterized yet, ECs in tumors and inflamed tissues likely

resemble highly activated ECs. Lactate dehydrogenase B (LDH-B) is

upregulated in tumor endothelium, and VEGF signaling increases

glycolytic flux by inducing GLUT1 and the glycolytic enzyme

6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase-3 (PFKFB3)

(Fig 2A) (van Beijnum et al, 2006; Yeh et al, 2008; De Bock et al,

2013b). PFKFB3 catalyzes the synthesis of fructose-2,6-bisphosphate

(F2,6P2), which is an allosteric activator of 6-phosphofructo-1-kinase

(PFK-1) (Van Schaftingen et al, 1982). PFK-1 converts fructose-6-

phosphate (F6P) to fructose-1,6-bisphosphate (F1,6P2) in the rate-

limiting step of glycolysis. EC-specific PFKFB3 deletion diminishes

EMBO Molecular Medicine Vol 6 | No 9 | 2014 ª 2014 The Authors

EMBO Molecular Medicine Targeting EC metabolism in disease Jermaine Goveia et al

1108

retinal and hindbrain vascularization in mice, showing that

increased glycolytic flux is required for growth factor-induced angio-

genesis (De Bock et al, 2013b). Moreover, PFKFB3 overexpression

in zebrafish drives EC specification into sprout forming tip cells,

even in the presence of tip cell-inhibitory Notch signals that

promote proliferating stalk elongating cells (De Bock et al, 2013b).

Increased glycolysis not only provides energy for proliferation and

biosynthesis, but also for locomotion. Specifically, PFKFB3 and

other glycolytic enzymes co-localize with F-actin bundles in filopodia

and lamellipodia to produce ATP needed for rapid actin remodeling,

underlying locomotion and tip cell formation (De Bock et al,

2013b).

The important role of glycolysis in angiogenesis provides

opportunities for therapeutic targeting. Indeed, pharmacological

blockade with 3-(3-pyridinyl)-1-(4-pyridinyl)-2-propen-1-one (3PO)

or EC-specific genetic silencing of PFKFB3 inhibits tumor growth in

vivo (Xu et al, 2014). In addition, 3PO inhibits glycolytic flux

partially and transiently and has recently shown efficacy in reducing

pathological angiogenesis in a variety of disease models (Schoors

et al, 2014b; Xu et al, 2014). The systemic harm caused by inhibit-

ing glycolysis is minimal, however, showing that even moderate,

short-term impairment of glycolysis renders ECs more quiescent

without overt detrimental side effects (Schoors et al, 2014b). The

finding that partial and transient reduction of glycolysis may be

sufficient to inhibit pathological angiogenesis provides a paradigm

shift in our thinking about anti-glycolytic therapies, away from

complete and permanent blockade of glycolysis, which can induce

undesired adverse systemic effects.

Aside from serving as an energy source or building blocks for

biosynthesis, glycolytic metabolites can also modulate angiogenesis

by acting as bona fide signaling molecules. This is evidenced by the

observation that glycolytic tumor cells secrete lactate, which is

taken up by ECs through the monocarboxylate transporter 1 (MCT1)

(Fig 2A) (Sonveaux et al, 2012). Instead of being metabolized,

A The glycolytic pathway drives pathological angiogenesis

TCA CYCLE

HEXOSAMINE BIOSYNTHESIS PATHA WAYAAGLYCOLL LYSISLL

PHD REGULATION

Pyruvate

Acetyl-CoA

Lactate Lactate

PHD2

PFKFB3

HIF1α

IL-8FGFVEGFR-2

MCT

Glucose

Glucose

GLUT1

F6P

F1,6P2

GlucN6P UDP-GlcNAc

VEGFR-2glycosylationPFK

LDH

Galectin-1 VEGFindependentsignaling

ReducedIncreased

αα-ketoglutarate--ketoglutarate

FATGFAT

Acetyl-CoAFatty acidoxidation

B Altered glucose metabolism and low NO levels are associatedwith pulmonary artery hypertension

PENTOSE PHOSPHATE PATHWAY

ORNITHINE CYCLE

POLYAMINEYY SYNTHESIS

TCA CYCLE

GLYCOLL LYSISLL

FATTYA ACID β-OXIDATIONA

Pyruvate

GLUT1

G6P

F6P

Lactate

Glucose

Citrate

Isocitrate

α-ketoglutarateIDH

MnSOD

ROS

Lactate

L-arginine

eNOS

Ornithine

ARG

NO

6PGXu5P

R5P

F6P

G3PG3P

Glucose

6PDG6PDRPIA

RPETKT

PutrescineSpermidineSpermineSRMSMS

ODC

PGK

ReducedIncreased

Ru5P

Figure 2. Metabolic pathways implicated in diseases characterized by pathological angiogenesis or hyperproliferative ECs.(A) Angiogenic ECs rely on glycolysis, instead of oxidative metabolism, for ATP production and upregulate PFKFB3 to increase the conversion of glucose into lactate throughglycolysis. Lactate is secreted and taken up through lactate transporters. High Lactate influx through MCT1 results in increased intracellular lactate levels that compete witha-ketoglutarate for PHD-2 binding, leading to HIF-1a stabilization and upregulation of pro-angiogenic genes. VEGFR-2 glycosylation is required for galectin-1-induced VEGF-independent signaling. (B) PAH ECs are metabolically characterized by high aerobic glycolysis and low oxidative metabolism. NO production through eNOS is impaired due toupregulation of arginase II and increased oxidative stress due to limited availability of MnSOD. In addition, several enzymes in the pentose phosphate pathway and polyaminebiosynthesis pathway are differentially expressed in PAH ECs, but the importance of these findings remains to be determined (B). Green font / bold line: upregulated, red font /broken line: downregulated. For clarity, not all metabolites and enzymes of the depicted pathways are shown. Abbreviations: as in Fig 1. FGF: fibroblast growth factor; HIF:hypoxia-inducible factor; IL: interleukin; PHD: prolyl hydroxylase domain; R5P: ribose-5-phosphate; RPE: ribulose-5-phosphate 3-epimerase; RPIA: ribose-5-phosphateisomerase; Ru5P: ribulose-5-phosphate; SRM: spermidine synthase; VEGFR: vascular endothelial growth factor receptor; Xu5P: xylulose-5-phosphate.

ª 2014 The Authors EMBO Molecular Medicine Vol 6 | No 9 | 2014

Jermaine Goveia et al Targeting EC metabolism in disease EMBO Molecular Medicine

1109

2009; Croci et al, 2014). The polyol pathway and methylglyoxal

pathways are glycolysis side-pathways that are mostly known for

their role in cardiovascular disease (Fig 1; see below) (Goldin

et al, 2006).

Other metabolic pathways are less well characterized in ECs.

Fatty acid (FA) oxidation (FAO) and glutamine oxidation have been

implicated in replenishing the TCA cycle to produce ATP via oxida-

tive phosphorylation (Fig 1) (Leighton et al, 1987; Hinshaw &

Burger, 1990; Dagher et al, 1999, 2001; De Bock et al, 2013b).

However, since ECs predominately rely on glucose metabolism to

provide ATP, the energetic function of FAO and glutamine oxidation

is not clear (De Bock et al, 2013b). FAs and amino acids can serve

as precursors for biomass production, but such a role in ECs has not

been demonstrated using isotope tracer labeling studies. FAO

produces significant amounts of nicotinamide adenine dinucleotide

phosphate (NADPH), which is an important co-factor in many

biosynthetic reactions and essential to maintain redox balance. In

addition, FAO generates acetyl-coA which is another important

precursor for biomolecule production.

For example, acetyl-CoA is used, among other things, for the

synthesis of cholesterol via the mevalonate pathway (Fig 1).

Although endothelial cholesterol metabolism has been poorly stud-

ied, perturbations in cholesterol homeostasis are known to affect

key EC functions such as intracellular signaling, inflammatory acti-

vation, nitric oxide synthesis and angiogenesis (Boger et al, 2000;

Ivashchenko et al, 2010; Whetzel et al, 2010; Xu et al, 2010; Fang

et al, 2013). ECs express all the cholesterol biosynthesis enzymes

and the LDL receptor for extracellular uptake (Fig 1). These proteins

are under transcriptional control of the sterol regulatory element

binding protein (SREBP1 and -2) and liver X receptors (LXR)

(Noghero et al, 2012). SREBP1 and LXRs inhibit cholesterol synthe-

sis and absorption, whereas SREBP2 induces synthesis and inhibits

cholesterol efflux via transcriptional repression of the ATP-binding

cassette (ABC) transporter 1 ABCA1, which together with ABCG1

mediates cholesterol efflux from ECs (Hassan et al, 2006). Notably,

endothelial SREBP2 also controls expression of arginine metabolism

enzymes, although the physiological significance of this interaction

between cholesterol and arginine metabolism remains to be deter-

mined (Zeng et al, 2004).

Arginine and glutamine are the best studied amino acids

(AAs) in ECs. Arginine is a metabolite in the ornithine cycle and

converted into citruline and nitric oxide (NO) by endothelial

nitric oxide synthase (eNOS) (Fig 1) (Sessa et al, 1990). Altera-

tions in arginine and eNOS metabolism are among the best-

characterized causes of EC dysfunction and a prime therapeutic

target (Leiper & Nandi, 2011). Glutamine is the most abundant

AA in the peripheral blood and preferentially taken up by ECs

via the solute carrier family 1 member 5 (SLC1A5) trans-

porter (Fig 1) (Herskowitz et al, 1991; Pan et al, 1995).

Glutamine-utilizing pathways are mainly biosynthetic and can be

divided into those that utilize the c-nitrogen (nucleotide biosyn-

thesis, hexosamine biosynthesis, asparagine synthesis) and those

that use the a-nitrogen or carbon backbone (DeBerardinis &

Cheng, 2010). The latter reactions use glutamine-derived gluta-

mate rather than glutamine itself and include glutathione (GSH)

synthesis, anaplerotic refueling of the TCA cycle and biosynthesis

of polyamines, proline and other non-essential AAs (NEAAs)

(Fig 1) (DeBerardinis & Cheng, 2010).

Serine and glycine are especially interesting examples of gluta-

mine / glutamate-derived NEAAs, not only because of their direct

effects on ECs (Weinberg et al, 1992; Rose et al, 1999; Yamashina

et al, 2001; Mishra et al, 2008; den Eynden et al, 2009; McCarty

et al, 2009; Stobart et al, 2013), but also since their synthesis

requires both the glutamate a-nitrogen and the glycolytic intermedi-

ate 3-phosphoglycerate (3PG) (Fig 1) (Locasale, 2013). Hence,

serine and glycine metabolism integrates metabolic input from

central carbon (glycolysis) and nitrogen (glutamine) metabolism.

Moreover, the reversible interconversion of serine and glycine is

directly coupled to one-carbon metabolism, intermediates of which

are considered important targets to treat cardiovascular disease

(Fig 1; see below) (Locasale, 2013). In fact, while EC metabolism is

largely understudied, several of the above-mentioned metabolic

pathways have been implicated as mediators of pathological angio-

genesis or EC dysfunction.

EC metabolism in diseases characterized by angiogenesisand EC hyperproliferation

Cancer

Tumors need blood vessels to supply oxygen and detoxify waste

products (Jain, 1987; Papetti & Herman, 2002; Welti et al, 2013).

When tumors become too large to allow adequate diffusion of

oxygen and nutrients from local vasculature they secrete pro-

angiogenic growth factors to induce angiogenesis (Bergers &

Benjamin, 2003). Pharmacological inhibition of growth factor

signaling (primarily vascular endothelial growth factor (VEGF)

signaling) is the only clinically approved anti-angiogenic strategy,

but the benefits are limited as tumors acquire resistance within

months after treatment initiation (Bergers & Hanahan, 2008; Carme-

liet & Jain, 2011; Ebos & Kerbel, 2011; Welti et al, 2013). Escape

from anti-angiogenic therapy is mediated by increased secretion of

pro-angiogenic factors, activation of alternative angiogenic signaling

pathways, recruitment of pro-angiogenic accessory cells and other

mechanisms (Loges et al, 2010; Sennino & McDonald, 2012). A

recent report indicated that glycosylation-dependent interactions of

galectin-1 with VEGF receptor 2 (VEGFR2) could activate pro-angio-

genic signaling even when the VEGF ligand is blocked (Fig 2A)

(Croci et al, 2014). Hence, angiogenic signaling is robust and redun-

dant, and inhibition of individual signaling molecules and growth

factors can be overcome by escape mechanisms.

The switch from a quiescent to an angiogenic phenotype (as

occurs in cancer) is metabolically demanding and mediated by

adaptations in EC metabolism (Fig 2A). While the changes in meta-

bolic fluxes of ECs, freshly isolated from tumors, have not been

characterized yet, ECs in tumors and inflamed tissues likely

resemble highly activated ECs. Lactate dehydrogenase B (LDH-B) is

upregulated in tumor endothelium, and VEGF signaling increases

glycolytic flux by inducing GLUT1 and the glycolytic enzyme

6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase-3 (PFKFB3)

(Fig 2A) (van Beijnum et al, 2006; Yeh et al, 2008; De Bock et al,

2013b). PFKFB3 catalyzes the synthesis of fructose-2,6-bisphosphate

(F2,6P2), which is an allosteric activator of 6-phosphofructo-1-kinase

(PFK-1) (Van Schaftingen et al, 1982). PFK-1 converts fructose-6-

phosphate (F6P) to fructose-1,6-bisphosphate (F1,6P2) in the rate-

limiting step of glycolysis. EC-specific PFKFB3 deletion diminishes

EMBO Molecular Medicine Vol 6 | No 9 | 2014 ª 2014 The Authors

EMBO Molecular Medicine Targeting EC metabolism in disease Jermaine Goveia et al

1108

of intracellular NADPH, which is necessary to convert oxidized

glutathione (GSSH) into reduced GSH, a critical ROS scavenger

(Fig 3A) (Leopold et al, 2003; Zhang et al, 2012). Therefore, by

reducing PPP flux, high glucose depletes NADPH levels and

contributes to ROS accumulation (Goldin et al, 2006). Interest-

ingly, G6PD overexpression restores redox homeostasis in high

glucose cultured ECs (Leopold et al, 2003; Zhang et al, 2012).

Some studies suggest that high glucose shifts the normally glyco-

lytic EC metabolism toward oxidative metabolism and increased

mitochondrial respiration (Fig 3). However, these results appear

contextual, as other studies did not report such an induction of

oxidative metabolism (Nishikawa et al, 2000; Koziel et al, 2012;

Pangare & Makino, 2012; Dymkowska et al, 2014). While the

precise effects on mitochondrial respiration require further study,

hyperglycemia-induced mitochondrial ROS induces DNA breaks

and thereby activates polyAPD-ribose polymerase (PARP-1) (Du

et al, 2000, 2003; Nishikawa et al, 2000; Giacco & Brownlee, 2010;

Blake & Trounce, 2013). PolyADP-ribosylation by PARP-1 inacti-

vates GAPDH and stalls glycolysis, allowing accumulation of glyco-

lytic metabolites (Du et al, 2003).

Accumulation of F6P increases the flux through the hexosamine

biosynthesis pathway (HBP), which produces UDP-GlcNac, an

important precursor of glycosylation reactions (Fig 3A) (Brownlee,

2001). While glycosylation is important for physiological EC func-

tion, hyperglycemia-induced protein glycosylation inhibits angio-

genic functions (Du et al, 2001; Federici et al, 2002; Luo et al,

2008). Other glycolytic intermediates are diverted into the polyol

and methylglyoxal pathways that produce damaging agents such

as ROS and advanced glycation end products (AGEs) (Fig 3A)

(Goldin et al, 2006). AGEs induce vascular dysfunction by altering

extracellular matrix protein function and dysregulating cytokine

expression (Yan et al, 2008). In addition, receptor of AGE (RAGE)

binding by AGEs in vascular cells causes inflammation and

reduced NO availability associated with vascular complications in

ReducedIncreased

Glucose

Acetyl-CoA

GLUT1

PPP

L-arginine

gnggi

npppupl

uplinin

up

ROS

NADH

ReducedIncreased

Figure 3. Metabolic pathways implicated in diseases characterized by EC dysfunction.(A) High glucose levels in diabetes pushes glycolytic flux and cause ROS production and AGE formation. (B) Metabolic alterations that cause eNOS dysfunction mediateatherosclerosis pathogenesis. Asymmetric dimethylarginine (ADMA) competes with arginine for binding to eNOS. Arginase expression is increased and eNOS expression isdecreased, leading to reduced eNOS activity. 1C metabolism and mevalonate metabolism provide eNOS coupling co-factors and inhibit ROS production. The mevalonatepathway also provides farnesyl pyrophosphate (FPP) and geranylgeranyl pyrophosphate (GGPP), required for GTPase prenylation. For clarity, not all metabolites and enzymesof the depicted pathways are shown. Green font / bold line: upregulated, red font / broken line: downregulated. Abbreviations: as in Figure 1. BH2: dihydrobiopterin; BH4:tetrahydrobiopterin; ADMA: asymmetric dimethylarginine; CoQ10: coenzyme Q10; DDAH: dimethylarginine dimethylaminohydrolase; DHF: dihydrofolate; DHFR:dihydrofolate reductase; FPP: farnesyl pyrophosphate; GGPP: geranylgeranyl pyrophosphate; GTP: Guanosine triphosphate; HMGCR: hydroxymethylglutaryl coenzyme Areductase; PRMT: protein arginine methyltransferase.

ª 2014 The Authors EMBO Molecular Medicine Vol 6 | No 9 | 2014

Jermaine Goveia et al Targeting EC metabolism in disease EMBO Molecular Medicine

1111

lactate induces HIF-1a activation leading to increased expression of

VEGFR2 and bFGF (Sonveaux et al, 2012). Moreover, lactate

competes with a-ketoglutarate for binding to the oxygen sensing

prolyl hydroxylase-2 (PHD-2), resulting in diminished PHD-2 activ-

ity and subsequent hypoxia-inducible factor-1a (HIF-1a) stabiliza-

tion (Fig 2A). Stabilized HIF-1a induces pro-angiogenic signaling

pathways such as nuclear factor kappa-light-chain-enhancer of acti-

vated B-cells (NFkB)/interleukin 8 (IL-8) leading to increased angio-

genesis (Fig 2A) (Hunt et al, 2007; Vegran et al, 2011; Sonveaux

et al, 2012). Exploratory studies found that lactate induces angio-

genesis in vivo and that pharmacological blockade of MCT1 inhibits

angiogenesis and reduces tumor growth in mice (Sonveaux et al,

2012). Together, these data suggest an intricate relationship

between classical pro-angiogenic signals such as VEGF, HIF-1a and

hypoxia, and EC glucose metabolism. Targeting EC glucose metabo-

lism to inhibit tumor angiogenesis is in its infancy as a therapeutic

strategy, but recent evidence suggests its viability.

Pulmonary arterial hypertension

Idiopathic pulmonary arterial hypertension (PAH) is characterized

by heightened pressure in pulmonary arteries caused by excessive

EC proliferation and vascular dysfunction (Xu & Erzurum, 2011).

Emerging evidence indicates that metabolic abnormalities underlie

PAH (Fig 2B) (Sutendra & Michelakis, 2014; Zhao et al, 2014). In

line with recent findings that glycolysis regulates angiogenesis,

hyperproliferative PAH ECs rely on increased glycolytic flux and

reduced oxygen consumption, which may be related to HIF-1aoverexpression (Fig 2B) (Xu et al, 2007; Fijalkowska et al, 2010;

Majmundar et al, 2010; Tuder et al, 2012). Human pulmonary

ECs expressing mutated bone morphogenetic protein receptor 2

(BMPR2), which confers PAH, show altered expression of several

glycolytic enzymes including GLUT1 and phosphoglycerate kinase 1

(PGK1). PAH ECs also show increased expression of enzymes of the

PPP (R5P isomerase, Ru5P-3-epimerase) and polyamine biosynthe-

sis pathway (ornithine decarboxylase (ODC), spermine synthase

(SMS)). These metabolic changes may underlie the rapid prolifera-

tion of PAH ECs, since glycolysis, the PPP and mitogenic polyam-

ines all promote cellular proliferation (Morrison & Seidel, 1995).

However, the expression of other PPP and polyamine enzymes

[G6PD, TKT, spermidine synthase (SRM)] is reduced—a finding

that requires further explanation (Fig 2B) (Atkinson et al, 2002;

Rudarakanchana et al, 2002; Long et al, 2006; Fessel et al, 2012). In

addition, ECs isolated from EC-specific BMPR2 mutant mice show

similarly increased expression of PGK1, indicating altogether that

alterations in glycolysis as well as PPP likely underlie PAH (Majka

et al, 2011).

In addition to alterations in glycolysis, idiopathic PAH ECs have

fewer mitochondria and decreased mitochondrial metabolic activity

(Xu et al, 2007). BMPR2 mutant ECs have reduced quantities of TCA

cycle intermediates, reduced fatty acid oxidation and transcriptional

reduction of several enzymes involved in fatty acid metabolism,

including the rate-limiting enzyme of fatty acid oxidation carnitine

palmitoyltransferase 1 (CPT1) (Fig 2) (Fessel et al, 2012). Together,

these findings suggest reduced oxidative metabolism. Indeed, phar-

macological inhibition of hyper-activated pyruvate dehydrogenase

kinase (PDK), an enzyme that shunts glucose-derived pyruvate away

from oxidative TCA metabolism, has shown therapeutic efficacy.

However, whether these effects are mediated via ECs specifically

remains to be determined (McMurtry et al, 2004). For unexplained

reasons, PAH patients also show increased isocitrate dehydrogenase

(IDH)-1 and IDH-2 serum activity, a finding that corroborates with

the increased IDH activity observed in BPMR2 mutant ECs (Fessel

et al, 2012). Still, the mechanisms that alter metabolic pathways in

PAH ECs and the importance of some of these metabolic adaptations

in the pathogenesis of PAH remain unclear.

Reduced nitric oxide (NO) levels are another hallmark of PAH

ECs (Fijalkowska et al, 2010). Low NO levels may be related to the

reduced levels of the mitochondrial antioxidant manganese superox-

ide dismutase (MnSOD) (Fijalkowska et al, 2010). Indeed, MnSOD

increases NO availability by clearing superoxide anion, which inac-

tivates NO to form peroxynitrite (Fig 2) (Masri et al, 2008).

However, other factors likely contribute to the low NO levels in

PAH ECs (Xu et al, 2004). Indeed, human PAH ECs express high

levels of arginase II, which competes with endothelial nitric oxide

synthetase (eNOS) for their common substrate L-arginine (Fig 2)

(Xu et al, 2004). Inhibition of endothelial arginase II increases NO

production in vitro, suggesting that arginase II can be targeted to

prevent EC hyperproliferation and restore NO availability (Krotova

et al, 2010). While the mechanisms that induce abnormal metabolic

activity in PAH ECs are understudied, restoring NO may provide

dual benefits in preventing excessive EC proliferation as well as

restoring EC vasoactivity.

The metabolic adaptations in PAH (high glycolytic rates and

reduced oxidative metabolism) are partly reminiscent of the meta-

bolic profile of angiogenic ECs. It would be thus interesting to deter-

mine if reducing glycolysis by pharmacological blockade of PFKFB3

can reduce the hyperproliferative rate in PAH ECs. Alternatively, the

beneficial effects of PDK inhibition in PAH to induce oxidative

metabolism could also be beneficial to block angiogenesis by

preventing the glycolytic switch in ECs. Indeed, PDK blockade

with dichloroacetate inhibits angiogenesis in glioblastoma patients

(Michelakis et al, 2010).

EC metabolism in diseases characterized byEC dysfunction

Diabetes

Diabetes is characterized by high blood glucose levels that affect EC

metabolism and cause dysfunction (Fig 3A) (Blake & Trounce,

2013). Hyperglycemia induces peroxisome proliferator-activated

receptor-gamma coactivator 1a (PGC-1a), an important regulator of

metabolic gene expression and mitochondrial biogenesis (Puigserver

et al, 1998; Herzig et al, 2001; Lin et al, 2002). PGC1a increases

angiogenesis when expressed in heart and muscle cells (Arany et al,

2008; Patten et al, 2012). In contrast, diabetes-induced PGC-1aexpression in ECs renders them less responsive to angiogenic factors

and blunts angiogenesis (Sawada et al, 2014).

In addition to affecting gene expression, high glucose levels

alter metabolism to induce the production of reactive oxygen

species (ROS) and reactive nitrogen species (RNS), which might be

mediators of EC dysfunction (Fig 3) (Blake & Trounce, 2013).

High glucose levels cause ECs to produce ROS via activation of

NADPH-dependent oxidases (Inoguchi et al, 2003). In addition,

hyperglycemia inhibits PPP flux by down-regulation of G6PD, the

rate-limiting enzyme of the PPP. The PPP is an important source

EMBO Molecular Medicine Vol 6 | No 9 | 2014 ª 2014 The Authors

EMBO Molecular Medicine Targeting EC metabolism in disease Jermaine Goveia et al

1110

of intracellular NADPH, which is necessary to convert oxidized

glutathione (GSSH) into reduced GSH, a critical ROS scavenger

(Fig 3A) (Leopold et al, 2003; Zhang et al, 2012). Therefore, by

reducing PPP flux, high glucose depletes NADPH levels and

contributes to ROS accumulation (Goldin et al, 2006). Interest-

ingly, G6PD overexpression restores redox homeostasis in high

glucose cultured ECs (Leopold et al, 2003; Zhang et al, 2012).

Some studies suggest that high glucose shifts the normally glyco-

lytic EC metabolism toward oxidative metabolism and increased

mitochondrial respiration (Fig 3). However, these results appear

contextual, as other studies did not report such an induction of

oxidative metabolism (Nishikawa et al, 2000; Koziel et al, 2012;

Pangare & Makino, 2012; Dymkowska et al, 2014). While the

precise effects on mitochondrial respiration require further study,

hyperglycemia-induced mitochondrial ROS induces DNA breaks

and thereby activates polyAPD-ribose polymerase (PARP-1) (Du

et al, 2000, 2003; Nishikawa et al, 2000; Giacco & Brownlee, 2010;

Blake & Trounce, 2013). PolyADP-ribosylation by PARP-1 inacti-

vates GAPDH and stalls glycolysis, allowing accumulation of glyco-

lytic metabolites (Du et al, 2003).

Accumulation of F6P increases the flux through the hexosamine

biosynthesis pathway (HBP), which produces UDP-GlcNac, an

important precursor of glycosylation reactions (Fig 3A) (Brownlee,

2001). While glycosylation is important for physiological EC func-

tion, hyperglycemia-induced protein glycosylation inhibits angio-

genic functions (Du et al, 2001; Federici et al, 2002; Luo et al,

2008). Other glycolytic intermediates are diverted into the polyol

and methylglyoxal pathways that produce damaging agents such

as ROS and advanced glycation end products (AGEs) (Fig 3A)

(Goldin et al, 2006). AGEs induce vascular dysfunction by altering

extracellular matrix protein function and dysregulating cytokine

expression (Yan et al, 2008). In addition, receptor of AGE (RAGE)

binding by AGEs in vascular cells causes inflammation and

reduced NO availability associated with vascular complications in

ReducedIncreased

Glucose

Acetyl-CoA

GLUT1

PPP

L-arginine

gnggi

npppupl

uplinin

up

ROS

NADH

ReducedIncreased

Figure 3. Metabolic pathways implicated in diseases characterized by EC dysfunction.(A) High glucose levels in diabetes pushes glycolytic flux and cause ROS production and AGE formation. (B) Metabolic alterations that cause eNOS dysfunction mediateatherosclerosis pathogenesis. Asymmetric dimethylarginine (ADMA) competes with arginine for binding to eNOS. Arginase expression is increased and eNOS expression isdecreased, leading to reduced eNOS activity. 1C metabolism and mevalonate metabolism provide eNOS coupling co-factors and inhibit ROS production. The mevalonatepathway also provides farnesyl pyrophosphate (FPP) and geranylgeranyl pyrophosphate (GGPP), required for GTPase prenylation. For clarity, not all metabolites and enzymesof the depicted pathways are shown. Green font / bold line: upregulated, red font / broken line: downregulated. Abbreviations: as in Figure 1. BH2: dihydrobiopterin; BH4:tetrahydrobiopterin; ADMA: asymmetric dimethylarginine; CoQ10: coenzyme Q10; DDAH: dimethylarginine dimethylaminohydrolase; DHF: dihydrofolate; DHFR:dihydrofolate reductase; FPP: farnesyl pyrophosphate; GGPP: geranylgeranyl pyrophosphate; GTP: Guanosine triphosphate; HMGCR: hydroxymethylglutaryl coenzyme Areductase; PRMT: protein arginine methyltransferase.

ª 2014 The Authors EMBO Molecular Medicine Vol 6 | No 9 | 2014

Jermaine Goveia et al Targeting EC metabolism in disease EMBO Molecular Medicine

1111

lactate induces HIF-1a activation leading to increased expression of

VEGFR2 and bFGF (Sonveaux et al, 2012). Moreover, lactate

competes with a-ketoglutarate for binding to the oxygen sensing

prolyl hydroxylase-2 (PHD-2), resulting in diminished PHD-2 activ-

ity and subsequent hypoxia-inducible factor-1a (HIF-1a) stabiliza-

tion (Fig 2A). Stabilized HIF-1a induces pro-angiogenic signaling

pathways such as nuclear factor kappa-light-chain-enhancer of acti-

vated B-cells (NFkB)/interleukin 8 (IL-8) leading to increased angio-

genesis (Fig 2A) (Hunt et al, 2007; Vegran et al, 2011; Sonveaux

et al, 2012). Exploratory studies found that lactate induces angio-

genesis in vivo and that pharmacological blockade of MCT1 inhibits

angiogenesis and reduces tumor growth in mice (Sonveaux et al,

2012). Together, these data suggest an intricate relationship

between classical pro-angiogenic signals such as VEGF, HIF-1a and

hypoxia, and EC glucose metabolism. Targeting EC glucose metabo-

lism to inhibit tumor angiogenesis is in its infancy as a therapeutic

strategy, but recent evidence suggests its viability.

Pulmonary arterial hypertension

Idiopathic pulmonary arterial hypertension (PAH) is characterized

by heightened pressure in pulmonary arteries caused by excessive

EC proliferation and vascular dysfunction (Xu & Erzurum, 2011).

Emerging evidence indicates that metabolic abnormalities underlie

PAH (Fig 2B) (Sutendra & Michelakis, 2014; Zhao et al, 2014). In

line with recent findings that glycolysis regulates angiogenesis,

hyperproliferative PAH ECs rely on increased glycolytic flux and

reduced oxygen consumption, which may be related to HIF-1aoverexpression (Fig 2B) (Xu et al, 2007; Fijalkowska et al, 2010;

Majmundar et al, 2010; Tuder et al, 2012). Human pulmonary

ECs expressing mutated bone morphogenetic protein receptor 2

(BMPR2), which confers PAH, show altered expression of several

glycolytic enzymes including GLUT1 and phosphoglycerate kinase 1

(PGK1). PAH ECs also show increased expression of enzymes of the

PPP (R5P isomerase, Ru5P-3-epimerase) and polyamine biosynthe-

sis pathway (ornithine decarboxylase (ODC), spermine synthase

(SMS)). These metabolic changes may underlie the rapid prolifera-

tion of PAH ECs, since glycolysis, the PPP and mitogenic polyam-

ines all promote cellular proliferation (Morrison & Seidel, 1995).

However, the expression of other PPP and polyamine enzymes

[G6PD, TKT, spermidine synthase (SRM)] is reduced—a finding

that requires further explanation (Fig 2B) (Atkinson et al, 2002;

Rudarakanchana et al, 2002; Long et al, 2006; Fessel et al, 2012). In

addition, ECs isolated from EC-specific BMPR2 mutant mice show

similarly increased expression of PGK1, indicating altogether that

alterations in glycolysis as well as PPP likely underlie PAH (Majka

et al, 2011).

In addition to alterations in glycolysis, idiopathic PAH ECs have

fewer mitochondria and decreased mitochondrial metabolic activity

(Xu et al, 2007). BMPR2 mutant ECs have reduced quantities of TCA

cycle intermediates, reduced fatty acid oxidation and transcriptional

reduction of several enzymes involved in fatty acid metabolism,

including the rate-limiting enzyme of fatty acid oxidation carnitine

palmitoyltransferase 1 (CPT1) (Fig 2) (Fessel et al, 2012). Together,

these findings suggest reduced oxidative metabolism. Indeed, phar-

macological inhibition of hyper-activated pyruvate dehydrogenase

kinase (PDK), an enzyme that shunts glucose-derived pyruvate away

from oxidative TCA metabolism, has shown therapeutic efficacy.

However, whether these effects are mediated via ECs specifically

remains to be determined (McMurtry et al, 2004). For unexplained

reasons, PAH patients also show increased isocitrate dehydrogenase

(IDH)-1 and IDH-2 serum activity, a finding that corroborates with

the increased IDH activity observed in BPMR2 mutant ECs (Fessel

et al, 2012). Still, the mechanisms that alter metabolic pathways in

PAH ECs and the importance of some of these metabolic adaptations

in the pathogenesis of PAH remain unclear.

Reduced nitric oxide (NO) levels are another hallmark of PAH

ECs (Fijalkowska et al, 2010). Low NO levels may be related to the

reduced levels of the mitochondrial antioxidant manganese superox-

ide dismutase (MnSOD) (Fijalkowska et al, 2010). Indeed, MnSOD

increases NO availability by clearing superoxide anion, which inac-

tivates NO to form peroxynitrite (Fig 2) (Masri et al, 2008).

However, other factors likely contribute to the low NO levels in

PAH ECs (Xu et al, 2004). Indeed, human PAH ECs express high

levels of arginase II, which competes with endothelial nitric oxide

synthetase (eNOS) for their common substrate L-arginine (Fig 2)

(Xu et al, 2004). Inhibition of endothelial arginase II increases NO

production in vitro, suggesting that arginase II can be targeted to

prevent EC hyperproliferation and restore NO availability (Krotova

et al, 2010). While the mechanisms that induce abnormal metabolic

activity in PAH ECs are understudied, restoring NO may provide

dual benefits in preventing excessive EC proliferation as well as

restoring EC vasoactivity.

The metabolic adaptations in PAH (high glycolytic rates and

reduced oxidative metabolism) are partly reminiscent of the meta-

bolic profile of angiogenic ECs. It would be thus interesting to deter-

mine if reducing glycolysis by pharmacological blockade of PFKFB3

can reduce the hyperproliferative rate in PAH ECs. Alternatively, the

beneficial effects of PDK inhibition in PAH to induce oxidative

metabolism could also be beneficial to block angiogenesis by

preventing the glycolytic switch in ECs. Indeed, PDK blockade

with dichloroacetate inhibits angiogenesis in glioblastoma patients

(Michelakis et al, 2010).

EC metabolism in diseases characterized byEC dysfunction

Diabetes

Diabetes is characterized by high blood glucose levels that affect EC

metabolism and cause dysfunction (Fig 3A) (Blake & Trounce,

2013). Hyperglycemia induces peroxisome proliferator-activated

receptor-gamma coactivator 1a (PGC-1a), an important regulator of

metabolic gene expression and mitochondrial biogenesis (Puigserver

et al, 1998; Herzig et al, 2001; Lin et al, 2002). PGC1a increases

angiogenesis when expressed in heart and muscle cells (Arany et al,

2008; Patten et al, 2012). In contrast, diabetes-induced PGC-1aexpression in ECs renders them less responsive to angiogenic factors

and blunts angiogenesis (Sawada et al, 2014).

In addition to affecting gene expression, high glucose levels

alter metabolism to induce the production of reactive oxygen

species (ROS) and reactive nitrogen species (RNS), which might be

mediators of EC dysfunction (Fig 3) (Blake & Trounce, 2013).

High glucose levels cause ECs to produce ROS via activation of

NADPH-dependent oxidases (Inoguchi et al, 2003). In addition,

hyperglycemia inhibits PPP flux by down-regulation of G6PD, the

rate-limiting enzyme of the PPP. The PPP is an important source

EMBO Molecular Medicine Vol 6 | No 9 | 2014 ª 2014 The Authors

EMBO Molecular Medicine Targeting EC metabolism in disease Jermaine Goveia et al

1110

atherosclerosis and promotes ROS production through eNOS

uncoupling (Fig 3B) (Pieper, 1997; Stroes et al, 1997; Heitzer et al,

2000). Endothelial BH4 levels are maintained by de novo biosyn-

thesis via the rate-limiting enzyme guanosine triphosphate

cyclohydrolase I (GTPCH) and by a salvage pathway from dihydro-

biopterin (BH2) via dihydrofolate reductase (DHFR) (Fig 3B)

(Bendall et al, 2014). Insufficient levels of GTPCH and DHFR,

important enzymes in GTP and folate metabolism, respectively, have

been associated with reduced BH4 availability, endothelial dysfunc-

tion and cardiovascular disease in several preclinical models

(Chalupsky & Cai, 2005; Crabtree et al, 2009b, 2011; Sugiyama

et al, 2009; Kidokoro et al, 2013). Interestingly, DHFR not only

regenerates active BH4 from oxidized inactive BH2 but is also a

key enzyme in folate and one-carbon metabolism, intermediates of

which in turn regulate BH4 biosynthesis and are associated with

cardiovascular disease (Humphrey et al, 2008).

One-carbon (1C) metabolism centers around the ability of

folate-derived co-enzymes to carry activated 1C units (Fig 3)

(Tibbetts & Appling, 2010). DHFR catalyzes the formation of

tetrahydrofolate (THF) from folate fueling 1C metabolism. THF

accepts 1C units from serine to produce 5,10-methylene-THF

(meTHF) and glycine. MeTHF is reduced to 5-methyl-THF (mTHF)

by methylenetetrahydrofolate reductase (MTHFR) (Fig 3).

Importantly, inactivating mutations in the MTHFR gene result in

hyperhomocysteinemia, which decreases GTPCH and DHFR levels

and may subsequently reduce BH4 levels (Bendall et al, 2014).

Indeed, MTHFR mutations have been associated with cardiovascu-

lar disease, but the exact association is still controversial (Kelly

et al, 2002; Klerk et al, 2002; Frederiksen et al, 2004; Yang et al,

2012). mTHF produced by MTHFR activity is required as a methyl

donor in the methionine synthase (MS) catalyzed reaction that

converts mTHF into THF (completing the folate cycle) and forms

methionine (MET) from homocysteine (hCYS) (Fig 3B) (Locasale,

2013). Methionine is used to generate S-adenosylmethionine

(SAM), which is an important methyl donor and plays a pivotal

role in methylation of lysine and arginine residues in proteins

(Fig 3B) (Leiper & Nandi, 2011). As discussed above, methylated

arginine residues are emerging as important mediators of EC

dysfunction. Moreover, SAM-mediated protein methylation

produces S-adenosylhomocysteine, which is converted back into

homocysteine. Homocysteine decreases the bioavailability of BH4

possibly through downregulation of GTPCH and DHFR, while BH4

supplementation alleviates homocysteine-induced EC dysfunction

(Dhillon et al, 2003; Topal et al, 2004). Together, these findings

suggest that dysregulation of endothelial 1C metabolism is

involved in the pathogenesis of cardiovascular disease, but the

VE

GG

VE

GG

VE

GG FGG

VE

GG

FGF

FGG

FGF

Anti-VEGFtreatment

Anti-VEGFtreatment

Angiogenic activityen a

ReducedIncreasedn

Figure 4. Targeting EC metabolism as an alternative to targeting growth factors in angiogenesis.(A) Vascular endothelial growth factor (VEGF) induces 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase-3 (PFKFB3) and increases glycolytic flux, required forangiogenesis. (B) Anti-VEGF treatment reduces glycolytic flux and angiogenesis. (C) Increased growth factor signaling through alternative pathways, in this case fibroblastgrowth factor (FGF), mediates resistance to anti-angiogenic therapy. (D) Pharmacological targeting of PFKFB3 with (3PO) reduces angiogenesis irrespective of growth factorsignaling and is therefore possibly less prone to resistance. Abbreviations: as in Figure 1. 3PO: 3-(3-pyridinyl)-1-(4-pyridinyl)-2-propen-1-one; FGF: fibroblast growth factor.

ª 2014 The Authors EMBO Molecular Medicine Vol 6 | No 9 | 2014

Jermaine Goveia et al Targeting EC metabolism in disease EMBO Molecular Medicine

1113

diabetic patients (Bucala et al, 1991; Vlassara et al, 1995; Min

et al, 1999; Wautier & Schmidt, 2004; Goldin et al, 2006; Manigrasso

et al, 2014).

Excess glucose that cannot be metabolized by glycolysis enters

the polyol pathway when converted into sorbitol by aldose reduc-

tase (AR) at the expense of NADPH, increasing ROS. Sorbitol

is subsequently converted into fructose and the highly reactive

3-deoxyglucosone (3DG), which promotes the formation of AGEs

(Fig 3A) (Kashiwagi et al, 1994; Oyama et al, 2006; Giacco &

Brownlee, 2010; Sena et al, 2012; Yoshida et al, 2012). Transgenic

overexpression of human AR in the endothelium of diabetic mice

accelerates atherosclerosis formation and inhibition of endothelial

AR reduces intracellular ROS, EC migration and proliferation

(Obrosova et al, 2003; Tammali et al, 2011; Vedantham et al, 2011;

Yadav et al, 2012). Methylglyoxal is another AGE precursor and

produced from the glycolytic intermediates glyceraldehyde-3-phosphate

(G3P) and dihydroxyacetone phosphate (DHAP). Methylglyoxal is

detoxified by conversion into pyruvate via the multienzyme

glyoxalase system, of which glyoxalase-I (GloI) is rate-limiting

(Fig 3A) (Thornalley, 1993). Glyoxalase-I overexpression reverses

hyperglycemia-induced angiogenesis defects in vitro and transgenic

overexpression of glyoxalase-I in rats reduces vascular AGE

formation and improves vasoreactivity (Brouwers et al, 2010, 2014)

(Ahmed et al, 2008). Together, these observations indicate that

targeting AR and glyoxalase might confer a therapeutic benefit in

diabetic patients.

Atherosclerosis

Atherosclerosis is a chronic inflammatory process in the blood

vessel wall leading to luminal narrowing and subsequent cardio-

vascular events (Hopkins, 2013). Systemic metabolic perturbations

are among the most important risk factors of atherosclerosis.

However, metabolic flux changes have not been studied in ECs

isolated from atherosclerotic lesions, and the effects of atheroscle-

rosis on central metabolism of ECs thus remains to be character-

ized. Nonetheless, EC metabolism is strongly associated with a key

pathophysiological feature of atherosclerosis: reduced and uncou-

pled eNOS activity resulting in low NO bioavailability and high

ROS production (Fig 3B) (Kawashima & Yokoyama, 2004). eNOS

activity critically depends on the availability of L-arginine,

co-factor tetrahydrobiopterin (BH4) (Fig 3B) and possibly co-

enzyme Q10 (CoQ10) (Gorren et al, 2000; Crabtree et al, 2009a;

Mugoni et al, 2013). If L-arginine, BH4 or CoQ10 become limited,

eNOS no longer oxidizes L-arginine to form citrulline and NO, but

instead produces ROS (a condition termed eNOS uncoupling)

(Fig 3B) (Stroes et al, 1998; Mugoni et al, 2013). Targeting L-arginine

and BH4 metabolism to increase eNOS activity in patients with

cardiovascular disease is potentially beneficial, but available

evidence indicates that the picture is more complex than initially

anticipated.

Small-scale clinical trials indicate that administration of L-arginine

to patients with coronary heart disease improves vasoresponsive-

ness, possibly by increasing NO production by eNOS (Lerman et al,

1998). Interestingly, however, intracellular and plasma arginine

levels are sufficiently high to support NO biosynthesis via eNOS.

Therefore, the benefits of L-arginine supplementation on

elevating NO levels are not readily explained by increasing the

supply of L-arginine; however, it is possible that L-arginine is

compartmentalized in poorly interchangeable pools. Another possi-

ble explanation of the beneficial effects of L-arginine is competition

with asymmetric methylated arginines, which bind and inhibit

eNOS (Fig 3B) (Boger, 2004; Chen et al, 2013). More in detail, post-

translational methylation of arginine residues in proteins by protein

arginine methyltransferase (PRMT) results in the addition of up to

two methyl groups to arginine. Protein turnover releases these

post-translationally modified amino acids as asymmetric dimethyl-

arginine (ADMA) and symmetric dimethylarginine (SDMA). The

asymmetric dimethylarginines bind and uncouple eNOS resulting in

increased ROS production and reduced NO availability (Fig 3B)

(Dhillon et al, 2003; Leiper & Nandi, 2011). Hence by competing

with ADMAs, supplemented L-arginine could maintain eNOS activ-

ity to produce NO (Bode-Boger et al, 2003). Additional potential

interventions to reduce eNOS inhibition by ADMA include PRMT

inhibition (to reduce arginine methylation) and activation of methyl-

arginine catabolism by dimethylarginine dimethylaminohydrolase

(DDAH) (Fig 3B) (Leiper & Nandi, 2011). Interestingly, DDAH1 is

predominantly expressed in ECs and EC-specific deletion attenuates

NO production and induces hypertension, indicating that ADMA

scavenging by ECs is important to maintain homeostasis (Hu et al,

2009).

Because L-arginine is a substrate for both eNOS and arginase

(Wu & Meininger, 1995), NO production depends on the relative

expression levels of each enzyme (Fig 3) (Chang et al, 1998; Ming

et al, 2004; Ryoo et al, 2008). Endothelial arginase expression is

induced by many risk factors for cardiovascular disease, while

reducing arginase expression restores NO production in vitro

and improves vasodilatation in vivo (Ryoo et al, 2006, 2008;

Thengchaisri et al, 2006; Romero et al, 2008). The activity of

eNOS and arginase is regulated by the RhoA/ROCK signaling

cascade. RhoA and Rock decrease eNOS expression, while RhoA

also increases arginase activity (Fig 3B) (Laufs et al, 1998;

Takemoto et al, 2002). For proper activation and localization to

the cell membrane, RhoA must be prenylated (more specifically,

geranylgeranylated) by geranylgeranyltransferase (GGT) using

geranylgeranyl pyrophosphate (GGPP) as a substrate (Laufs &

Liao, 1998). This isoprenoid is an intermediate of the mevalonate

pathway, which produces cholesterol from acetyl-coA (Fig 3B).

Blocking the mevalonate pathway by inhibiting HMG-coA reduc-

tase using statins lowers cholesterol synthesis and is clinically

approved to prevent cardiovascular events in dyslipidemia

patients. In addition, HMG-coA blockade also decreases geranyl-

geranyl production, which reduces RhoA activity and restores a

more beneficial eNOS/arginase balance (Goldstein & Brown, 1990;

Liao & Laufs, 2005). Interestingly, UBIAD1 was recently identified

as a novel prenyltransferase that produces non-mitochondrial

CoQ10 from farnesyl pyrophosphate (FPP), another isoprenoid

produced in the mevalonate pathway (Fig 3) (Mugoni et al, 2013).

CoQ10 is an important anti-oxidant with beneficial effects on EC

function and hypothesized to be a novel co-factor required for

eNOS coupling (Gao et al, 2012; Mugoni et al, 2013). Hence, in

contrast to the above-mentioned beneficial effects, HMG-coA

reductase inhibition might thus also have a less favorable effect by

increasing ROS levels through reducing CoQ10 synthesis (Fig 3)

(Mugoni et al, 2013).

In addition to CoQ10, eNOS requires BH4 as a co-factor.

Reduced BH4 availability is found in patients at risk of

EMBO Molecular Medicine Vol 6 | No 9 | 2014 ª 2014 The Authors

EMBO Molecular Medicine Targeting EC metabolism in disease Jermaine Goveia et al

1112

atherosclerosis and promotes ROS production through eNOS

uncoupling (Fig 3B) (Pieper, 1997; Stroes et al, 1997; Heitzer et al,

2000). Endothelial BH4 levels are maintained by de novo biosyn-

thesis via the rate-limiting enzyme guanosine triphosphate

cyclohydrolase I (GTPCH) and by a salvage pathway from dihydro-

biopterin (BH2) via dihydrofolate reductase (DHFR) (Fig 3B)

(Bendall et al, 2014). Insufficient levels of GTPCH and DHFR,

important enzymes in GTP and folate metabolism, respectively, have

been associated with reduced BH4 availability, endothelial dysfunc-

tion and cardiovascular disease in several preclinical models

(Chalupsky & Cai, 2005; Crabtree et al, 2009b, 2011; Sugiyama

et al, 2009; Kidokoro et al, 2013). Interestingly, DHFR not only

regenerates active BH4 from oxidized inactive BH2 but is also a

key enzyme in folate and one-carbon metabolism, intermediates of

which in turn regulate BH4 biosynthesis and are associated with

cardiovascular disease (Humphrey et al, 2008).

One-carbon (1C) metabolism centers around the ability of

folate-derived co-enzymes to carry activated 1C units (Fig 3)

(Tibbetts & Appling, 2010). DHFR catalyzes the formation of

tetrahydrofolate (THF) from folate fueling 1C metabolism. THF

accepts 1C units from serine to produce 5,10-methylene-THF

(meTHF) and glycine. MeTHF is reduced to 5-methyl-THF (mTHF)

by methylenetetrahydrofolate reductase (MTHFR) (Fig 3).

Importantly, inactivating mutations in the MTHFR gene result in

hyperhomocysteinemia, which decreases GTPCH and DHFR levels

and may subsequently reduce BH4 levels (Bendall et al, 2014).

Indeed, MTHFR mutations have been associated with cardiovascu-

lar disease, but the exact association is still controversial (Kelly

et al, 2002; Klerk et al, 2002; Frederiksen et al, 2004; Yang et al,

2012). mTHF produced by MTHFR activity is required as a methyl

donor in the methionine synthase (MS) catalyzed reaction that

converts mTHF into THF (completing the folate cycle) and forms

methionine (MET) from homocysteine (hCYS) (Fig 3B) (Locasale,

2013). Methionine is used to generate S-adenosylmethionine

(SAM), which is an important methyl donor and plays a pivotal

role in methylation of lysine and arginine residues in proteins

(Fig 3B) (Leiper & Nandi, 2011). As discussed above, methylated

arginine residues are emerging as important mediators of EC

dysfunction. Moreover, SAM-mediated protein methylation

produces S-adenosylhomocysteine, which is converted back into

homocysteine. Homocysteine decreases the bioavailability of BH4

possibly through downregulation of GTPCH and DHFR, while BH4

supplementation alleviates homocysteine-induced EC dysfunction

(Dhillon et al, 2003; Topal et al, 2004). Together, these findings

suggest that dysregulation of endothelial 1C metabolism is

involved in the pathogenesis of cardiovascular disease, but the

VE

GG

VE

GG

VE

GG FGG

VE

GG

FGF

FGG

FGF

Anti-VEGFtreatment

Anti-VEGFtreatment

Angiogenic activityen a

ReducedIncreasedn

Figure 4. Targeting EC metabolism as an alternative to targeting growth factors in angiogenesis.(A) Vascular endothelial growth factor (VEGF) induces 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase-3 (PFKFB3) and increases glycolytic flux, required forangiogenesis. (B) Anti-VEGF treatment reduces glycolytic flux and angiogenesis. (C) Increased growth factor signaling through alternative pathways, in this case fibroblastgrowth factor (FGF), mediates resistance to anti-angiogenic therapy. (D) Pharmacological targeting of PFKFB3 with (3PO) reduces angiogenesis irrespective of growth factorsignaling and is therefore possibly less prone to resistance. Abbreviations: as in Figure 1. 3PO: 3-(3-pyridinyl)-1-(4-pyridinyl)-2-propen-1-one; FGF: fibroblast growth factor.

ª 2014 The Authors EMBO Molecular Medicine Vol 6 | No 9 | 2014

Jermaine Goveia et al Targeting EC metabolism in disease EMBO Molecular Medicine

1113

diabetic patients (Bucala et al, 1991; Vlassara et al, 1995; Min

et al, 1999; Wautier & Schmidt, 2004; Goldin et al, 2006; Manigrasso

et al, 2014).

Excess glucose that cannot be metabolized by glycolysis enters

the polyol pathway when converted into sorbitol by aldose reduc-

tase (AR) at the expense of NADPH, increasing ROS. Sorbitol

is subsequently converted into fructose and the highly reactive

3-deoxyglucosone (3DG), which promotes the formation of AGEs

(Fig 3A) (Kashiwagi et al, 1994; Oyama et al, 2006; Giacco &

Brownlee, 2010; Sena et al, 2012; Yoshida et al, 2012). Transgenic

overexpression of human AR in the endothelium of diabetic mice

accelerates atherosclerosis formation and inhibition of endothelial

AR reduces intracellular ROS, EC migration and proliferation

(Obrosova et al, 2003; Tammali et al, 2011; Vedantham et al, 2011;

Yadav et al, 2012). Methylglyoxal is another AGE precursor and

produced from the glycolytic intermediates glyceraldehyde-3-phosphate

(G3P) and dihydroxyacetone phosphate (DHAP). Methylglyoxal is

detoxified by conversion into pyruvate via the multienzyme

glyoxalase system, of which glyoxalase-I (GloI) is rate-limiting

(Fig 3A) (Thornalley, 1993). Glyoxalase-I overexpression reverses

hyperglycemia-induced angiogenesis defects in vitro and transgenic

overexpression of glyoxalase-I in rats reduces vascular AGE

formation and improves vasoreactivity (Brouwers et al, 2010, 2014)

(Ahmed et al, 2008). Together, these observations indicate that

targeting AR and glyoxalase might confer a therapeutic benefit in

diabetic patients.

Atherosclerosis

Atherosclerosis is a chronic inflammatory process in the blood

vessel wall leading to luminal narrowing and subsequent cardio-

vascular events (Hopkins, 2013). Systemic metabolic perturbations

are among the most important risk factors of atherosclerosis.

However, metabolic flux changes have not been studied in ECs

isolated from atherosclerotic lesions, and the effects of atheroscle-

rosis on central metabolism of ECs thus remains to be character-

ized. Nonetheless, EC metabolism is strongly associated with a key

pathophysiological feature of atherosclerosis: reduced and uncou-

pled eNOS activity resulting in low NO bioavailability and high

ROS production (Fig 3B) (Kawashima & Yokoyama, 2004). eNOS

activity critically depends on the availability of L-arginine,

co-factor tetrahydrobiopterin (BH4) (Fig 3B) and possibly co-

enzyme Q10 (CoQ10) (Gorren et al, 2000; Crabtree et al, 2009a;

Mugoni et al, 2013). If L-arginine, BH4 or CoQ10 become limited,

eNOS no longer oxidizes L-arginine to form citrulline and NO, but

instead produces ROS (a condition termed eNOS uncoupling)

(Fig 3B) (Stroes et al, 1998; Mugoni et al, 2013). Targeting L-arginine

and BH4 metabolism to increase eNOS activity in patients with

cardiovascular disease is potentially beneficial, but available

evidence indicates that the picture is more complex than initially

anticipated.

Small-scale clinical trials indicate that administration of L-arginine

to patients with coronary heart disease improves vasoresponsive-

ness, possibly by increasing NO production by eNOS (Lerman et al,

1998). Interestingly, however, intracellular and plasma arginine

levels are sufficiently high to support NO biosynthesis via eNOS.

Therefore, the benefits of L-arginine supplementation on

elevating NO levels are not readily explained by increasing the

supply of L-arginine; however, it is possible that L-arginine is

compartmentalized in poorly interchangeable pools. Another possi-

ble explanation of the beneficial effects of L-arginine is competition

with asymmetric methylated arginines, which bind and inhibit

eNOS (Fig 3B) (Boger, 2004; Chen et al, 2013). More in detail, post-

translational methylation of arginine residues in proteins by protein

arginine methyltransferase (PRMT) results in the addition of up to

two methyl groups to arginine. Protein turnover releases these

post-translationally modified amino acids as asymmetric dimethyl-

arginine (ADMA) and symmetric dimethylarginine (SDMA). The

asymmetric dimethylarginines bind and uncouple eNOS resulting in

increased ROS production and reduced NO availability (Fig 3B)

(Dhillon et al, 2003; Leiper & Nandi, 2011). Hence by competing

with ADMAs, supplemented L-arginine could maintain eNOS activ-

ity to produce NO (Bode-Boger et al, 2003). Additional potential

interventions to reduce eNOS inhibition by ADMA include PRMT

inhibition (to reduce arginine methylation) and activation of methyl-

arginine catabolism by dimethylarginine dimethylaminohydrolase

(DDAH) (Fig 3B) (Leiper & Nandi, 2011). Interestingly, DDAH1 is

predominantly expressed in ECs and EC-specific deletion attenuates

NO production and induces hypertension, indicating that ADMA

scavenging by ECs is important to maintain homeostasis (Hu et al,

2009).

Because L-arginine is a substrate for both eNOS and arginase

(Wu & Meininger, 1995), NO production depends on the relative

expression levels of each enzyme (Fig 3) (Chang et al, 1998; Ming

et al, 2004; Ryoo et al, 2008). Endothelial arginase expression is

induced by many risk factors for cardiovascular disease, while

reducing arginase expression restores NO production in vitro

and improves vasodilatation in vivo (Ryoo et al, 2006, 2008;

Thengchaisri et al, 2006; Romero et al, 2008). The activity of

eNOS and arginase is regulated by the RhoA/ROCK signaling

cascade. RhoA and Rock decrease eNOS expression, while RhoA

also increases arginase activity (Fig 3B) (Laufs et al, 1998;

Takemoto et al, 2002). For proper activation and localization to

the cell membrane, RhoA must be prenylated (more specifically,

geranylgeranylated) by geranylgeranyltransferase (GGT) using

geranylgeranyl pyrophosphate (GGPP) as a substrate (Laufs &

Liao, 1998). This isoprenoid is an intermediate of the mevalonate

pathway, which produces cholesterol from acetyl-coA (Fig 3B).

Blocking the mevalonate pathway by inhibiting HMG-coA reduc-

tase using statins lowers cholesterol synthesis and is clinically

approved to prevent cardiovascular events in dyslipidemia

patients. In addition, HMG-coA blockade also decreases geranyl-

geranyl production, which reduces RhoA activity and restores a

more beneficial eNOS/arginase balance (Goldstein & Brown, 1990;

Liao & Laufs, 2005). Interestingly, UBIAD1 was recently identified

as a novel prenyltransferase that produces non-mitochondrial

CoQ10 from farnesyl pyrophosphate (FPP), another isoprenoid

produced in the mevalonate pathway (Fig 3) (Mugoni et al, 2013).

CoQ10 is an important anti-oxidant with beneficial effects on EC

function and hypothesized to be a novel co-factor required for

eNOS coupling (Gao et al, 2012; Mugoni et al, 2013). Hence, in

contrast to the above-mentioned beneficial effects, HMG-coA

reductase inhibition might thus also have a less favorable effect by

increasing ROS levels through reducing CoQ10 synthesis (Fig 3)

(Mugoni et al, 2013).

In addition to CoQ10, eNOS requires BH4 as a co-factor.

Reduced BH4 availability is found in patients at risk of

EMBO Molecular Medicine Vol 6 | No 9 | 2014 ª 2014 The Authors

EMBO Molecular Medicine Targeting EC metabolism in disease Jermaine Goveia et al

1112

Conflict of interestThe authors declare that they have no conflict of interest.

References

Ahmed U, Dobler D, Larkin SJ, Rabbani N, Thornalley PJ (2008) Reversal of

hyperglycemia-induced angiogenesis deficit of human endothelial cells

by overexpression of glyoxalase 1 in vitro. Ann N Y Acad Sci 1126:

262 – 264

Arany Z, Foo SY, Ma Y, Ruas JL, Bommi-Reddy A, Girnun G, Cooper M, Laznik

D, Chinsomboon J, Rangwala SM et al (2008) HIF-independent regulation

of VEGF and angiogenesis by the transcriptional coactivator PGC-1alpha.

Nature 451: 1008 – 1012

Atkinson C, Stewart S, Upton PD, Machado R, Thomson JR, Trembath RC,

Morrell NW (2002) Primary pulmonary hypertension is associated with

reduced pulmonary vascular expression of type II bone morphogenetic

protein receptor. Circulation 105: 1672 – 1678

van Beijnum JR, Dings RP, van der Linden E, Zwaans BM, Ramaekers FC,

Mayo KH, Griffioen AW (2006) Gene expression of tumor angiogenesis

dissected: specific targeting of colon cancer angiogenic vasculature. Blood

108: 2339 – 2348

Bendall JK, Douglas G, McNeill E, Channon KM, Crabtree MJ (2014)

Tetrahydrobiopterin in cardiovascular health and disease. Antioxid Redox

Signal 20: 3040 – 3077

Benedito R, Roca C, Sorensen I, Adams S, Gossler A, Fruttiger M, Adams RH

(2009) The notch ligands Dll4 and Jagged1 have opposing effects on

angiogenesis. Cell 137: 1124 – 1135

Bergers G, Benjamin LE (2003) Tumorigenesis and the angiogenic switch. Nat

Rev 3: 401 – 410

Bergers G, Hanahan D (2008) Modes of resistance to anti-angiogenic therapy.

Nat Rev 8: 592 – 603

Blake R, Trounce IA (2013) Mitochondrial dysfunction and complications

associated with diabetes. Biochim Biophys Acta 1840: 1404 – 1412

Bode-Boger SM, Muke J, Surdacki A, Brabant G, Boger RH, Frolich JC (2003)

Oral L-arginine improves endothelial function in healthy individuals older

than 70 years. Vas Med 8: 77 – 81

Boger RH, Sydow K, Borlak J, Thum T, Lenzen H, Schubert B, Tsikas D,

Bode-Boger SM (2000) LDL cholesterol upregulates synthesis of

asymmetrical dimethylarginine in human endothelial cells: involvement of

S-adenosylmethionine-dependent methyltransferases. Circ Res 87: 99 – 105

Boger RH (2004) Asymmetric dimethylarginine, an endogenous inhibitor of

nitric oxide synthase, explains the “L-arginine paradox” and acts as a novel

cardiovascular risk factor. J Nutr 134: 2842S – 2847S; discussion 2853S

Brouwers O, Niessen PM, Haenen G, Miyata T, Brownlee M, Stehouwer CD,

De Mey JG, Schalkwijk CG (2010) Hyperglycaemia-induced impairment of

endothelium-dependent vasorelaxation in rat mesenteric arteries is

mediated by intracellular methylglyoxal levels in a pathway dependent on

oxidative stress. Diabetologia 53: 989 – 1000

Brouwers O, Niessen PM, Miyata T, Ostergaard JA, Flyvbjerg A, Peutz-Kootstra

CJ, Sieber J, Mundel PH, Brownlee M, Janssen BJ et al (2014) Glyoxalase-1

overexpression reduces endothelial dysfunction and attenuates early renal

impairment in a rat model of diabetes. Diabetologia 57: 224 – 235

Brownlee M (2001) Biochemistry and molecular cell biology of diabetic

complications. Nature 414: 813 – 820

Bucala R, Tracey KJ, Cerami A (1991) Advanced glycosylation products quench

nitric oxide and mediate defective endothelium-dependent vasodilatation

in experimental diabetes. J Clin Investig 87: 432 – 438

Carmeliet P (2003) Angiogenesis in health and disease. Nat Med 9: 653 – 660

Carmeliet P, Jain RK (2011) Molecular mechanisms and clinical applications

of angiogenesis. Nature 473: 298 – 307

Chalupsky K, Cai H (2005) Endothelial dihydrofolate reductase: critical for

nitric oxide bioavailability and role in angiotensin II uncoupling of

endothelial nitric oxide synthase. Proc Natl Acad Sci USA 102: 9056 – 9061

Chang CI, Liao JC, Kuo L (1998) Arginase modulates nitric oxide production in

activated macrophages. Am J Physiol 274: H342 –H348

Chen F, Lucas R, Fulton D (2013) The subcellular compartmentalization of

arginine metabolizing enzymes and their role in endothelial dysfunction.

Front Immunol 4: 184

Cines DB, Pollak ES, Buck CA, Loscalzo J, Zimmerman GA, McEver RP, Pober

JS, Wick TM, Konkle BA, Schwartz BS et al (1998) Endothelial cells in

physiology and in the pathophysiology of vascular disorders. Blood 91:

3527 – 3561

Clarke R, Halsey J, Lewington S, Lonn E, Armitage J, Manson JE, Bonaa KH,

Spence JD, Nygard O, Jamison R et al (2010) Effects of lowering

homocysteine levels with B vitamins on cardiovascular disease, cancer,

and cause-specific mortality: Meta-analysis of 8 randomized trials

involving 37 485 individuals. Arch Intern Med 170: 1622 – 1631

Crabtree MJ, Hale AB, Channon KM (2011) Dihydrofolate reductase protects

endothelial nitric oxide synthase from uncoupling in tetrahydrobiopterin

deficiency. Free Radical Biol Med 50: 1639 – 1646

Crabtree MJ, Tatham AL, Al-Wakeel Y, Warrick N, Hale AB, Cai S, Channon

KM, Alp NJ (2009a) Quantitative regulation of intracellular endothelial

nitric-oxide synthase (eNOS) coupling by both tetrahydrobiopterin-eNOS

stoichiometry and biopterin redox status: insights from cells with

tet-regulated GTP cyclohydrolase I expression. J Biol Chem 284:

1136 – 1144

Crabtree MJ, Tatham AL, Hale AB, Alp NJ, Channon KM (2009b) Critical role

for tetrahydrobiopterin recycling by dihydrofolate reductase in regulation

of endothelial nitric-oxide synthase coupling: relative importance of the

de novo biopterin synthesis versus salvage pathways. J Biol Chem 284:

28128 – 28136

Croci DO, Cerliani JP, Dalotto-Moreno T, Mendez-Huergo SP, Mascanfroni ID,

Dergan-Dylon S, Toscano MA, Caramelo JJ, Garcia-Vallejo JJ, Ouyang J et al

(2014) Glycosylation-dependent lectin-receptor interactions preserve

angiogenesis in anti-VEGF refractory tumors. Cell 156: 744 – 758

Culic O, Gruwel ML, Schrader J (1997) Energy turnover of vascular endothelial

cells. Am J Physiol 273: C205 –C213

Cunnington C, Van Assche T, Shirodaria C, Kylintireas I, Lindsay AC, Lee JM,

Antoniades C, Margaritis M, Lee R, Cerrato R et al (2012) Systemic and

vascular oxidation limits the efficacy of oral tetrahydrobiopterin treatment

in patients with coronary artery disease. Circulation 125: 1356 – 1366

Dagher Z, Ruderman N, Tornheim K, Ido Y (1999) The effect of AMP-activated

protein kinase and its activator AICAR on the metabolism of human

umbilical vein endothelial cells. Biochem Biophys Res Commun 265:

112 – 115

Dagher Z, Ruderman N, Tornheim K, Ido Y (2001) Acute regulation of fatty

acid oxidation and amp-activated protein kinase in human umbilical vein

endothelial cells. Circ Res 88: 1276 – 1282

Davignon J, Ganz P (2004) Role of endothelial dysfunction in atherosclerosis.

Circulation 109: III27 – III32

De Bock K, Georgiadou M, Carmeliet P (2013a) Role of endothelial cell

metabolism in vessel sprouting. Cell Metab 18: 634 – 647

De Bock K, Georgiadou M, Schoors S, Kuchnio A, Wong BW, Cantelmo AR,

Quaegebeur A, Ghesquiere B, Cauwenberghs S, Eelen G et al (2013b) Role

of PFKFB3-driven glycolysis in vessel sprouting. Cell 154: 651 – 663

ª 2014 The Authors EMBO Molecular Medicine Vol 6 | No 9 | 2014

Jermaine Goveia et al Targeting EC metabolism in disease EMBO Molecular Medicine

1115

exact mechanisms remain to be elucidated. Nonetheless, early

clinical and preclinical studies have found that therapeutic target-

ing of 1C metabolism, for example, via folate supplementation

lowers levels of homocysteinemia and increases BH4 regeneration

from BH2 (Verhaar et al, 2002). However, large-scale clinical trials

failed to show benefits of folate or BH4 supplementation to

prevent cardiovascular disease (Clarke et al, 2010; Cunnington

et al, 2012; Marti-Carvajal et al, 2013). These clinical and preclini-

cal findings suggest that while L-arginine, folate, methionine,

COQ10 and homocysteine metabolism are potential therapeutic

targets, a more detailed understanding of how these pathways

cause dysfunction is required to design more rational therapeutic

agents.

EC metabolism in the pathogenesis of other diseases

EC metabolism is best characterized in the diseases discussed

above. However, these represent only a minor fraction of the disor-

ders in which pathological EC responses are presumably involved.

Indeed, it is highly likely that EC metabolic alterations are also

involved in the pathogenesis of other diseases such as ischemia,

pre-eclampsia, vasculitis, vascular neoplasms and others although

this has hardly been studied.

On the other hand, many of the EC metabolic alterations that

lead to EC dysfunction are likely induced by cardiovascular risk

factors such as those that characterize metabolic syndrome, hyper-

homocysteinemia and hyperuricemia. For example, elevated serum

uric acid (a breakdown product of purine nucleotides generated by

xanthine oxidase with potent anti-oxidant activity) is common in

patients with hypertension and may even be a root cause of EC

dysfunction leading to cardiovascular disease (Feig et al, 2008).

Interestingly, while uric acid has been described as major anti-

oxidant in human plasma, ECs exposed to uric acid display

increased ROS production creating a paradox that has not been

resolved (Lippi et al, 2008; Sautin & Johnson, 2008). Regardless, in

cardiovascular disease models uric acid reduces mitochondrial

content, intracellular ATP and arginase activity (Zharikov et al,

2008; Sanchez-Lozada et al, 2012). In addition, uric acid inhibits

NO production in ECs in vitro, and in vivo levels of serum nitrites

(an indicator of NO production) are inversely proportional to serum

uric acid concentrations (Khosla et al, 2005). Interestingly, ECs

exposed to uric acid increase expression of AR and alter expression

of several other proteins linked to metabolism (Zhang et al, 2014).

These studies suggest that hyperuricemia induces EC dysfunction

through metabolic alterations. Whether the same is true for other

cardiovascular risk factors remains in question.

A broader characterization of EC metabolism in the future might

reveal novel therapeutic targets in metabolic pathways that are

generally not considered to be important in pathological EC func-

tion. Recent findings that endothelial cholesterol efflux to high-

density lipoprotein regulates angiogenesis (Fang et al, 2013), and

that EC-specific insulin receptor knock-out accelerates atheroscle-

rotic plaque formation (Gage et al, 2013) point to a key role for EC

metabolism in the pathogenesis of disease and indicate that many

more yet to be identified non-traditional but potentially druggable

metabolic enzymes, transporters and pathways may play a role in

vascular disease.

Therapeutic targeting of EC metabolism

Overall, it is clear that pathological blood vessel responses are

associated with metabolic alterations in ECs. These metabolic

adaptations are not just innocent bystanders, but in many cases

mediate important aspects of disease. Increased EC glucose metabo-

lism is emerging as a key feature of angiogenic and hyper-prolifera-

tive ECs. Targeting EC glucose metabolism has recently been shown

as a viable strategy to curb pathological angiogenesis, but is still in

its infancy (Schoors et al, 2014b). Recent technical and conceptual

advances, however, now make it possible to perform comprehen-

sive metabolic studies. These technical breakthroughs have led to a

resurgent interest in targeting cell metabolism for therapeutic gains.

As a proof of concept, targeting EC metabolism by pharmacological

inhibition of the glycolytic enzyme PFKFB3 has shown recent

success in inhibiting pathological angiogenesis (Fig 4) (De Bock

et al, 2013b; Schoors et al, 2014b; Xu et al, 2014). These results,

together with the observation that EC metabolism is altered in many

diseases, suggest that EC metabolism is an attractive and viable but

understudied therapeutic target.

For more informationAuthor website: http://www.vrc-lab.be/peter-carmeliet

AcknowledgementsWe apologize for not being able to cite the work of all other studies related to

this topic because of space restrictions. The authors gratefully acknowledge

Massimo M. Santoro and Richard C. Cubbon for their valuable comments that

helped improve the manuscript. J.G. is a PhD student supported by a BOF

fellowship from the University of Leuven. The work of P.C. is supported by a

Federal Government Belgium grant (IUAP P7/03), long-term structural

Methusalem funding by the Flemish Government, grants from the Research

Foundation Flanders (FWO), the Foundation of Leducq Transatlantic Network

(ARTEMIS), Foundation against Cancer, an European Research Council (ERC)

Advanced Research Grant (EU-ERC269073) and the AXA Research Fund.

Pending issues

The findings in this review suggest that blood vessel pathology is medi-ated, or at least characterized, by disease-specific alterations. However,at present, there are no studies that incorporate state-of-the-art meta-bolomics tools to characterize EC metabolism in disease. Metabolicprofiling using isotope incorporation studies and metabolic flux analysiscould greatly increase our understanding of the metabolic alterationsthat underlie EC pathology.

In vivo studies to characterize EC metabolism in animal models ofhuman disease could provide highly relevant insight in disease-specific metabolic alterations. However, this requires isolation of ECsfrom diseased tissue, which at present poses technical and interpreta-tional challenges for proper analysis of metabolism using advancedmetabolomics methods.

Another pressing issue is the lack of studies characterizing metabolism inpatient-derived tissue using either in or ex vivo models. The recent devel-opment of new protocols to isolate ECs from patient tissue offers thepossibility to study metabolism in clinically relevant systems. Accordingly,such studies could greatly advance the identification of novel biomarkersand therapeutic targets in EC metabolism.

EMBO Molecular Medicine Vol 6 | No 9 | 2014 ª 2014 The Authors

EMBO Molecular Medicine Targeting EC metabolism in disease Jermaine Goveia et al

1114

Conflict of interestThe authors declare that they have no conflict of interest.

References

Ahmed U, Dobler D, Larkin SJ, Rabbani N, Thornalley PJ (2008) Reversal of

hyperglycemia-induced angiogenesis deficit of human endothelial cells

by overexpression of glyoxalase 1 in vitro. Ann N Y Acad Sci 1126:

262 – 264

Arany Z, Foo SY, Ma Y, Ruas JL, Bommi-Reddy A, Girnun G, Cooper M, Laznik

D, Chinsomboon J, Rangwala SM et al (2008) HIF-independent regulation

of VEGF and angiogenesis by the transcriptional coactivator PGC-1alpha.

Nature 451: 1008 – 1012

Atkinson C, Stewart S, Upton PD, Machado R, Thomson JR, Trembath RC,

Morrell NW (2002) Primary pulmonary hypertension is associated with

reduced pulmonary vascular expression of type II bone morphogenetic

protein receptor. Circulation 105: 1672 – 1678

van Beijnum JR, Dings RP, van der Linden E, Zwaans BM, Ramaekers FC,

Mayo KH, Griffioen AW (2006) Gene expression of tumor angiogenesis

dissected: specific targeting of colon cancer angiogenic vasculature. Blood

108: 2339 – 2348

Bendall JK, Douglas G, McNeill E, Channon KM, Crabtree MJ (2014)

Tetrahydrobiopterin in cardiovascular health and disease. Antioxid Redox

Signal 20: 3040 – 3077

Benedito R, Roca C, Sorensen I, Adams S, Gossler A, Fruttiger M, Adams RH

(2009) The notch ligands Dll4 and Jagged1 have opposing effects on

angiogenesis. Cell 137: 1124 – 1135

Bergers G, Benjamin LE (2003) Tumorigenesis and the angiogenic switch. Nat

Rev 3: 401 – 410

Bergers G, Hanahan D (2008) Modes of resistance to anti-angiogenic therapy.

Nat Rev 8: 592 – 603

Blake R, Trounce IA (2013) Mitochondrial dysfunction and complications

associated with diabetes. Biochim Biophys Acta 1840: 1404 – 1412

Bode-Boger SM, Muke J, Surdacki A, Brabant G, Boger RH, Frolich JC (2003)

Oral L-arginine improves endothelial function in healthy individuals older

than 70 years. Vas Med 8: 77 – 81

Boger RH, Sydow K, Borlak J, Thum T, Lenzen H, Schubert B, Tsikas D,

Bode-Boger SM (2000) LDL cholesterol upregulates synthesis of

asymmetrical dimethylarginine in human endothelial cells: involvement of

S-adenosylmethionine-dependent methyltransferases. Circ Res 87: 99 – 105

Boger RH (2004) Asymmetric dimethylarginine, an endogenous inhibitor of

nitric oxide synthase, explains the “L-arginine paradox” and acts as a novel

cardiovascular risk factor. J Nutr 134: 2842S – 2847S; discussion 2853S

Brouwers O, Niessen PM, Haenen G, Miyata T, Brownlee M, Stehouwer CD,

De Mey JG, Schalkwijk CG (2010) Hyperglycaemia-induced impairment of

endothelium-dependent vasorelaxation in rat mesenteric arteries is

mediated by intracellular methylglyoxal levels in a pathway dependent on

oxidative stress. Diabetologia 53: 989 – 1000

Brouwers O, Niessen PM, Miyata T, Ostergaard JA, Flyvbjerg A, Peutz-Kootstra

CJ, Sieber J, Mundel PH, Brownlee M, Janssen BJ et al (2014) Glyoxalase-1

overexpression reduces endothelial dysfunction and attenuates early renal

impairment in a rat model of diabetes. Diabetologia 57: 224 – 235

Brownlee M (2001) Biochemistry and molecular cell biology of diabetic

complications. Nature 414: 813 – 820

Bucala R, Tracey KJ, Cerami A (1991) Advanced glycosylation products quench

nitric oxide and mediate defective endothelium-dependent vasodilatation

in experimental diabetes. J Clin Investig 87: 432 – 438

Carmeliet P (2003) Angiogenesis in health and disease. Nat Med 9: 653 – 660

Carmeliet P, Jain RK (2011) Molecular mechanisms and clinical applications

of angiogenesis. Nature 473: 298 – 307

Chalupsky K, Cai H (2005) Endothelial dihydrofolate reductase: critical for

nitric oxide bioavailability and role in angiotensin II uncoupling of

endothelial nitric oxide synthase. Proc Natl Acad Sci USA 102: 9056 – 9061

Chang CI, Liao JC, Kuo L (1998) Arginase modulates nitric oxide production in

activated macrophages. Am J Physiol 274: H342 –H348

Chen F, Lucas R, Fulton D (2013) The subcellular compartmentalization of

arginine metabolizing enzymes and their role in endothelial dysfunction.

Front Immunol 4: 184

Cines DB, Pollak ES, Buck CA, Loscalzo J, Zimmerman GA, McEver RP, Pober

JS, Wick TM, Konkle BA, Schwartz BS et al (1998) Endothelial cells in

physiology and in the pathophysiology of vascular disorders. Blood 91:

3527 – 3561

Clarke R, Halsey J, Lewington S, Lonn E, Armitage J, Manson JE, Bonaa KH,

Spence JD, Nygard O, Jamison R et al (2010) Effects of lowering

homocysteine levels with B vitamins on cardiovascular disease, cancer,

and cause-specific mortality: Meta-analysis of 8 randomized trials

involving 37 485 individuals. Arch Intern Med 170: 1622 – 1631

Crabtree MJ, Hale AB, Channon KM (2011) Dihydrofolate reductase protects

endothelial nitric oxide synthase from uncoupling in tetrahydrobiopterin

deficiency. Free Radical Biol Med 50: 1639 – 1646

Crabtree MJ, Tatham AL, Al-Wakeel Y, Warrick N, Hale AB, Cai S, Channon

KM, Alp NJ (2009a) Quantitative regulation of intracellular endothelial

nitric-oxide synthase (eNOS) coupling by both tetrahydrobiopterin-eNOS

stoichiometry and biopterin redox status: insights from cells with

tet-regulated GTP cyclohydrolase I expression. J Biol Chem 284:

1136 – 1144

Crabtree MJ, Tatham AL, Hale AB, Alp NJ, Channon KM (2009b) Critical role

for tetrahydrobiopterin recycling by dihydrofolate reductase in regulation

of endothelial nitric-oxide synthase coupling: relative importance of the

de novo biopterin synthesis versus salvage pathways. J Biol Chem 284:

28128 – 28136

Croci DO, Cerliani JP, Dalotto-Moreno T, Mendez-Huergo SP, Mascanfroni ID,

Dergan-Dylon S, Toscano MA, Caramelo JJ, Garcia-Vallejo JJ, Ouyang J et al

(2014) Glycosylation-dependent lectin-receptor interactions preserve

angiogenesis in anti-VEGF refractory tumors. Cell 156: 744 – 758

Culic O, Gruwel ML, Schrader J (1997) Energy turnover of vascular endothelial

cells. Am J Physiol 273: C205 –C213

Cunnington C, Van Assche T, Shirodaria C, Kylintireas I, Lindsay AC, Lee JM,

Antoniades C, Margaritis M, Lee R, Cerrato R et al (2012) Systemic and

vascular oxidation limits the efficacy of oral tetrahydrobiopterin treatment

in patients with coronary artery disease. Circulation 125: 1356 – 1366

Dagher Z, Ruderman N, Tornheim K, Ido Y (1999) The effect of AMP-activated

protein kinase and its activator AICAR on the metabolism of human

umbilical vein endothelial cells. Biochem Biophys Res Commun 265:

112 – 115

Dagher Z, Ruderman N, Tornheim K, Ido Y (2001) Acute regulation of fatty

acid oxidation and amp-activated protein kinase in human umbilical vein

endothelial cells. Circ Res 88: 1276 – 1282

Davignon J, Ganz P (2004) Role of endothelial dysfunction in atherosclerosis.

Circulation 109: III27 – III32

De Bock K, Georgiadou M, Carmeliet P (2013a) Role of endothelial cell

metabolism in vessel sprouting. Cell Metab 18: 634 – 647

De Bock K, Georgiadou M, Schoors S, Kuchnio A, Wong BW, Cantelmo AR,

Quaegebeur A, Ghesquiere B, Cauwenberghs S, Eelen G et al (2013b) Role

of PFKFB3-driven glycolysis in vessel sprouting. Cell 154: 651 – 663

ª 2014 The Authors EMBO Molecular Medicine Vol 6 | No 9 | 2014

Jermaine Goveia et al Targeting EC metabolism in disease EMBO Molecular Medicine

1115

exact mechanisms remain to be elucidated. Nonetheless, early

clinical and preclinical studies have found that therapeutic target-

ing of 1C metabolism, for example, via folate supplementation

lowers levels of homocysteinemia and increases BH4 regeneration

from BH2 (Verhaar et al, 2002). However, large-scale clinical trials

failed to show benefits of folate or BH4 supplementation to

prevent cardiovascular disease (Clarke et al, 2010; Cunnington

et al, 2012; Marti-Carvajal et al, 2013). These clinical and preclini-

cal findings suggest that while L-arginine, folate, methionine,

COQ10 and homocysteine metabolism are potential therapeutic

targets, a more detailed understanding of how these pathways

cause dysfunction is required to design more rational therapeutic

agents.

EC metabolism in the pathogenesis of other diseases

EC metabolism is best characterized in the diseases discussed

above. However, these represent only a minor fraction of the disor-

ders in which pathological EC responses are presumably involved.

Indeed, it is highly likely that EC metabolic alterations are also

involved in the pathogenesis of other diseases such as ischemia,

pre-eclampsia, vasculitis, vascular neoplasms and others although

this has hardly been studied.

On the other hand, many of the EC metabolic alterations that

lead to EC dysfunction are likely induced by cardiovascular risk

factors such as those that characterize metabolic syndrome, hyper-

homocysteinemia and hyperuricemia. For example, elevated serum

uric acid (a breakdown product of purine nucleotides generated by

xanthine oxidase with potent anti-oxidant activity) is common in

patients with hypertension and may even be a root cause of EC

dysfunction leading to cardiovascular disease (Feig et al, 2008).

Interestingly, while uric acid has been described as major anti-

oxidant in human plasma, ECs exposed to uric acid display

increased ROS production creating a paradox that has not been

resolved (Lippi et al, 2008; Sautin & Johnson, 2008). Regardless, in

cardiovascular disease models uric acid reduces mitochondrial

content, intracellular ATP and arginase activity (Zharikov et al,

2008; Sanchez-Lozada et al, 2012). In addition, uric acid inhibits

NO production in ECs in vitro, and in vivo levels of serum nitrites

(an indicator of NO production) are inversely proportional to serum

uric acid concentrations (Khosla et al, 2005). Interestingly, ECs

exposed to uric acid increase expression of AR and alter expression

of several other proteins linked to metabolism (Zhang et al, 2014).

These studies suggest that hyperuricemia induces EC dysfunction

through metabolic alterations. Whether the same is true for other

cardiovascular risk factors remains in question.

A broader characterization of EC metabolism in the future might

reveal novel therapeutic targets in metabolic pathways that are

generally not considered to be important in pathological EC func-

tion. Recent findings that endothelial cholesterol efflux to high-

density lipoprotein regulates angiogenesis (Fang et al, 2013), and

that EC-specific insulin receptor knock-out accelerates atheroscle-

rotic plaque formation (Gage et al, 2013) point to a key role for EC

metabolism in the pathogenesis of disease and indicate that many

more yet to be identified non-traditional but potentially druggable

metabolic enzymes, transporters and pathways may play a role in

vascular disease.

Therapeutic targeting of EC metabolism

Overall, it is clear that pathological blood vessel responses are

associated with metabolic alterations in ECs. These metabolic

adaptations are not just innocent bystanders, but in many cases

mediate important aspects of disease. Increased EC glucose metabo-

lism is emerging as a key feature of angiogenic and hyper-prolifera-

tive ECs. Targeting EC glucose metabolism has recently been shown

as a viable strategy to curb pathological angiogenesis, but is still in

its infancy (Schoors et al, 2014b). Recent technical and conceptual

advances, however, now make it possible to perform comprehen-

sive metabolic studies. These technical breakthroughs have led to a

resurgent interest in targeting cell metabolism for therapeutic gains.

As a proof of concept, targeting EC metabolism by pharmacological

inhibition of the glycolytic enzyme PFKFB3 has shown recent

success in inhibiting pathological angiogenesis (Fig 4) (De Bock

et al, 2013b; Schoors et al, 2014b; Xu et al, 2014). These results,

together with the observation that EC metabolism is altered in many

diseases, suggest that EC metabolism is an attractive and viable but

understudied therapeutic target.

For more informationAuthor website: http://www.vrc-lab.be/peter-carmeliet

AcknowledgementsWe apologize for not being able to cite the work of all other studies related to

this topic because of space restrictions. The authors gratefully acknowledge

Massimo M. Santoro and Richard C. Cubbon for their valuable comments that

helped improve the manuscript. J.G. is a PhD student supported by a BOF

fellowship from the University of Leuven. The work of P.C. is supported by a

Federal Government Belgium grant (IUAP P7/03), long-term structural

Methusalem funding by the Flemish Government, grants from the Research

Foundation Flanders (FWO), the Foundation of Leducq Transatlantic Network

(ARTEMIS), Foundation against Cancer, an European Research Council (ERC)

Advanced Research Grant (EU-ERC269073) and the AXA Research Fund.

Pending issues

The findings in this review suggest that blood vessel pathology is medi-ated, or at least characterized, by disease-specific alterations. However,at present, there are no studies that incorporate state-of-the-art meta-bolomics tools to characterize EC metabolism in disease. Metabolicprofiling using isotope incorporation studies and metabolic flux analysiscould greatly increase our understanding of the metabolic alterationsthat underlie EC pathology.

In vivo studies to characterize EC metabolism in animal models ofhuman disease could provide highly relevant insight in disease-specific metabolic alterations. However, this requires isolation of ECsfrom diseased tissue, which at present poses technical and interpreta-tional challenges for proper analysis of metabolism using advancedmetabolomics methods.

Another pressing issue is the lack of studies characterizing metabolism inpatient-derived tissue using either in or ex vivo models. The recent devel-opment of new protocols to isolate ECs from patient tissue offers thepossibility to study metabolism in clinically relevant systems. Accordingly,such studies could greatly advance the identification of novel biomarkersand therapeutic targets in EC metabolism.

EMBO Molecular Medicine Vol 6 | No 9 | 2014 ª 2014 The Authors

EMBO Molecular Medicine Targeting EC metabolism in disease Jermaine Goveia et al

1114

DeBerardinis RJ, Cheng T (2010) Q’s next: the diverse functions of glutamine

in metabolism, cell biology and cancer. Oncogene 29: 313 – 324

Dhillon B, Badiwala MV, Maitland A, Rao V, Li SH, Verma S (2003)

Tetrahydrobiopterin attenuates homocysteine induced endothelial

dysfunction. Mol Cell Biochem 247: 223 – 227

Dobrina A, Rossi F (1983) Metabolic properties of freshly isolated bovine

endothelial cells. Biochim Biophys Acta 762: 295 – 301

Dranka BP, Hill BG, Darley-Usmar VM (2010) Mitochondrial reserve capacity

in endothelial cells: The impact of nitric oxide and reactive oxygen

species. Free Radical Biol Med 48: 905 – 914

Du XL, Edelstein D, Rossetti L, Fantus IG, Goldberg H, Ziyadeh F, Wu J,

Brownlee M (2000) Hyperglycemia-induced mitochondrial superoxide

overproduction activates the hexosamine pathway and induces

plasminogen activator inhibitor-1 expression by increasing Sp1

glycosylation. Proc Natl Acad Sci USA 97: 12222 – 12226

Du XL, Edelstein D, Dimmeler S, Ju Q, Sui C, Brownlee M (2001)

Hyperglycemia inhibits endothelial nitric oxide synthase activity by

posttranslational modification at the Akt site. J Clin Invest 108: 1341 – 1348

Du X, Matsumura T, Edelstein D, Rossetti L, Zsengeller Z, Szabo C, Brownlee

M (2003) Inhibition of GAPDH activity by poly(ADP-ribose) polymerase

activates three major pathways of hyperglycemic damage in endothelial

cells. J Clin Investig 112: 1049 – 1057

Dymkowska D, Drabarek B, Podszywalow-Bartnicka P, Szczepanowska J,

Zablocki K (2014) Hyperglycaemia modifies energy metabolism and

reactive oxygen species formation in endothelial cells in vitro. Arch

Biochem Biophys 542: 7 – 13

Ebos JM, Kerbel RS (2011) Antiangiogenic therapy: impact on invasion,

disease progression, and metastasis. Nat Rev Clin Oncol 8: 210 – 221

den Eynden JV, Ali SS, Horwood N, Carmans S, Brone B, Hellings N, Steels P,

Harvey RJ, Rigo JM (2009) Glycine and glycine receptor signalling in

non-neuronal cells. Front Mol Neurosci 2: 9

Fang L, Choi SH, Baek JS, Liu C, Almazan F, Ulrich F, Wiesner P, Taleb A, Deer

E, Pattison J et al (2013) Control of angiogenesis by AIBP-mediated

cholesterol efflux. Nature 498: 118 – 122

Federici M, Menghini R, Mauriello A, Hribal ML, Ferrelli F, Lauro D, Sbraccia P,

Spagnoli LG, Sesti G, Lauro R (2002) Insulin-dependent activation of

endothelial nitric oxide synthase is impaired by O-linked glycosylation

modification of signaling proteins in human coronary endothelial cells.

Circulation 106: 466 – 472

Feig DI, Kang DH, Johnson RJ (2008) Uric acid and cardiovascular risk. New

Engl J Med 359: 1811 – 1821

Fessel JP, Hamid R, Wittmann BM, Robinson LJ, Blackwell T, Tada Y, Tanabe

N, Tatsumi K, Hemnes AR, West JD (2012) Metabolomic analysis of bone

morphogenetic protein receptor type 2 mutations in human pulmonary

endothelium reveals widespread metabolic reprogramming. Pulm Circ 2:

201 – 213

Fijalkowska I, Xu W, Comhair SA, Janocha AJ, Mavrakis LA, Krishnamachary B,

Zhen L, Mao T, Richter A, Erzurum SC et al (2010) Hypoxia

inducible-factor1alpha regulates the metabolic shift of pulmonary

hypertensive endothelial cells. Am J Pathol 176: 1130 – 1138

Frederiksen J, Juul K, Grande P, Jensen GB, Schroeder TV, Tybjaerg-Hansen A,

Nordestgaard BG (2004) Methylenetetrahydrofolate reductase

polymorphism (C677T), hyperhomocysteinemia, and risk of ischemic

cardiovascular disease and venous thromboembolism: prospective and

case-control studies from the Copenhagen City Heart Study. Blood 104:

3046 – 3051

Gage MC, Yuldasheva NY, Viswambharan H, Sukumar P, Cubbon RM,

Galloway S, Imrie H, Skromna A, Smith J, Jackson CL et al (2013)

Endothelium-specific insulin resistance leads to accelerated

atherosclerosis in areas with disturbed flow patterns: a role for reactive

oxygen species. Atherosclerosis 230: 131 – 139

Gao L, Mao Q, Cao J, Wang Y, Zhou X, Fan L (2012) Effects of coenzyme Q10

on vascular endothelial function in humans: a meta-analysis of

randomized controlled trials. Atherosclerosis 221: 311 – 316

Gaudreault N, Scriven DR, Moore ED (2004) Characterisation of

glucose transporters in the intact coronary artery endothelium in rats:

GLUT-2 upregulated by long-term hyperglycaemia. Diabetologia 47:

2081 – 2092

Gaudreault N, Scriven DR, Laher I, Moore ED (2008) Subcellular

characterization of glucose uptake in coronary endothelial cells. Microvasc

Res 75: 73 – 82

Giacco F, Brownlee M (2010) Oxidative stress and diabetic complications. Circ

Res 107: 1058 – 1070

Goldin A, Beckman JA, Schmidt AM, Creager MA (2006) Advanced glycation

end products: sparking the development of diabetic vascular injury.

Circulation 114: 597 – 605

Goldstein JL, Brown MS (1990) Regulation of the mevalonate pathway. Nature

343: 425 – 430

Gorren AC, Bec N, Schrammel A, Werner ER, Lange R, Mayer B (2000)

Low-temperature optical absorption spectra suggest a redox role for

tetrahydrobiopterin in both steps of nitric oxide synthase catalysis.

Biochemistry 39: 11763 – 11770

Harjes U, Bensaad K, Harris AL (2012) Endothelial cell metabolism and

implications for cancer therapy. Br J Cancer 107: 1207 – 1212

Hassan HH, Denis M, Krimbou L, Marcil M, Genest J (2006) Cellular

cholesterol homeostasis in vascular endothelial cells. Can J Cardiol 22

(Suppl B): 35B – 40B

Heitzer T, Krohn K, Albers S, Meinertz T (2000) Tetrahydrobiopterin

improves endothelium-dependent vasodilation by increasing nitric oxide

activity in patients with Type II diabetes mellitus. Diabetologia 43:

1435 – 1438

Herskowitz K, Bode BP, Block ER, Souba WW (1991) Characterization of

L-glutamine transport by pulmonary artery endothelial cells. Am J Physiol

260: L241 – L246

Herzig S, Long F, Jhala US, Hedrick S, Quinn R, Bauer A, Rudolph D, Schutz G,

Yoon C, Puigserver P et al (2001) CREB regulates hepatic gluconeogenesis

through the coactivator PGC-1. Nature 413: 179 – 183

Hinshaw DB, Burger JM (1990) Protective effect of glutamine on endothelial

cell ATP in oxidant injury. J Surg Res 49: 222 – 227

Hopkins PN (2013) Molecular biology of atherosclerosis. Physiol Rev 93:

1317 – 1542

Hu X, Xu X, Zhu G, Atzler D, Kimoto M, Chen J, Schwedhelm E, Luneburg N,

Boger RH, Zhang P et al (2009) Vascular endothelial-specific

dimethylarginine dimethylaminohydrolase-1-deficient mice reveal that

vascular endothelium plays an important role in removing asymmetric

dimethylarginine. Circulation 120: 2222 – 2229

Humphrey LL, Fu R, Rogers K, Freeman M, Helfand M (2008) Homocysteine

level and coronary heart disease incidence: a systematic review and

meta-analysis. Mayo Clin Proc 83: 1203 – 1212

Hunt TK, Aslam RS, Beckert S, Wagner S, Ghani QP, Hussain MZ, Roy S,

Sen CK (2007) Aerobically derived lactate stimulates revascularization

and tissue repair via redox mechanisms. Antioxid Redox Signal 9:

1115 – 1124

Inoguchi T, Sonta T, Tsubouchi H, Etoh T, Kakimoto M, Sonoda N, Sato N,

Sekiguchi N, Kobayashi K, Sumimoto H, et al (2003) Protein kinase

C-dependent increase in reactive oxygen species (ROS) production in

EMBO Molecular Medicine Vol 6 | No 9 | 2014 ª 2014 The Authors

EMBO Molecular Medicine Targeting EC metabolism in disease Jermaine Goveia et al

1116

vascular tissues of diabetes: role of vascular NAD(P)H oxidase. J Am Soc

Nephrol 14: S227 – S232

Ivashchenko CY, Bradley BT, Ao Z, Leiper J, Vallance P, Johns DG (2010)

Regulation of the ADMA-DDAH system in endothelial cells: a novel

mechanism for the sterol response element binding proteins, SREBP1c and

-2. Am J Physiol Heart Circ Physiol 298: H251 –H258

Jain RK (1987) Transport of molecules in the tumor interstitium: a review.

Cancer Res 47: 3039 – 3051

Jongkind JF, Verkerk A, Baggen RG (1989) Glutathione metabolism of human

vascular endothelial cells under peroxidative stress. Free Radical Biol Med

7: 507 – 512

Kashiwagi A, Asahina T, Ikebuchi M, Tanaka Y, Takagi Y, Nishio Y, Kikkawa R,

Shigeta Y (1994) Abnormal glutathione metabolism and increased

cytotoxicity caused by H2O2 in human umbilical vein endothelial cells

cultured in high glucose medium. Diabetologia 37: 264 – 269

Kawashima S, Yokoyama M (2004) Dysfunction of endothelial nitric oxide

synthase and atherosclerosis. Arterioscler Thromb Vasc Biol 24:

998 – 1005

Kelly PJ, Rosand J, Kistler JP, Shih VE, Silveira S, Plomaritoglou A, Furie KL

(2002) Homocysteine, MTHFR 677C–>T polymorphism, and risk of ischemic

stroke: results of a meta-analysis. Neurology 59: 529 – 536

Khosla UM, Zharikov S, Finch JL, Nakagawa T, Roncal C, Mu W, Krotova K,

Block ER, Prabhakar S, Johnson RJ (2005) Hyperuricemia induces

endothelial dysfunction. Kidney Int 67: 1739 – 1742

Kidokoro K, Satoh M, Channon KM, Yada T, Sasaki T, Kashihara N (2013)

Maintenance of endothelial guanosine triphosphate cyclohydrolase I

ameliorates diabetic nephropathy. J Am Soc Nephrol 24: 1139 – 1150

Klerk M, Verhoef P, Clarke R, Blom HJ, Kok FJ, Schouten EG (2002) MTHFR

677C–>T polymorphism and risk of coronary heart disease: a

meta-analysis. JAMA 288: 2023 – 2031

Koziel A, Woyda-Ploszczyca A, Kicinska A, Jarmuszkiewicz W (2012) The

influence of high glucose on the aerobic metabolism of endothelial

EA.hy926 cells. Pflugers Arch 464: 657 – 669

Krotova K, Patel JM, Block ER, Zharikov S (2010) Endothelial arginase II

responds to pharmacological inhibition by elevation in protein level. Mol

Cell Biochem 343: 211 – 216

Krutzfeldt A, Spahr R, Mertens S, Siegmund B, Piper HM (1990) Metabolism of

exogenous substrates by coronary endothelial cells in culture. J Mol Cell

Cardiol 22: 1393 – 1404

Laczy B, Hill BG, Wang K, Paterson AJ, White CR, Xing D, Chen YF,

Darley-Usmar V, Oparil S, Chatham JC (2009) Protein O-GlcNAcylation: a

new signaling paradigm for the cardiovascular system. Am J Physiol Heart

Circ Physiol 296: H13 –H28

Laufs U, La Fata V, Plutzky J, Liao JK (1998) Upregulation of endothelial nitric

oxide synthase by HMG CoA reductase inhibitors. Circulation 97: 1129 – 1135

Laufs U, Liao JK (1998) Post-transcriptional regulation of endothelial nitric

oxide synthase mRNA stability by Rho GTPase. J Biol Chem 273:

24266 – 24271

Lee R, Channon KM, Antoniades C (2012) Therapeutic strategies targeting

endothelial function in humans: clinical implications. Curr Vasc Pharmacol

10: 77 – 93

Leighton B, Curi R, Hussein A, Newsholme EA (1987) Maximum activities of

some key enzymes of glycolysis, glutaminolysis, Krebs cycle and fatty acid

utilization in bovine pulmonary endothelial cells. FEBS Lett 225: 93 – 96

Leiper J, Nandi M (2011) The therapeutic potential of targeting endogenous

inhibitors of nitric oxide synthesis. Nat Rev Drug Discovery 10: 277 – 291

Leopold JA, Zhang YY, Scribner AW, Stanton RC, Loscalzo J (2003)

Glucose-6-phosphate dehydrogenase overexpression decreases endothelial

cell oxidant stress and increases bioavailable nitric oxide. Arterioscler

Thromb Vasc Biol 23: 411 – 417

Lerman A, Burnett JC Jr, Higano ST, McKinley LJ, Holmes DR Jr (1998)

Long-term L-arginine supplementation improves small-vessel coronary

endothelial function in humans. Circulation 97: 2123 – 2128

Liao JK, Laufs U (2005) Pleiotropic effects of statins. Annu Rev Pharmacol

Toxicol 45: 89 – 118

Lin J, Wu H, Tarr PT, Zhang CY, Wu Z, Boss O, Michael LF, Puigserver P,

Isotani E, Olson EN et al (2002) Transcriptional co-activator PGC-1

alpha drives the formation of slow-twitch muscle fibres. Nature 418:

797 – 801

Lippi G, Montagnana M, Franchini M, Favaloro EJ, Targher G (2008) The

paradoxical relationship between serum uric acid and cardiovascular

disease. Clin Chim Acta 392: 1 – 7

Locasale JW (2013) Serine, glycine and one-carbon units: cancer metabolism

in full circle. Nat Rev 13: 572 – 583

Loges S, Schmidt T, Carmeliet P (2010) Mechanisms of resistance to

anti-angiogenic therapy and development of third-generation

anti-angiogenic drug candidates. Genes & cancer 1: 12 – 25

Long L, MacLean MR, Jeffery TK, Morecroft I, Yang X, Rudarakanchana N,

Southwood M, James V, Trembath RC, Morrell NW (2006) Serotonin

increases susceptibility to pulmonary hypertension in BMPR2-deficient

mice. Circ Res 98: 818 – 827

Luo B, Soesanto Y, McClain DA (2008) Protein modification by O-linked

GlcNAc reduces angiogenesis by inhibiting Akt activity in endothelial cells.

Arterioscler Thromb Vasc Biol 28: 651 – 657

Majka S, Hagen M, Blackwell T, Harral J, Johnson JA, Gendron R, Paradis H,

Crona D, Loyd JE, Nozik-Grayck E et al (2011) Physiologic and molecular

consequences of endothelial Bmpr2 mutation. Respir Res 12: 84

Majmundar AJ, Wong WJ, Simon MC (2010) Hypoxia-inducible factors and

the response to hypoxic stress. Mol Cell 40: 294 – 309

Manigrasso MB, Juranek J, Ramasamy R, Schmidt AM (2014) Unlocking the

biology of RAGE in diabetic microvascular complications. Trends Endocrinol

Metab 25: 15 – 22

Mann GE, Yudilevich DL, Sobrevia L (2003) Regulation of amino acid and

glucose transporters in endothelial and smooth muscle cells. Physiol Rev

83: 183 – 252

Marti-Carvajal AJ, Sola I, Lathyris D, Karakitsiou DE, Simancas-Racines D

(2013) Homocysteine-lowering interventions for preventing cardiovascular

events. Cochrane Database Syst Rev 1: CD006612

Masri FA, Comhair SA, Dostanic-Larson I, Kaneko FT, Dweik RA, Arroliga AC,

Erzurum SC (2008) Deficiency of lung antioxidants in idiopathic

pulmonary arterial hypertension. Clin Transl Sci 1: 99 – 106

McCarty MF, Barroso-Aranda J, Contreras F (2009) The hyperpolarizing impact

of glycine on endothelial cells may be anti-atherogenic. Med Hypotheses

73: 263– 264

McMurtry MS, Bonnet S, Wu X, Dyck JR, Haromy A, Hashimoto K, Michelakis

ED (2004) Dichloroacetate prevents and reverses pulmonary hypertension

by inducing pulmonary artery smooth muscle cell apoptosis. Circ Res 95:

830 – 840

Michelakis ED, Sutendra G, Dromparis P, Webster L, Haromy A, Niven E,

Maguire C, Gammer TL, Mackey JR, Fulton D et al (2010) Metabolic

modulation of glioblastoma with dichloroacetate. Sci Transl Med 2:

31ra34

Min C, Kang E, Yu SH, Shinn SH, Kim YS (1999) Advanced glycation end

products induce apoptosis and procoagulant activity in cultured

human umbilical vein endothelial cells. Diabetes Res Clin Pract 46:

197 – 202

ª 2014 The Authors EMBO Molecular Medicine Vol 6 | No 9 | 2014

Jermaine Goveia et al Targeting EC metabolism in disease EMBO Molecular Medicine

1117

DeBerardinis RJ, Cheng T (2010) Q’s next: the diverse functions of glutamine

in metabolism, cell biology and cancer. Oncogene 29: 313 – 324

Dhillon B, Badiwala MV, Maitland A, Rao V, Li SH, Verma S (2003)

Tetrahydrobiopterin attenuates homocysteine induced endothelial

dysfunction. Mol Cell Biochem 247: 223 – 227

Dobrina A, Rossi F (1983) Metabolic properties of freshly isolated bovine

endothelial cells. Biochim Biophys Acta 762: 295 – 301

Dranka BP, Hill BG, Darley-Usmar VM (2010) Mitochondrial reserve capacity

in endothelial cells: The impact of nitric oxide and reactive oxygen

species. Free Radical Biol Med 48: 905 – 914

Du XL, Edelstein D, Rossetti L, Fantus IG, Goldberg H, Ziyadeh F, Wu J,

Brownlee M (2000) Hyperglycemia-induced mitochondrial superoxide

overproduction activates the hexosamine pathway and induces

plasminogen activator inhibitor-1 expression by increasing Sp1

glycosylation. Proc Natl Acad Sci USA 97: 12222 – 12226

Du XL, Edelstein D, Dimmeler S, Ju Q, Sui C, Brownlee M (2001)

Hyperglycemia inhibits endothelial nitric oxide synthase activity by

posttranslational modification at the Akt site. J Clin Invest 108: 1341 – 1348

Du X, Matsumura T, Edelstein D, Rossetti L, Zsengeller Z, Szabo C, Brownlee

M (2003) Inhibition of GAPDH activity by poly(ADP-ribose) polymerase

activates three major pathways of hyperglycemic damage in endothelial

cells. J Clin Investig 112: 1049 – 1057

Dymkowska D, Drabarek B, Podszywalow-Bartnicka P, Szczepanowska J,

Zablocki K (2014) Hyperglycaemia modifies energy metabolism and

reactive oxygen species formation in endothelial cells in vitro. Arch

Biochem Biophys 542: 7 – 13

Ebos JM, Kerbel RS (2011) Antiangiogenic therapy: impact on invasion,

disease progression, and metastasis. Nat Rev Clin Oncol 8: 210 – 221

den Eynden JV, Ali SS, Horwood N, Carmans S, Brone B, Hellings N, Steels P,

Harvey RJ, Rigo JM (2009) Glycine and glycine receptor signalling in

non-neuronal cells. Front Mol Neurosci 2: 9

Fang L, Choi SH, Baek JS, Liu C, Almazan F, Ulrich F, Wiesner P, Taleb A, Deer

E, Pattison J et al (2013) Control of angiogenesis by AIBP-mediated

cholesterol efflux. Nature 498: 118 – 122

Federici M, Menghini R, Mauriello A, Hribal ML, Ferrelli F, Lauro D, Sbraccia P,

Spagnoli LG, Sesti G, Lauro R (2002) Insulin-dependent activation of

endothelial nitric oxide synthase is impaired by O-linked glycosylation

modification of signaling proteins in human coronary endothelial cells.

Circulation 106: 466 – 472

Feig DI, Kang DH, Johnson RJ (2008) Uric acid and cardiovascular risk. New

Engl J Med 359: 1811 – 1821

Fessel JP, Hamid R, Wittmann BM, Robinson LJ, Blackwell T, Tada Y, Tanabe

N, Tatsumi K, Hemnes AR, West JD (2012) Metabolomic analysis of bone

morphogenetic protein receptor type 2 mutations in human pulmonary

endothelium reveals widespread metabolic reprogramming. Pulm Circ 2:

201 – 213

Fijalkowska I, Xu W, Comhair SA, Janocha AJ, Mavrakis LA, Krishnamachary B,

Zhen L, Mao T, Richter A, Erzurum SC et al (2010) Hypoxia

inducible-factor1alpha regulates the metabolic shift of pulmonary

hypertensive endothelial cells. Am J Pathol 176: 1130 – 1138

Frederiksen J, Juul K, Grande P, Jensen GB, Schroeder TV, Tybjaerg-Hansen A,

Nordestgaard BG (2004) Methylenetetrahydrofolate reductase

polymorphism (C677T), hyperhomocysteinemia, and risk of ischemic

cardiovascular disease and venous thromboembolism: prospective and

case-control studies from the Copenhagen City Heart Study. Blood 104:

3046 – 3051

Gage MC, Yuldasheva NY, Viswambharan H, Sukumar P, Cubbon RM,

Galloway S, Imrie H, Skromna A, Smith J, Jackson CL et al (2013)

Endothelium-specific insulin resistance leads to accelerated

atherosclerosis in areas with disturbed flow patterns: a role for reactive

oxygen species. Atherosclerosis 230: 131 – 139

Gao L, Mao Q, Cao J, Wang Y, Zhou X, Fan L (2012) Effects of coenzyme Q10

on vascular endothelial function in humans: a meta-analysis of

randomized controlled trials. Atherosclerosis 221: 311 – 316

Gaudreault N, Scriven DR, Moore ED (2004) Characterisation of

glucose transporters in the intact coronary artery endothelium in rats:

GLUT-2 upregulated by long-term hyperglycaemia. Diabetologia 47:

2081 – 2092

Gaudreault N, Scriven DR, Laher I, Moore ED (2008) Subcellular

characterization of glucose uptake in coronary endothelial cells. Microvasc

Res 75: 73 – 82

Giacco F, Brownlee M (2010) Oxidative stress and diabetic complications. Circ

Res 107: 1058 – 1070

Goldin A, Beckman JA, Schmidt AM, Creager MA (2006) Advanced glycation

end products: sparking the development of diabetic vascular injury.

Circulation 114: 597 – 605

Goldstein JL, Brown MS (1990) Regulation of the mevalonate pathway. Nature

343: 425 – 430

Gorren AC, Bec N, Schrammel A, Werner ER, Lange R, Mayer B (2000)

Low-temperature optical absorption spectra suggest a redox role for

tetrahydrobiopterin in both steps of nitric oxide synthase catalysis.

Biochemistry 39: 11763 – 11770

Harjes U, Bensaad K, Harris AL (2012) Endothelial cell metabolism and

implications for cancer therapy. Br J Cancer 107: 1207 – 1212

Hassan HH, Denis M, Krimbou L, Marcil M, Genest J (2006) Cellular

cholesterol homeostasis in vascular endothelial cells. Can J Cardiol 22

(Suppl B): 35B – 40B

Heitzer T, Krohn K, Albers S, Meinertz T (2000) Tetrahydrobiopterin

improves endothelium-dependent vasodilation by increasing nitric oxide

activity in patients with Type II diabetes mellitus. Diabetologia 43:

1435 – 1438

Herskowitz K, Bode BP, Block ER, Souba WW (1991) Characterization of

L-glutamine transport by pulmonary artery endothelial cells. Am J Physiol

260: L241 – L246

Herzig S, Long F, Jhala US, Hedrick S, Quinn R, Bauer A, Rudolph D, Schutz G,

Yoon C, Puigserver P et al (2001) CREB regulates hepatic gluconeogenesis

through the coactivator PGC-1. Nature 413: 179 – 183

Hinshaw DB, Burger JM (1990) Protective effect of glutamine on endothelial

cell ATP in oxidant injury. J Surg Res 49: 222 – 227

Hopkins PN (2013) Molecular biology of atherosclerosis. Physiol Rev 93:

1317 – 1542

Hu X, Xu X, Zhu G, Atzler D, Kimoto M, Chen J, Schwedhelm E, Luneburg N,

Boger RH, Zhang P et al (2009) Vascular endothelial-specific

dimethylarginine dimethylaminohydrolase-1-deficient mice reveal that

vascular endothelium plays an important role in removing asymmetric

dimethylarginine. Circulation 120: 2222 – 2229

Humphrey LL, Fu R, Rogers K, Freeman M, Helfand M (2008) Homocysteine

level and coronary heart disease incidence: a systematic review and

meta-analysis. Mayo Clin Proc 83: 1203 – 1212

Hunt TK, Aslam RS, Beckert S, Wagner S, Ghani QP, Hussain MZ, Roy S,

Sen CK (2007) Aerobically derived lactate stimulates revascularization

and tissue repair via redox mechanisms. Antioxid Redox Signal 9:

1115 – 1124

Inoguchi T, Sonta T, Tsubouchi H, Etoh T, Kakimoto M, Sonoda N, Sato N,

Sekiguchi N, Kobayashi K, Sumimoto H, et al (2003) Protein kinase

C-dependent increase in reactive oxygen species (ROS) production in

EMBO Molecular Medicine Vol 6 | No 9 | 2014 ª 2014 The Authors

EMBO Molecular Medicine Targeting EC metabolism in disease Jermaine Goveia et al

1116

vascular tissues of diabetes: role of vascular NAD(P)H oxidase. J Am Soc

Nephrol 14: S227 – S232

Ivashchenko CY, Bradley BT, Ao Z, Leiper J, Vallance P, Johns DG (2010)

Regulation of the ADMA-DDAH system in endothelial cells: a novel

mechanism for the sterol response element binding proteins, SREBP1c and

-2. Am J Physiol Heart Circ Physiol 298: H251 –H258

Jain RK (1987) Transport of molecules in the tumor interstitium: a review.

Cancer Res 47: 3039 – 3051

Jongkind JF, Verkerk A, Baggen RG (1989) Glutathione metabolism of human

vascular endothelial cells under peroxidative stress. Free Radical Biol Med

7: 507 – 512

Kashiwagi A, Asahina T, Ikebuchi M, Tanaka Y, Takagi Y, Nishio Y, Kikkawa R,

Shigeta Y (1994) Abnormal glutathione metabolism and increased

cytotoxicity caused by H2O2 in human umbilical vein endothelial cells

cultured in high glucose medium. Diabetologia 37: 264 – 269

Kawashima S, Yokoyama M (2004) Dysfunction of endothelial nitric oxide

synthase and atherosclerosis. Arterioscler Thromb Vasc Biol 24:

998 – 1005

Kelly PJ, Rosand J, Kistler JP, Shih VE, Silveira S, Plomaritoglou A, Furie KL

(2002) Homocysteine, MTHFR 677C–>T polymorphism, and risk of ischemic

stroke: results of a meta-analysis. Neurology 59: 529 – 536

Khosla UM, Zharikov S, Finch JL, Nakagawa T, Roncal C, Mu W, Krotova K,

Block ER, Prabhakar S, Johnson RJ (2005) Hyperuricemia induces

endothelial dysfunction. Kidney Int 67: 1739 – 1742

Kidokoro K, Satoh M, Channon KM, Yada T, Sasaki T, Kashihara N (2013)

Maintenance of endothelial guanosine triphosphate cyclohydrolase I

ameliorates diabetic nephropathy. J Am Soc Nephrol 24: 1139 – 1150

Klerk M, Verhoef P, Clarke R, Blom HJ, Kok FJ, Schouten EG (2002) MTHFR

677C–>T polymorphism and risk of coronary heart disease: a

meta-analysis. JAMA 288: 2023 – 2031

Koziel A, Woyda-Ploszczyca A, Kicinska A, Jarmuszkiewicz W (2012) The

influence of high glucose on the aerobic metabolism of endothelial

EA.hy926 cells. Pflugers Arch 464: 657 – 669

Krotova K, Patel JM, Block ER, Zharikov S (2010) Endothelial arginase II

responds to pharmacological inhibition by elevation in protein level. Mol

Cell Biochem 343: 211 – 216

Krutzfeldt A, Spahr R, Mertens S, Siegmund B, Piper HM (1990) Metabolism of

exogenous substrates by coronary endothelial cells in culture. J Mol Cell

Cardiol 22: 1393 – 1404

Laczy B, Hill BG, Wang K, Paterson AJ, White CR, Xing D, Chen YF,

Darley-Usmar V, Oparil S, Chatham JC (2009) Protein O-GlcNAcylation: a

new signaling paradigm for the cardiovascular system. Am J Physiol Heart

Circ Physiol 296: H13 –H28

Laufs U, La Fata V, Plutzky J, Liao JK (1998) Upregulation of endothelial nitric

oxide synthase by HMG CoA reductase inhibitors. Circulation 97: 1129 – 1135

Laufs U, Liao JK (1998) Post-transcriptional regulation of endothelial nitric

oxide synthase mRNA stability by Rho GTPase. J Biol Chem 273:

24266 – 24271

Lee R, Channon KM, Antoniades C (2012) Therapeutic strategies targeting

endothelial function in humans: clinical implications. Curr Vasc Pharmacol

10: 77 – 93

Leighton B, Curi R, Hussein A, Newsholme EA (1987) Maximum activities of

some key enzymes of glycolysis, glutaminolysis, Krebs cycle and fatty acid

utilization in bovine pulmonary endothelial cells. FEBS Lett 225: 93 – 96

Leiper J, Nandi M (2011) The therapeutic potential of targeting endogenous

inhibitors of nitric oxide synthesis. Nat Rev Drug Discovery 10: 277 – 291

Leopold JA, Zhang YY, Scribner AW, Stanton RC, Loscalzo J (2003)

Glucose-6-phosphate dehydrogenase overexpression decreases endothelial

cell oxidant stress and increases bioavailable nitric oxide. Arterioscler

Thromb Vasc Biol 23: 411 – 417

Lerman A, Burnett JC Jr, Higano ST, McKinley LJ, Holmes DR Jr (1998)

Long-term L-arginine supplementation improves small-vessel coronary

endothelial function in humans. Circulation 97: 2123 – 2128

Liao JK, Laufs U (2005) Pleiotropic effects of statins. Annu Rev Pharmacol

Toxicol 45: 89 – 118

Lin J, Wu H, Tarr PT, Zhang CY, Wu Z, Boss O, Michael LF, Puigserver P,

Isotani E, Olson EN et al (2002) Transcriptional co-activator PGC-1

alpha drives the formation of slow-twitch muscle fibres. Nature 418:

797 – 801

Lippi G, Montagnana M, Franchini M, Favaloro EJ, Targher G (2008) The

paradoxical relationship between serum uric acid and cardiovascular

disease. Clin Chim Acta 392: 1 – 7

Locasale JW (2013) Serine, glycine and one-carbon units: cancer metabolism

in full circle. Nat Rev 13: 572 – 583

Loges S, Schmidt T, Carmeliet P (2010) Mechanisms of resistance to

anti-angiogenic therapy and development of third-generation

anti-angiogenic drug candidates. Genes & cancer 1: 12 – 25

Long L, MacLean MR, Jeffery TK, Morecroft I, Yang X, Rudarakanchana N,

Southwood M, James V, Trembath RC, Morrell NW (2006) Serotonin

increases susceptibility to pulmonary hypertension in BMPR2-deficient

mice. Circ Res 98: 818 – 827

Luo B, Soesanto Y, McClain DA (2008) Protein modification by O-linked

GlcNAc reduces angiogenesis by inhibiting Akt activity in endothelial cells.

Arterioscler Thromb Vasc Biol 28: 651 – 657

Majka S, Hagen M, Blackwell T, Harral J, Johnson JA, Gendron R, Paradis H,

Crona D, Loyd JE, Nozik-Grayck E et al (2011) Physiologic and molecular

consequences of endothelial Bmpr2 mutation. Respir Res 12: 84

Majmundar AJ, Wong WJ, Simon MC (2010) Hypoxia-inducible factors and

the response to hypoxic stress. Mol Cell 40: 294 – 309

Manigrasso MB, Juranek J, Ramasamy R, Schmidt AM (2014) Unlocking the

biology of RAGE in diabetic microvascular complications. Trends Endocrinol

Metab 25: 15 – 22

Mann GE, Yudilevich DL, Sobrevia L (2003) Regulation of amino acid and

glucose transporters in endothelial and smooth muscle cells. Physiol Rev

83: 183 – 252

Marti-Carvajal AJ, Sola I, Lathyris D, Karakitsiou DE, Simancas-Racines D

(2013) Homocysteine-lowering interventions for preventing cardiovascular

events. Cochrane Database Syst Rev 1: CD006612

Masri FA, Comhair SA, Dostanic-Larson I, Kaneko FT, Dweik RA, Arroliga AC,

Erzurum SC (2008) Deficiency of lung antioxidants in idiopathic

pulmonary arterial hypertension. Clin Transl Sci 1: 99 – 106

McCarty MF, Barroso-Aranda J, Contreras F (2009) The hyperpolarizing impact

of glycine on endothelial cells may be anti-atherogenic. Med Hypotheses

73: 263– 264

McMurtry MS, Bonnet S, Wu X, Dyck JR, Haromy A, Hashimoto K, Michelakis

ED (2004) Dichloroacetate prevents and reverses pulmonary hypertension

by inducing pulmonary artery smooth muscle cell apoptosis. Circ Res 95:

830 – 840

Michelakis ED, Sutendra G, Dromparis P, Webster L, Haromy A, Niven E,

Maguire C, Gammer TL, Mackey JR, Fulton D et al (2010) Metabolic

modulation of glioblastoma with dichloroacetate. Sci Transl Med 2:

31ra34

Min C, Kang E, Yu SH, Shinn SH, Kim YS (1999) Advanced glycation end

products induce apoptosis and procoagulant activity in cultured

human umbilical vein endothelial cells. Diabetes Res Clin Pract 46:

197 – 202

ª 2014 The Authors EMBO Molecular Medicine Vol 6 | No 9 | 2014

Jermaine Goveia et al Targeting EC metabolism in disease EMBO Molecular Medicine

1117

Ming XF, Barandier C, Viswambharan H, Kwak BR, Mach F, Mazzolai L, Hayoz

D, Ruffieux J, Rusconi S, Montani JP et al (2004) Thrombin stimulates

human endothelial arginase enzymatic activity via RhoA/ROCK pathway:

implications for atherosclerotic endothelial dysfunction. Circulation 110:

3708 – 3714

Mishra RC, Tripathy S, Desai KM, Quest D, Lu Y, Akhtar J, Gopalakrishnan V

(2008) Nitric oxide synthase inhibition promotes endothelium-dependent

vasodilatation and the antihypertensive effect of L-serine. Hypertension 51:

791 – 796

Morrison RF, Seidel ER (1995) Vascular endothelial cell proliferation:

regulation of cellular polyamines. Cardiovasc Res 29: 841 – 847

Mugoni V, Postel R, Catanzaro V, De Luca E, Turco E, Digilio G, Silengo L,

Murphy MP, Medana C, Stainier DY et al (2013) Ubiad1 is an antioxidant

enzyme that regulates eNOS activity by CoQ10 synthesis. Cell 152: 504 – 518

Nishikawa T, Edelstein D, Du XL, Yamagishi S, Matsumura T, Kaneda Y, Yorek

MA, Beebe D, Oates PJ, Hammes HP et al (2000) Normalizing

mitochondrial superoxide production blocks three pathways of

hyperglycaemic damage. Nature 404: 787 – 790

Noghero A, Perino A, Seano G, Saglio E, Lo Sasso G, Veglio F, Primo L, Hirsch

E, Bussolino F, Morello F (2012) Liver X receptor activation reduces

angiogenesis by impairing lipid raft localization and signaling of vascular

endothelial growth factor receptor-2. Arterioscler Thromb Vasc Biol 32:

2280 – 2288

Obrosova IG, Minchenko AG, Vasupuram R, White L, Abatan OI, Kumagai AK,

Frank RN, Stevens MJ (2003) Aldose reductase inhibitor fidarestat prevents

retinal oxidative stress and vascular endothelial growth factor

overexpression in streptozotocin-diabetic rats. Diabetes 52: 864 – 871

Oyama T, Miyasita Y, Watanabe H, Shirai K (2006) The role of polyol pathway

in high glucose-induced endothelial cell damages. Diabetes Res Clin Pract

73: 227 – 234

Paik JY, Lee KH, Ko BH, Choe YS, Choi Y, Kim BT (2005) Nitric oxide stimulates

18F-FDG uptake in human endothelial cells through increased hexokinase

activity and GLUT1 expression. J Nucl Med 46: 365 – 370

Pan M, Wasa M, Ryan U, Souba W (1995) Inhibition of pulmonary

microvascular endothelial glutamine transport by glucocorticoids and

endotoxin. JPEN J Parenter Enteral Nutr 19: 477 – 481

Pan S, World CJ, Kovacs CJ, Berk BC (2009) Glucose 6-phosphate dehydrogenase

is regulated through c-Src-mediated tyrosine phosphorylation in

endothelial cells. Arterioscler Thromb Vasc Biol 29: 895 – 901.

Pandolfi PP, Sonati F, Rivi R, Mason P, Grosveld F, Luzzatto L (1995)

Targeted disruption of the housekeeping gene encoding glucose

6-phosphate dehydrogenase (G6PD): G6PD is dispensable for pentose

synthesis but essential for defense against oxidative stress. EMBO J 14:

5209 – 5215

Pangare M, Makino A (2012) Mitochondrial function in vascular endothelial

cell in diabetes. J Smooth Muscle Res 48: 1 – 26

Papetti M, Herman IM (2002) Mechanisms of normal and tumor-derived

angiogenesis. Am J Physiol Cell Physiol 282: C947 –C970

Parra-Bonilla G, Alvarez DF, Al-Mehdi AB, Alexeyev M, Stevens T (2010)

Critical role for lactate dehydrogenase A in aerobic glycolysis that sustains

pulmonary microvascular endothelial cell proliferation. Am J Physiol Lung

Cell Mol Physiol 299: L513 – L522

Patten IS, Rana S, Shahul S, Rowe GC, Jang C, Liu L, Hacker MR, Rhee JS,

Mitchell J, Mahmood F et al (2012) Cardiac angiogenic imbalance leads to

peripartum cardiomyopathy. Nature 485: 333 – 338

Pieper GM (1997) Acute amelioration of diabetic endothelial dysfunction with

a derivative of the nitric oxide synthase cofactor, tetrahydrobiopterin. J

Cardiovasc Pharmacol 29: 8 – 15

Pober JS, Min W, Bradley JR (2009) Mechanisms of endothelial dysfunction,

injury, and death. Annu Rev Pathol 4: 71 – 95

Potente M, Gerhardt H, Carmeliet P (2011) Basic and therapeutic aspects of

angiogenesis. Cell 146: 873 – 887

Puigserver P, Wu Z, Park CW, Graves R, Wright M, Spiegelman BM (1998) A

cold-inducible coactivator of nuclear receptors linked to adaptive

thermogenesis. Cell 92: 829 – 839

Romero MJ, Platt DH, Tawfik HE, Labazi M, El-Remessy AB, Bartoli M,

Caldwell RB, Caldwell RW (2008) Diabetes-induced coronary vascular

dysfunction involves increased arginase activity. Circ Res 102: 95 – 102

Rose ML, Madren J, Bunzendahl H, Thurman RG (1999) Dietary glycine

inhibits the growth of B16 melanoma tumors in mice. Carcinogenesis 20:

793 – 798

Rudarakanchana N, Flanagan JA, Chen H, Upton PD, Machado R, Patel D,

Trembath RC, Morrell NW (2002) Functional analysis of bone

morphogenetic protein type II receptor mutations underlying primary

pulmonary hypertension. Hum Mol Genet 11: 1517 – 1525

Ryoo S, Lemmon CA, Soucy KG, Gupta G, White AR, Nyhan D, Shoukas A,

Romer LH, Berkowitz DE (2006) Oxidized low-density

lipoprotein-dependent endothelial arginase II activation contributes to

impaired nitric oxide signaling. Circ Res 99: 951 – 960

Ryoo S, Gupta G, Benjo A, Lim HK, Camara A, Sikka G, Lim HK, Sohi J,

Santhanam L, Soucy K et al (2008) Endothelial arginase II: a novel target

for the treatment of atherosclerosis. Circ Res 102: 923 – 932

Sahoo S, Aurich MK, Jonsson JJ, Thiele I (2014) Membrane transporters in a

human genome-scale metabolic knowledgebase and their implications for

disease. Front Physiol 5: 91

Sanchez-Lozada LG, Lanaspa MA, Cristobal-Garcia M, Garcia-Arroyo F, Soto V,

Cruz-Robles D, Nakagawa T, Yu MA, Kang DH, Johnson RJ (2012) Uric

acid-induced endothelial dysfunction is associated with mitochondrial

alterations and decreased intracellular ATP concentrations. Nephron Exp

Nephrol 121: e71 – e78

Sautin YY, Johnson RJ (2008) Uric acid: the oxidant-antioxidant paradox.

Nucleosides, Nucleotides Nucleic Acids 27: 608 – 619

Sawada N, Jiang A, Takizawa F, Safdar A, Manika A, Tesmenitsky Y, Kang

KT, Bischoff J, Kalwa H, Sartoretto JL et al (2014) Endothelial

PGC-1alpha mediates vascular dysfunction in diabetes. Cell Metab 19:

246 – 258

Schoors S, Cantelmo AR, Georgiadou M, Stapor PC, Wang X, Wong BW, Bifari

F, Quaegebeur A, Decimo I, Schoonjans L et al (2014a) Incomplete and

transitory decrease of glycolysis: a new paradigm for anti-angiogenic

therapy? Cell cycle 13: 16 – 22

Schoors S, De Bock K, Cantelmo AR, Georgiadou M, Ghesquiere B,

Cauwenberghs S, Kuchnio A, Wong BW, Quaegebeur A, Goveia J et al

(2014b) Partial and transient reduction of glycolysis by PFKFB3 blockade

reduces pathological angiogenesis. Cell Metab 19: 37 – 48

Sena CM, Matafome P, Crisostomo J, Rodrigues L, Fernandes R, Pereira P,

Seica RM (2012) Methylglyoxal promotes oxidative stress and endothelial

dysfunction. Pharmacol Res 65: 497 – 506

Sennino B, McDonald DM (2012) Controlling escape from angiogenesis

inhibitors. Nat Rev 12: 699 – 709

Sessa WC, Hecker M, Mitchell JA, Vane JR (1990) The metabolism of

L-arginine and its significance for the biosynthesis of

endothelium-derived relaxing factor: L-glutamine inhibits the generation

of L-arginine by cultured endothelial cells. Proc Natl Acad Sci USA 87:

8607 – 8611

Sonveaux P, Copetti T, De Saedeleer CJ, Vegran F, Verrax J, Kennedy KM,

Moon EJ, Dhup S, Danhier P, Frerart F et al (2012) Targeting the lactate

EMBO Molecular Medicine Vol 6 | No 9 | 2014 ª 2014 The Authors

EMBO Molecular Medicine Targeting EC metabolism in disease Jermaine Goveia et al

1118

transporter MCT1 in endothelial cells inhibits lactate-induced HIF-1

activation and tumor angiogenesis. PLoS ONE 7: e33418

Spolarics Z, Spitzer JJ (1993) Augmented glucose use and pentose cycle

activity in hepatic endothelial cells after in vivo endotoxemia. Hepatology

17: 615 – 620

Spolarics Z, Wu JX (1997) Role of glutathione and catalase in H2O2

detoxification in LPS-activated hepatic endothelial and Kupffer cells. Am J

Physiol 273: G1304 –G1311

Stobart JL, Lu L, Anderson HD, Mori H, Anderson CM (2013) Astrocyte-induced

cortical vasodilation is mediated by D-serine and endothelial nitric oxide

synthase. Proc Natl Acad Sci USA 110: 3149 – 3154

Stroes E, Kastelein J, Cosentino F, Erkelens W, Wever R, Koomans H, Luscher

T, Rabelink T (1997) Tetrahydrobiopterin restores endothelial function in

hypercholesterolemia. J Clin Invest 99: 41 – 46

Stroes E, Hijmering M, van Zandvoort M, Wever R, Rabelink TJ, van Faassen

EE (1998) Origin of superoxide production by endothelial nitric oxide

synthase. FEBS Lett 438: 161 – 164

Sugiyama T, Levy BD, Michel T (2009) Tetrahydrobiopterin recycling, a key

determinant of endothelial nitric-oxide synthase-dependent signaling

pathways in cultured vascular endothelial cells. J Biol Chem 284:

12691 – 12700

Sutendra G, Michelakis ED (2014) The metabolic basis of pulmonary arterial

hypertension. Cell Metab 19: 558 – 573

Takemoto M, Sun J, Hiroki J, Shimokawa H, Liao JK (2002) Rho-kinase

mediates hypoxia-induced downregulation of endothelial nitric oxide

synthase. Circulation 106: 57 – 62

Tammali R, Reddy AB, Srivastava SK, Ramana KV (2011) Inhibition of aldose

reductase prevents angiogenesis in vitro and in vivo. Angiogenesis 14:

209 – 221

Thengchaisri N, Hein TW, Wang W, Xu X, Li Z, Fossum TW, Kuo L (2006)

Upregulation of arginase by H2O2 impairs endothelium-dependent nitric

oxide-mediated dilation of coronary arterioles. Arterioscler Thromb Vasc

Biol 26: 2035 – 2042

Thornalley PJ (1993) The glyoxalase system in health and disease. Mol Aspects

Med 14: 287 – 371

Tibbetts AS, Appling DR (2010) Compartmentalization of Mammalian

folate-mediated one-carbon metabolism. Annu Rev Nutr 30: 57 – 81

Topal G, Brunet A, Millanvoye E, Boucher JL, Rendu F, Devynck MA,

David-Dufilho M (2004) Homocysteine induces oxidative stress by

uncoupling of NO synthase activity through reduction of

tetrahydrobiopterin. Free Radical Biol Med 36: 1532 – 1541

Tuder RM, Davis LA, Graham BB (2012) Targeting energetic metabolism: a

new frontier in the pathogenesis and treatment of pulmonary

hypertension. Am J Respir Crit Care Med 185: 260 – 266

Van Schaftingen E, Lederer B, Bartrons R, Hers HG (1982) A kinetic study of

pyrophosphate: fructose-6-phosphate phosphotransferase from potato

tubers. Application to a microassay of fructose 2,6-bisphosphate. Eur J

Biochem 129: 191 – 195

Vedantham S, Noh H, Ananthakrishnan R, Son N, Hallam K, Hu Y, Yu S, Shen

X, Rosario R, Lu Y et al (2011) Human aldose reductase expression

accelerates atherosclerosis in diabetic apolipoprotein E-/- mice. Arterioscler

Thromb Vasc Biol 31: 1805 – 1813

Vegran F, Boidot R, Michiels C, Sonveaux P, Feron O (2011) Lactate influx

through the endothelial cell monocarboxylate transporter MCT1 supports

an NF-kappaB/IL-8 pathway that drives tumor angiogenesis. Cancer Res

71: 2550 – 2560

Verhaar MC, Stroes E, Rabelink TJ (2002) Folates and cardiovascular disease.

Arterioscler Thromb Vasc Biol 22: 6 – 13

Versari D, Daghini E, Virdis A, Ghiadoni L, Taddei S (2009) Endothelial

dysfunction as a target for prevention of cardiovascular disease. Diabetes

Care 32(Suppl 2): S314 – S321

Vizan P, Sanchez-Tena S, Alcarraz-Vizan G, Soler M, Messeguer R, Pujol MD,

Lee WN, Cascante M (2009) Characterization of the metabolic changes

underlying growth factor angiogenic activation: identification of new

potential therapeutic targets. Carcinogenesis 30: 946 – 952

Vlassara H, Fuh H, Donnelly T, Cybulsky M (1995) Advanced glycation

endproducts promote adhesion molecule (VCAM-1, ICAM-1) expression

and atheroma formation in normal rabbits. Mol Med 1: 447 – 456

Wautier JL, Schmidt AM (2004) Protein glycation: a firm link to endothelial

cell dysfunction. Circ Res 95: 233 – 238

Weinberg JM, Varani J, Johnson KJ, Roeser NF, Dame MK, Davis JA,

Venkatachalam MA (1992) Protection of human umbilical vein

endothelial cells by glycine and structurally similar amino acids against

calcium and hydrogen peroxide-induced lethal cell injury. Am J Pathol

140: 457 – 471

Welti J, Loges S, Dimmeler S, Carmeliet P (2013) Recent molecular discoveries

in angiogenesis and antiangiogenic therapies in cancer. J Clin Invest 123:

3190 – 3200

Whetzel AM, Sturek JM, Nagelin MH, Bolick DT, Gebre AK, Parks JS, Bruce AC,

Skaflen MD, Hedrick CC (2010) ABCG1 deficiency in mice promotes

endothelial activation and monocyte-endothelial interactions. Arterioscler

Thromb Vasc Biol 30: 809 – 817

Wu G, Meininger CJ (1995) Impaired arginine metabolism and NO synthesis

in coronary endothelial cells of the spontaneously diabetic BB rat. Am J

Physiol 269: H1312 –H1318

Xu W, Kaneko FT, Zheng S, Comhair SA, Janocha AJ, Goggans T, Thunnissen

FB, Farver C, Hazen SL, Jennings C et al (2004) Increased arginase II and

decreased NO synthesis in endothelial cells of patients with pulmonary

arterial hypertension. FASEB J 18: 1746 – 1748

Xu W, Koeck T, Lara AR, Neumann D, DiFilippo FP, Koo M, Janocha AJ, Masri

FA, Arroliga AC, Jennings C et al (2007) Alterations of cellular bioenergetics

in pulmonary artery endothelial cells. Proc Natl Acad Sci USA 104:

1342 – 1347

Xu J, Dang Y, Ren YR, Liu JO (2010) Cholesterol trafficking is required for

mTOR activation in endothelial cells. Proc Natl Acad Sci USA 107:

4764 – 4769

Xu W, Erzurum SC (2011) Endothelial cell energy metabolism, proliferation,

and apoptosis in pulmonary hypertension. Compr Physiol 1: 357 – 372

Xu Y, An X, Guo X, Habtetsion TG, Wang Y, Xu X, Kandala S, Li Q, Li H, Zhang

C et al (2014) Endothelial 6-Phosphofructo-2-Kinase/Fructose-2,

6-Bisphosphatase, Isoform 3 Plays a Critical Role in Angiogenesis.

Arterioscler Thromb Vasc Biol 34: 1231 – 1239

Yadav UC, Srivastava SK, Ramana KV (2012) Prevention of VEGF-induced

growth and tube formation in human retinal endothelial cells by aldose

reductase inhibition. J Diabetes Complications 26: 369 – 377

Yamashina S, Konno A, Wheeler MD, Rusyn I, Rusyn EV, Cox AD, Thurman RG

(2001) Endothelial cells contain a glycine-gated chloride channel. Nutr

Cancer 40: 197 – 204

Yan SF, Ramasamy R, Schmidt AM (2008) Mechanisms of disease: advanced

glycation end-products and their receptor in inflammation and diabetes

complications. Nat Clin Pract Endocrinol Metab 4: 285 – 293

Yang Q, Bailey L, Clarke R, Flanders WD, Liu T, Yesupriya A, Khoury MJ,

Friedman JM (2012) Prospective study of methylenetetrahydrofolate

reductase (MTHFR) variant C677T and risk of all-cause and

cardiovascular disease mortality among 6000 US adults. Am J Clin Nutr

95: 1245 – 1253

ª 2014 The Authors EMBO Molecular Medicine Vol 6 | No 9 | 2014

Jermaine Goveia et al Targeting EC metabolism in disease EMBO Molecular Medicine

1119

Ming XF, Barandier C, Viswambharan H, Kwak BR, Mach F, Mazzolai L, Hayoz

D, Ruffieux J, Rusconi S, Montani JP et al (2004) Thrombin stimulates

human endothelial arginase enzymatic activity via RhoA/ROCK pathway:

implications for atherosclerotic endothelial dysfunction. Circulation 110:

3708 – 3714

Mishra RC, Tripathy S, Desai KM, Quest D, Lu Y, Akhtar J, Gopalakrishnan V

(2008) Nitric oxide synthase inhibition promotes endothelium-dependent

vasodilatation and the antihypertensive effect of L-serine. Hypertension 51:

791 – 796

Morrison RF, Seidel ER (1995) Vascular endothelial cell proliferation:

regulation of cellular polyamines. Cardiovasc Res 29: 841 – 847

Mugoni V, Postel R, Catanzaro V, De Luca E, Turco E, Digilio G, Silengo L,

Murphy MP, Medana C, Stainier DY et al (2013) Ubiad1 is an antioxidant

enzyme that regulates eNOS activity by CoQ10 synthesis. Cell 152: 504 – 518

Nishikawa T, Edelstein D, Du XL, Yamagishi S, Matsumura T, Kaneda Y, Yorek

MA, Beebe D, Oates PJ, Hammes HP et al (2000) Normalizing

mitochondrial superoxide production blocks three pathways of

hyperglycaemic damage. Nature 404: 787 – 790

Noghero A, Perino A, Seano G, Saglio E, Lo Sasso G, Veglio F, Primo L, Hirsch

E, Bussolino F, Morello F (2012) Liver X receptor activation reduces

angiogenesis by impairing lipid raft localization and signaling of vascular

endothelial growth factor receptor-2. Arterioscler Thromb Vasc Biol 32:

2280 – 2288

Obrosova IG, Minchenko AG, Vasupuram R, White L, Abatan OI, Kumagai AK,

Frank RN, Stevens MJ (2003) Aldose reductase inhibitor fidarestat prevents

retinal oxidative stress and vascular endothelial growth factor

overexpression in streptozotocin-diabetic rats. Diabetes 52: 864 – 871

Oyama T, Miyasita Y, Watanabe H, Shirai K (2006) The role of polyol pathway

in high glucose-induced endothelial cell damages. Diabetes Res Clin Pract

73: 227 – 234

Paik JY, Lee KH, Ko BH, Choe YS, Choi Y, Kim BT (2005) Nitric oxide stimulates

18F-FDG uptake in human endothelial cells through increased hexokinase

activity and GLUT1 expression. J Nucl Med 46: 365 – 370

Pan M, Wasa M, Ryan U, Souba W (1995) Inhibition of pulmonary

microvascular endothelial glutamine transport by glucocorticoids and

endotoxin. JPEN J Parenter Enteral Nutr 19: 477 – 481

Pan S, World CJ, Kovacs CJ, Berk BC (2009) Glucose 6-phosphate dehydrogenase

is regulated through c-Src-mediated tyrosine phosphorylation in

endothelial cells. Arterioscler Thromb Vasc Biol 29: 895 – 901.

Pandolfi PP, Sonati F, Rivi R, Mason P, Grosveld F, Luzzatto L (1995)

Targeted disruption of the housekeeping gene encoding glucose

6-phosphate dehydrogenase (G6PD): G6PD is dispensable for pentose

synthesis but essential for defense against oxidative stress. EMBO J 14:

5209 – 5215

Pangare M, Makino A (2012) Mitochondrial function in vascular endothelial

cell in diabetes. J Smooth Muscle Res 48: 1 – 26

Papetti M, Herman IM (2002) Mechanisms of normal and tumor-derived

angiogenesis. Am J Physiol Cell Physiol 282: C947 –C970

Parra-Bonilla G, Alvarez DF, Al-Mehdi AB, Alexeyev M, Stevens T (2010)

Critical role for lactate dehydrogenase A in aerobic glycolysis that sustains

pulmonary microvascular endothelial cell proliferation. Am J Physiol Lung

Cell Mol Physiol 299: L513 – L522

Patten IS, Rana S, Shahul S, Rowe GC, Jang C, Liu L, Hacker MR, Rhee JS,

Mitchell J, Mahmood F et al (2012) Cardiac angiogenic imbalance leads to

peripartum cardiomyopathy. Nature 485: 333 – 338

Pieper GM (1997) Acute amelioration of diabetic endothelial dysfunction with

a derivative of the nitric oxide synthase cofactor, tetrahydrobiopterin. J

Cardiovasc Pharmacol 29: 8 – 15

Pober JS, Min W, Bradley JR (2009) Mechanisms of endothelial dysfunction,

injury, and death. Annu Rev Pathol 4: 71 – 95

Potente M, Gerhardt H, Carmeliet P (2011) Basic and therapeutic aspects of

angiogenesis. Cell 146: 873 – 887

Puigserver P, Wu Z, Park CW, Graves R, Wright M, Spiegelman BM (1998) A

cold-inducible coactivator of nuclear receptors linked to adaptive

thermogenesis. Cell 92: 829 – 839

Romero MJ, Platt DH, Tawfik HE, Labazi M, El-Remessy AB, Bartoli M,

Caldwell RB, Caldwell RW (2008) Diabetes-induced coronary vascular

dysfunction involves increased arginase activity. Circ Res 102: 95 – 102

Rose ML, Madren J, Bunzendahl H, Thurman RG (1999) Dietary glycine

inhibits the growth of B16 melanoma tumors in mice. Carcinogenesis 20:

793 – 798

Rudarakanchana N, Flanagan JA, Chen H, Upton PD, Machado R, Patel D,

Trembath RC, Morrell NW (2002) Functional analysis of bone

morphogenetic protein type II receptor mutations underlying primary

pulmonary hypertension. Hum Mol Genet 11: 1517 – 1525

Ryoo S, Lemmon CA, Soucy KG, Gupta G, White AR, Nyhan D, Shoukas A,

Romer LH, Berkowitz DE (2006) Oxidized low-density

lipoprotein-dependent endothelial arginase II activation contributes to

impaired nitric oxide signaling. Circ Res 99: 951 – 960

Ryoo S, Gupta G, Benjo A, Lim HK, Camara A, Sikka G, Lim HK, Sohi J,

Santhanam L, Soucy K et al (2008) Endothelial arginase II: a novel target

for the treatment of atherosclerosis. Circ Res 102: 923 – 932

Sahoo S, Aurich MK, Jonsson JJ, Thiele I (2014) Membrane transporters in a

human genome-scale metabolic knowledgebase and their implications for

disease. Front Physiol 5: 91

Sanchez-Lozada LG, Lanaspa MA, Cristobal-Garcia M, Garcia-Arroyo F, Soto V,

Cruz-Robles D, Nakagawa T, Yu MA, Kang DH, Johnson RJ (2012) Uric

acid-induced endothelial dysfunction is associated with mitochondrial

alterations and decreased intracellular ATP concentrations. Nephron Exp

Nephrol 121: e71 – e78

Sautin YY, Johnson RJ (2008) Uric acid: the oxidant-antioxidant paradox.

Nucleosides, Nucleotides Nucleic Acids 27: 608 – 619

Sawada N, Jiang A, Takizawa F, Safdar A, Manika A, Tesmenitsky Y, Kang

KT, Bischoff J, Kalwa H, Sartoretto JL et al (2014) Endothelial

PGC-1alpha mediates vascular dysfunction in diabetes. Cell Metab 19:

246 – 258

Schoors S, Cantelmo AR, Georgiadou M, Stapor PC, Wang X, Wong BW, Bifari

F, Quaegebeur A, Decimo I, Schoonjans L et al (2014a) Incomplete and

transitory decrease of glycolysis: a new paradigm for anti-angiogenic

therapy? Cell cycle 13: 16 – 22

Schoors S, De Bock K, Cantelmo AR, Georgiadou M, Ghesquiere B,

Cauwenberghs S, Kuchnio A, Wong BW, Quaegebeur A, Goveia J et al

(2014b) Partial and transient reduction of glycolysis by PFKFB3 blockade

reduces pathological angiogenesis. Cell Metab 19: 37 – 48

Sena CM, Matafome P, Crisostomo J, Rodrigues L, Fernandes R, Pereira P,

Seica RM (2012) Methylglyoxal promotes oxidative stress and endothelial

dysfunction. Pharmacol Res 65: 497 – 506

Sennino B, McDonald DM (2012) Controlling escape from angiogenesis

inhibitors. Nat Rev 12: 699 – 709

Sessa WC, Hecker M, Mitchell JA, Vane JR (1990) The metabolism of

L-arginine and its significance for the biosynthesis of

endothelium-derived relaxing factor: L-glutamine inhibits the generation

of L-arginine by cultured endothelial cells. Proc Natl Acad Sci USA 87:

8607 – 8611

Sonveaux P, Copetti T, De Saedeleer CJ, Vegran F, Verrax J, Kennedy KM,

Moon EJ, Dhup S, Danhier P, Frerart F et al (2012) Targeting the lactate

EMBO Molecular Medicine Vol 6 | No 9 | 2014 ª 2014 The Authors

EMBO Molecular Medicine Targeting EC metabolism in disease Jermaine Goveia et al

1118

transporter MCT1 in endothelial cells inhibits lactate-induced HIF-1

activation and tumor angiogenesis. PLoS ONE 7: e33418

Spolarics Z, Spitzer JJ (1993) Augmented glucose use and pentose cycle

activity in hepatic endothelial cells after in vivo endotoxemia. Hepatology

17: 615 – 620

Spolarics Z, Wu JX (1997) Role of glutathione and catalase in H2O2

detoxification in LPS-activated hepatic endothelial and Kupffer cells. Am J

Physiol 273: G1304 –G1311

Stobart JL, Lu L, Anderson HD, Mori H, Anderson CM (2013) Astrocyte-induced

cortical vasodilation is mediated by D-serine and endothelial nitric oxide

synthase. Proc Natl Acad Sci USA 110: 3149 – 3154

Stroes E, Kastelein J, Cosentino F, Erkelens W, Wever R, Koomans H, Luscher

T, Rabelink T (1997) Tetrahydrobiopterin restores endothelial function in

hypercholesterolemia. J Clin Invest 99: 41 – 46

Stroes E, Hijmering M, van Zandvoort M, Wever R, Rabelink TJ, van Faassen

EE (1998) Origin of superoxide production by endothelial nitric oxide

synthase. FEBS Lett 438: 161 – 164

Sugiyama T, Levy BD, Michel T (2009) Tetrahydrobiopterin recycling, a key

determinant of endothelial nitric-oxide synthase-dependent signaling

pathways in cultured vascular endothelial cells. J Biol Chem 284:

12691 – 12700

Sutendra G, Michelakis ED (2014) The metabolic basis of pulmonary arterial

hypertension. Cell Metab 19: 558 – 573

Takemoto M, Sun J, Hiroki J, Shimokawa H, Liao JK (2002) Rho-kinase

mediates hypoxia-induced downregulation of endothelial nitric oxide

synthase. Circulation 106: 57 – 62

Tammali R, Reddy AB, Srivastava SK, Ramana KV (2011) Inhibition of aldose

reductase prevents angiogenesis in vitro and in vivo. Angiogenesis 14:

209 – 221

Thengchaisri N, Hein TW, Wang W, Xu X, Li Z, Fossum TW, Kuo L (2006)

Upregulation of arginase by H2O2 impairs endothelium-dependent nitric

oxide-mediated dilation of coronary arterioles. Arterioscler Thromb Vasc

Biol 26: 2035 – 2042

Thornalley PJ (1993) The glyoxalase system in health and disease. Mol Aspects

Med 14: 287 – 371

Tibbetts AS, Appling DR (2010) Compartmentalization of Mammalian

folate-mediated one-carbon metabolism. Annu Rev Nutr 30: 57 – 81

Topal G, Brunet A, Millanvoye E, Boucher JL, Rendu F, Devynck MA,

David-Dufilho M (2004) Homocysteine induces oxidative stress by

uncoupling of NO synthase activity through reduction of

tetrahydrobiopterin. Free Radical Biol Med 36: 1532 – 1541

Tuder RM, Davis LA, Graham BB (2012) Targeting energetic metabolism: a

new frontier in the pathogenesis and treatment of pulmonary

hypertension. Am J Respir Crit Care Med 185: 260 – 266

Van Schaftingen E, Lederer B, Bartrons R, Hers HG (1982) A kinetic study of

pyrophosphate: fructose-6-phosphate phosphotransferase from potato

tubers. Application to a microassay of fructose 2,6-bisphosphate. Eur J

Biochem 129: 191 – 195

Vedantham S, Noh H, Ananthakrishnan R, Son N, Hallam K, Hu Y, Yu S, Shen

X, Rosario R, Lu Y et al (2011) Human aldose reductase expression

accelerates atherosclerosis in diabetic apolipoprotein E-/- mice. Arterioscler

Thromb Vasc Biol 31: 1805 – 1813

Vegran F, Boidot R, Michiels C, Sonveaux P, Feron O (2011) Lactate influx

through the endothelial cell monocarboxylate transporter MCT1 supports

an NF-kappaB/IL-8 pathway that drives tumor angiogenesis. Cancer Res

71: 2550 – 2560

Verhaar MC, Stroes E, Rabelink TJ (2002) Folates and cardiovascular disease.

Arterioscler Thromb Vasc Biol 22: 6 – 13

Versari D, Daghini E, Virdis A, Ghiadoni L, Taddei S (2009) Endothelial

dysfunction as a target for prevention of cardiovascular disease. Diabetes

Care 32(Suppl 2): S314 – S321

Vizan P, Sanchez-Tena S, Alcarraz-Vizan G, Soler M, Messeguer R, Pujol MD,

Lee WN, Cascante M (2009) Characterization of the metabolic changes

underlying growth factor angiogenic activation: identification of new

potential therapeutic targets. Carcinogenesis 30: 946 – 952

Vlassara H, Fuh H, Donnelly T, Cybulsky M (1995) Advanced glycation

endproducts promote adhesion molecule (VCAM-1, ICAM-1) expression

and atheroma formation in normal rabbits. Mol Med 1: 447 – 456

Wautier JL, Schmidt AM (2004) Protein glycation: a firm link to endothelial

cell dysfunction. Circ Res 95: 233 – 238

Weinberg JM, Varani J, Johnson KJ, Roeser NF, Dame MK, Davis JA,

Venkatachalam MA (1992) Protection of human umbilical vein

endothelial cells by glycine and structurally similar amino acids against

calcium and hydrogen peroxide-induced lethal cell injury. Am J Pathol

140: 457 – 471

Welti J, Loges S, Dimmeler S, Carmeliet P (2013) Recent molecular discoveries

in angiogenesis and antiangiogenic therapies in cancer. J Clin Invest 123:

3190 – 3200

Whetzel AM, Sturek JM, Nagelin MH, Bolick DT, Gebre AK, Parks JS, Bruce AC,

Skaflen MD, Hedrick CC (2010) ABCG1 deficiency in mice promotes

endothelial activation and monocyte-endothelial interactions. Arterioscler

Thromb Vasc Biol 30: 809 – 817

Wu G, Meininger CJ (1995) Impaired arginine metabolism and NO synthesis

in coronary endothelial cells of the spontaneously diabetic BB rat. Am J

Physiol 269: H1312 –H1318

Xu W, Kaneko FT, Zheng S, Comhair SA, Janocha AJ, Goggans T, Thunnissen

FB, Farver C, Hazen SL, Jennings C et al (2004) Increased arginase II and

decreased NO synthesis in endothelial cells of patients with pulmonary

arterial hypertension. FASEB J 18: 1746 – 1748

Xu W, Koeck T, Lara AR, Neumann D, DiFilippo FP, Koo M, Janocha AJ, Masri

FA, Arroliga AC, Jennings C et al (2007) Alterations of cellular bioenergetics

in pulmonary artery endothelial cells. Proc Natl Acad Sci USA 104:

1342 – 1347

Xu J, Dang Y, Ren YR, Liu JO (2010) Cholesterol trafficking is required for

mTOR activation in endothelial cells. Proc Natl Acad Sci USA 107:

4764 – 4769

Xu W, Erzurum SC (2011) Endothelial cell energy metabolism, proliferation,

and apoptosis in pulmonary hypertension. Compr Physiol 1: 357 – 372

Xu Y, An X, Guo X, Habtetsion TG, Wang Y, Xu X, Kandala S, Li Q, Li H, Zhang

C et al (2014) Endothelial 6-Phosphofructo-2-Kinase/Fructose-2,

6-Bisphosphatase, Isoform 3 Plays a Critical Role in Angiogenesis.

Arterioscler Thromb Vasc Biol 34: 1231 – 1239

Yadav UC, Srivastava SK, Ramana KV (2012) Prevention of VEGF-induced

growth and tube formation in human retinal endothelial cells by aldose

reductase inhibition. J Diabetes Complications 26: 369 – 377

Yamashina S, Konno A, Wheeler MD, Rusyn I, Rusyn EV, Cox AD, Thurman RG

(2001) Endothelial cells contain a glycine-gated chloride channel. Nutr

Cancer 40: 197 – 204

Yan SF, Ramasamy R, Schmidt AM (2008) Mechanisms of disease: advanced

glycation end-products and their receptor in inflammation and diabetes

complications. Nat Clin Pract Endocrinol Metab 4: 285 – 293

Yang Q, Bailey L, Clarke R, Flanders WD, Liu T, Yesupriya A, Khoury MJ,

Friedman JM (2012) Prospective study of methylenetetrahydrofolate

reductase (MTHFR) variant C677T and risk of all-cause and

cardiovascular disease mortality among 6000 US adults. Am J Clin Nutr

95: 1245 – 1253

ª 2014 The Authors EMBO Molecular Medicine Vol 6 | No 9 | 2014

Jermaine Goveia et al Targeting EC metabolism in disease EMBO Molecular Medicine

1119

Yeh WL, Lin CJ, Fu WM (2008) Enhancement of glucose transporter

expression of brain endothelial cells by vascular endothelial growth factor

derived from glioma exposed to hypoxia. Mol Pharmacol 73: 170 – 177

Yoshida A, Wei D, Nomura W, Izawa S, Inoue Y (2012) Reduction of glucose

uptake through inhibition of hexose transporters and enhancement of

their endocytosis by methylglyoxal in Saccharomyces cerevisiae. J Biol

Chem 287: 701 – 711

Zeng L, Liao H, Liu Y, Lee TS, Zhu M, Wang X, Stemerman MB, Zhu Y, Shyy JY

(2004) Sterol-responsive element-binding protein (SREBP) 2

down-regulates ATP-binding cassette transporter A1 in vascular

endothelial cells: a novel role of SREBP in regulating cholesterol

metabolism. J Biol Chem 279: 48801 – 48807

Zhang Z, Yang Z, Zhu B, Hu J, Liew CW, Zhang Y, Leopold JA, Handy DE,

Loscalzo J, Stanton RC (2012) Increasing glucose 6-phosphate

dehydrogenase activity restores redox balance in vascular endothelial cells

exposed to high glucose. PLoS One 7: e49128

Zhang Y, Hong Q, Huang Z, Xue P, Lv Y, Fu B, Chen X, Wu D (2014) ALDR

enhanced endothelial injury in hyperuricemia screened using SILAC. Cell

Physiol Biochem 33: 479 – 490

Zhao Y, Peng J, Lu C, Hsin M, Mura M, Wu L, Chu L, Zamel R, Machuca T,

Waddell T et al (2014) Metabolomic heterogeneity of pulmonary arterial

hypertension. PLoS ONE 9: e88727

Zharikov S, Krotova K, Hu H, Baylis C, Johnson RJ, Block ER, Patel J (2008) Uric

acid decreases NO production and increases arginase activity in cultured

pulmonary artery endothelial cells. Am J Physiol Cell Physiol 295:

C1183 –C1190

License: This is an open access article under the

terms of the Creative Commons Attribution 4.0

License, which permits use, distribution and reproduc-

tion in any medium, provided the original work is

properly cited.

1120 EMBO Molecular Medicine Vol 6 | No 9 | 2014 ª 2014 The Authors

EMBO Molecular Medicine Targeting EC metabolism in disease Jermaine Goveia et al

Closeup

The heart-liver metabolic axis: defectivecommunication exacerbates diseaseKedryn K Baskin, Angie L Bookout & Eric N Olson

The heart has been recognized as an endo-crine organ for over 30 years (de Bold,2011); however, little is known about howthe heart communicates with other organsin the body, and even less is known aboutthis process in the diseased heart. In thisissue of EMBO Molecular Medicine, Magidaand Leinwand (2014) introduce the conceptthat a primary genetic defect in the heartresults in aberrant hepatic lipid metabo-lism, which consequently exacerbateshypertrophic cardiomyopathy (HCM). Thisstudy provides evidence in support of thehypothesis that crosstalk occurs betweenthe heart and liver, and that this becomesdisrupted in the diseased state.

See also: JAMagida& LA Leinwand (April 2014)

H CM is an inherited cardiovascular

disease primarily caused by muta-

tions in genes encoding proteins in

the sarcomere, the contractile apparatus of

cardiac myocytes. HCM is characterized by

increased heart mass and abnormal cardiac

function with susceptibility to arrhythmias

and sudden cardiac death. Histological

manifestations of the disease include cardiac

myocyte hypertrophy, myocardial fibrosis,

extracellular matrix disorganization, and

myocyte disarray. While many affected indi-

viduals are asymptomatic and remain undi-

agnosed, HCM is the most frequent cause of

sudden death in young athletes (Seidman &

Seidman, 2011; Maron & Maron, 2013).

To date, 13 genes containing over 900

distinct mutations have been identified as

genetic causes of HCM. Most of these genes

encode for proteins found within the thick

and thin filaments of sarcomeres, such as

b-myosin heavy chain (MYH7) and troponin

T (TTNT2). Mutations in MYH7 increase

force generation and actin-myosin sliding

velocity within sarcomeres. These findings

indicate that genetic mutations in HCM

patients are the primary cause of cardiac

hypertrophy (Wang et al, 2010).

Numerous animal models have been

generated to investigate HCM (Maass &

Leinwand, 2000), and much focus has been

given to an R403Q mutation in MYH7, which

causes an especially severe clinical pheno-

type (Seidman & Seidman, 2011). While the

various animal models of R403Q highlight

different aspects of HCM, they share

common traits of HCM including cardiac

hypertrophy and fibrosis (Maass & Leinw-

and, 2000). An interesting, and poorly

understood characteristic of hypertrophic

cardiomyopathy, as opposed to other types

of cardiomyopathies, is that systemic meta-

bolic alterations occur secondary to the

cardiomyopathy (Maron & Maron, 2013).

This is recapitulated in the R403Q model

used in the study published by Magida and

Leinwand (2014).

Clinical studies have revealed that HCM

patients harboring mutations in sarcomeric

genes present with deficient cardiac energet-

ics (Crilley et al, 2003). In the present study,

the authors demonstrate that the R403Q

HCM mouse model has diminished cardiac

ATP levels and impaired lipid utilization in

the heart, assessed by decreased cardiac

triglycerides and fatty acid content, and

decreased expression of fatty acid translo-

case (CD36), lipoprotein lipase (LPL), and

very low density lipoprotein receptor

(VLDLR). Notably, they observed an approx-

imate two-fold reduction in active CD36

protein at the plasma membrane, coupled

with a similarly decreased level of nonesteri-

fied fatty acid (NEFA) released from VLDL

by the left ventricle. The authors suggest

that this decreased lipid uptake in the heart

leads to the observed lipid elevation in the

plasma, ultimately resulting in hepatic lipid

accumulation and pathologically enhanced

gluconeogenesis. The authors propose that

this elevation in hepatic glucose production

creates a vicious cycle between the heart

and the liver in which the spillover of VLDL

triglyceride and oleic acid from the heart

insults the liver via elevated protein kinase

C signaling. The liver responds by increasing

blood glucose levels leading to exacerbation

of the primary cardiac disease (summarized

in Fig 1). Importantly, features of the dis-

eased state can be rescued either by restor-

ing the energetic deficit at the level of the

cardiomyocyte via AMPK agonism, or by

blocking the deleterious elevation in hepatic

glucose output using the phosphoenol-

pyruvate carboxykinase (PEPCK) inhibitor

3-MPA (Magida & Leinwand, 2014).

......................................................

“These findings raise the inter-esting concept that the lack ofuse of a specific metabolic sub-strate by one tissue directlyaffects another”......................................................

These findings raise the interesting con-

cept that the lack of use of a specific meta-

bolic substrate by one tissue directly affects

another, perhaps revealing an inter-tissue

homeostatic feedback mechanism. Namely,

Department of Molecular Biology, University of Texas Southwestern Medical Center, Dallas, TX, USA. E-mail: eric.olson@utsouthwestern.eduDOI 10.1002/emmm.201303800 | Published online 12 March 2014

EMBO Molecular Medicine Vol 6 | No 4 | 2014 ª 2014 The Authors. Published under the terms of the CC BY license.436

relationship between sarcomeric structural

integrity and metabolically-derived energy at

the organismal level, and opens up many

more avenues for future investigation.

Conflict of interestThe authors declare that they have no con-

flict of interest.

ReferencesBhatia LS, Curzen NP, Calder PC, Byrne CD (2012)

Non-alcoholic fatty liver disease: a new and

important cardiovascular risk factor? Eur Heart

J 33: 1190 – 1200

Blad CC, Tang C, Offermanns S (2012) G

protein-coupled receptors for energy

metabolites as new therapeutic targets. Nat

Rev Drug Discov 11: 603 – 619

de Bold AJ (2011) Thirty years of research on atrial

natriuretic factor: historical background and

emerging concepts. Can J Physiol Pharmacol 89:

527 – 531

Bordicchia M, Liu D, Amri EZ, Ailhaud G,

Dessi-Fulgheri P, Zhang C, Takahashi N, Sarzani

R, Collins S (2012) Cardiac natriuretic peptides

act via p38 MAPK to induce the brown fat

thermogenic program in mouse and human

adipocytes. J Clin Invest 122: 1022 – 1036

Bryzgalova G, Lundholm L, Portwood N,

Gustafsson JA, Khan A, Efendic S,

Dahlman-Wright K (2008) Mechanisms of

antidiabetogenic and body weight-lowering

effects of estrogen in high-fat diet-fed

mice. Am J Physiol Endocrinol Metab 295:

E904 – E912

Crilley JG, Boehm EA, Blair E, Rajagopalan B,

Blamire AM, Styles P, McKenna WJ,

Ostman-Smith I, Clarke K, Watkins H (2003)

Hypertrophic cardiomyopathy due to

sarcomeric gene mutations is characterized by

impaired energy metabolism irrespective of the

degree of hypertrophy. J Am Coll Cardiol 41:

1776 – 1782

D’Eon TM, Souza SC, Aronovitz M, Obin MS, Fried

SK, Greenberg AS (2005) Estrogen regulation of

adiposity and fuel partitioning. Evidence of

genomic and non-genomic regulation of

lipogenic and oxidative pathways. J Biol Chem

280: 35983 – 35991

Grueter CE, van Rooij E, Johnson BA, DeLeon SM,

Sutherland LB, Qi X, Gautron L, Elmquist JK,

Bassel-Duby R, Olson EN (2012) A cardiac

microRNA governs systemic energy homeostasis

by regulation of MED13. Cell 149: 671 – 683

Liu S, Brown JD, Stanya KJ, Homan E, Leidl M,

Inouye K, Bhargava P, Gangl MR, Dai L, Hatano

B et al (2013) A diurnal serum lipid integrates

hepatic lipogenesis and peripheral fatty acid

use. Nature 502: 550 – 554

Maass A, Leinwand LA (2000) Animal models of

hypertrophic cardiomyopathy. Curr Opin Cardiol

15: 189 – 196

Magida JA, Leinwand LA (2014) Metabolic

crosstalk between the heart and liver

impacts familial hypertrophic cardiomyopathy.

EMBO Mol Med 6: 482 – 495

Maron BJ, Maron MS (2013) Hypertrophic

cardiomyopathy. Lancet 381: 242 – 255

Rashed HM, Nair BG, Patel TB (1992) Regulation of

hepatic glycolysis and gluconeogenesis by

atrial natriuretic peptide. Arch Biochem Biophys

298: 640 – 645

Roberts LD, Bostrom P, O’Sullivan JF, Schinzel RT,

Lewis GD, Dejam A, Lee YK, Palma MJ, Calhoun

S, Georgiadi A et al (2014) Beta-Amino-

isobutyric Acid Induces Browning of White Fat

and Hepatic beta-Oxidation and Is Inversely

Correlated with Cardiometabolic Risk Factors.

Cell Metab 19: 96 – 108

Seidman CE, Seidman JG (2011) Identifying

sarcomere gene mutations in hypertrophic

cardiomyopathy: a personal history. Circ Res

108: 743 – 750

Wang L, Seidman JG, Seidman CE (2010) Narrative

review: harnessing molecular genetics for the

diagnosis and management of hypertrophic

cardiomyopathy. Ann Intern Med 152: 513 – 520,

W181

License: This is an open access article under the

terms of the Creative Commons Attribution License,

which permits use, distribution and reproduction

in any medium, provided the original work is prop-

erly cited.

EMBO Molecular Medicine Vol 6 | No 4 | 2014 ª 2014 The Authors

EMBO Molecular Medicine The heart-liver metabolic axis Kedryn K Baskin et al

438

that the heart signals to the liver to elevate

glucose production by selectively excluding

uptake and use of oleic acid and triglyceride

in VLDL particles. Indeed, an emerging

theme in homeostatic feedback is the recog-

nition of metabolites as signaling effectors

between tissues as means of physiologic

integration within the body [see (Blad et al,

2012; Liu et al, 2013; Roberts et al, 2014)

for examples]. However, in the setting of an

HCM genotype, the current work suggests

this relationship is injurious.

Many metabolic diseases, such as diabe-

tes and obesity, are ultimately detrimental to

cardiac function, but the reverse has yet to

be investigated. There is a clear relationship

between cardiac metabolism and cardiac

function, but diminished cardiac function,

per se, has thus far not been reported to

negatively influence systemic metabolism.

There is a clear link between liver dysfunc-

tion, specifically non-alcoholic fatty liver

disease, and cardiac dysfunction (Bhatia

et al, 2012), but new evidence reported in

this issue of EMBO Molecular Medicine sug-

gests the reverse is also true.

While the link between cardiac dysfunc-

tion, specifically the alteration of cardiac

metabolism, and deregulated hepatic lipid

metabolism is interesting, the mechanisms

regulating this crosstalk are not resolved by

the work of Magida and Leinwand (2014).

Further studies are required to clarify

whether HCM-induced metabolic abnormali-

ties are the primary cause of liver dysfunc-

tion. It remains unclear whether hepatic lipid

accumulation in this mouse model results

from decreased fatty acid uptake in the heart

alone. Certainly, the relationship between the

heart and liver is not monogamous, and

other tissues such as skeletal muscle, pan-

creas, and adipose are likely to be directly

affected by elevated circulating oleic acid and

VLDL triglyceride. Indeed it is likely that lipid

uptake, utilization, or storage in each of these

tissues contributes to the metabolic pheno-

type described by Magida and Leinwand

(2014) and would be influenced by systemic

agonism of AMPK. Further, PEPCK inhibition

not only affects glucose production by the

liver, kidney, and intestine, but also glycero-

neogenesis in adipocytes. Additionally, it

would be interesting to know if other sarco-

meric mutations also decrease liver function

in end-stage disease, and if so, if a similar

mechanism is involved.

Other aspects of HCM can also be

explored in the R403Q HCM mouse model

within the framework of metabolic abnor-

malities. For example, what role does

calcium homeostasis play in the develop-

ment of cardiac and metabolic dysfunction?

Calcium is an important regulator of energy

metabolism and calcium levels and homeo-

stasis are altered in human HCM patients

(Wang et al, 2010). Perhaps restoring cal-

cium homeostasis in the heart could restore

metabolism in this mouse as well? More-

over, what is the basis for the phenotypic

gender differences in HCM? Is there likely a

protective role for estrogen at the level of

cardiac energetics as well as liver metabo-

lism in the HCM patient? Estrogen certainly

has a role both as it relates to AMPK and

hepatic lipid metabolism (D’Eon et al, 2005;

Bryzgalova et al, 2008), properties which

could be therapeutically exploited.

......................................................

“Certainly, the relationshipbetween the heart and liver isnot monogamous”......................................................

The studies of Magida and Leinwand

(2014) add to a growing number of exam-

ples in which the heart modulates energy

homeostasis and metabolism in non-cardiac

tissues. In this regard, the cardiac natriuretic

peptides, ANP and BNP, have been shown

to improve metabolic parameters by induc-

ing the “browning” of white adipocytes

(Bordicchia et al, 2012). While the thermo-

genic action by ANP is restricted to human,

but not rodent adipocytes (Bordicchia et al,

2012), ANP was shown to induce gluconeo-

genesis in rat hepatocytes (Rashed et al,

1992). Therefore, it is curious that ANP

expression is dramatically enhanced in

HCM, but this mechanism for hepatic glu-

cose output was left unexplored in these

studies. Similarly, elevated expression of the

Mediator subunit MED13 in the heart con-

fers metabolic benefits in mice. MED13 is

negatively regulated by a cardiac specific

microRNA, miR-208, which plays a key role

in cardiac hypertrophy (Grueter et al, 2012).

Whether the miR-208/MED13 axis influ-

ences the metabolic consequences associ-

ated with HCM is an interesting question for

the future. Perhaps a miR-208 inhibitor can

remedy the metabolic defects observed in

HCM by activating cardiac MED13, thus

enhancing systemic metabolism, and revers-

ing or preventing liver steatosis.

In summary, the work of Magida and

Leinwand (2014) highlights the inextricable

Lipid storage

Gluconeogenesis

Plasma lipids Blood glucose

Other organs

Other organs

Lipid uptake

R403Q

HCM

?

?

Figure 1. Crosstalk between the heart and liver is altered in the setting of hypertrophiccardiomyopathy. The HCM-causing mutation in myosin (R403Q) decreases cardiac lipid uptake resulting inincreased plasma lipid content. Consequently, lipid storage is increased in liver, leading to increasedgluconeogenesis, increased blood glucose, ultimately exacerbating cardiac disease. It is still unclear whetherother organs are involved in this crosstalk in HCM (denoted in the figure as ‘?’).

ª 2014 The Authors EMBO Molecular Medicine Vol 6 | No 4 | 2014

Kedryn K Baskin et al The heart-liver metabolic axis EMBO Molecular Medicine

437

relationship between sarcomeric structural

integrity and metabolically-derived energy at

the organismal level, and opens up many

more avenues for future investigation.

Conflict of interestThe authors declare that they have no con-

flict of interest.

ReferencesBhatia LS, Curzen NP, Calder PC, Byrne CD (2012)

Non-alcoholic fatty liver disease: a new and

important cardiovascular risk factor? Eur Heart

J 33: 1190 – 1200

Blad CC, Tang C, Offermanns S (2012) G

protein-coupled receptors for energy

metabolites as new therapeutic targets. Nat

Rev Drug Discov 11: 603 – 619

de Bold AJ (2011) Thirty years of research on atrial

natriuretic factor: historical background and

emerging concepts. Can J Physiol Pharmacol 89:

527 – 531

Bordicchia M, Liu D, Amri EZ, Ailhaud G,

Dessi-Fulgheri P, Zhang C, Takahashi N, Sarzani

R, Collins S (2012) Cardiac natriuretic peptides

act via p38 MAPK to induce the brown fat

thermogenic program in mouse and human

adipocytes. J Clin Invest 122: 1022 – 1036

Bryzgalova G, Lundholm L, Portwood N,

Gustafsson JA, Khan A, Efendic S,

Dahlman-Wright K (2008) Mechanisms of

antidiabetogenic and body weight-lowering

effects of estrogen in high-fat diet-fed

mice. Am J Physiol Endocrinol Metab 295:

E904 – E912

Crilley JG, Boehm EA, Blair E, Rajagopalan B,

Blamire AM, Styles P, McKenna WJ,

Ostman-Smith I, Clarke K, Watkins H (2003)

Hypertrophic cardiomyopathy due to

sarcomeric gene mutations is characterized by

impaired energy metabolism irrespective of the

degree of hypertrophy. J Am Coll Cardiol 41:

1776 – 1782

D’Eon TM, Souza SC, Aronovitz M, Obin MS, Fried

SK, Greenberg AS (2005) Estrogen regulation of

adiposity and fuel partitioning. Evidence of

genomic and non-genomic regulation of

lipogenic and oxidative pathways. J Biol Chem

280: 35983 – 35991

Grueter CE, van Rooij E, Johnson BA, DeLeon SM,

Sutherland LB, Qi X, Gautron L, Elmquist JK,

Bassel-Duby R, Olson EN (2012) A cardiac

microRNA governs systemic energy homeostasis

by regulation of MED13. Cell 149: 671 – 683

Liu S, Brown JD, Stanya KJ, Homan E, Leidl M,

Inouye K, Bhargava P, Gangl MR, Dai L, Hatano

B et al (2013) A diurnal serum lipid integrates

hepatic lipogenesis and peripheral fatty acid

use. Nature 502: 550 – 554

Maass A, Leinwand LA (2000) Animal models of

hypertrophic cardiomyopathy. Curr Opin Cardiol

15: 189 – 196

Magida JA, Leinwand LA (2014) Metabolic

crosstalk between the heart and liver

impacts familial hypertrophic cardiomyopathy.

EMBO Mol Med 6: 482 – 495

Maron BJ, Maron MS (2013) Hypertrophic

cardiomyopathy. Lancet 381: 242 – 255

Rashed HM, Nair BG, Patel TB (1992) Regulation of

hepatic glycolysis and gluconeogenesis by

atrial natriuretic peptide. Arch Biochem Biophys

298: 640 – 645

Roberts LD, Bostrom P, O’Sullivan JF, Schinzel RT,

Lewis GD, Dejam A, Lee YK, Palma MJ, Calhoun

S, Georgiadi A et al (2014) Beta-Amino-

isobutyric Acid Induces Browning of White Fat

and Hepatic beta-Oxidation and Is Inversely

Correlated with Cardiometabolic Risk Factors.

Cell Metab 19: 96 – 108

Seidman CE, Seidman JG (2011) Identifying

sarcomere gene mutations in hypertrophic

cardiomyopathy: a personal history. Circ Res

108: 743 – 750

Wang L, Seidman JG, Seidman CE (2010) Narrative

review: harnessing molecular genetics for the

diagnosis and management of hypertrophic

cardiomyopathy. Ann Intern Med 152: 513 – 520,

W181

License: This is an open access article under the

terms of the Creative Commons Attribution License,

which permits use, distribution and reproduction

in any medium, provided the original work is prop-

erly cited.

EMBO Molecular Medicine Vol 6 | No 4 | 2014 ª 2014 The Authors

EMBO Molecular Medicine The heart-liver metabolic axis Kedryn K Baskin et al

438

that the heart signals to the liver to elevate

glucose production by selectively excluding

uptake and use of oleic acid and triglyceride

in VLDL particles. Indeed, an emerging

theme in homeostatic feedback is the recog-

nition of metabolites as signaling effectors

between tissues as means of physiologic

integration within the body [see (Blad et al,

2012; Liu et al, 2013; Roberts et al, 2014)

for examples]. However, in the setting of an

HCM genotype, the current work suggests

this relationship is injurious.

Many metabolic diseases, such as diabe-

tes and obesity, are ultimately detrimental to

cardiac function, but the reverse has yet to

be investigated. There is a clear relationship

between cardiac metabolism and cardiac

function, but diminished cardiac function,

per se, has thus far not been reported to

negatively influence systemic metabolism.

There is a clear link between liver dysfunc-

tion, specifically non-alcoholic fatty liver

disease, and cardiac dysfunction (Bhatia

et al, 2012), but new evidence reported in

this issue of EMBO Molecular Medicine sug-

gests the reverse is also true.

While the link between cardiac dysfunc-

tion, specifically the alteration of cardiac

metabolism, and deregulated hepatic lipid

metabolism is interesting, the mechanisms

regulating this crosstalk are not resolved by

the work of Magida and Leinwand (2014).

Further studies are required to clarify

whether HCM-induced metabolic abnormali-

ties are the primary cause of liver dysfunc-

tion. It remains unclear whether hepatic lipid

accumulation in this mouse model results

from decreased fatty acid uptake in the heart

alone. Certainly, the relationship between the

heart and liver is not monogamous, and

other tissues such as skeletal muscle, pan-

creas, and adipose are likely to be directly

affected by elevated circulating oleic acid and

VLDL triglyceride. Indeed it is likely that lipid

uptake, utilization, or storage in each of these

tissues contributes to the metabolic pheno-

type described by Magida and Leinwand

(2014) and would be influenced by systemic

agonism of AMPK. Further, PEPCK inhibition

not only affects glucose production by the

liver, kidney, and intestine, but also glycero-

neogenesis in adipocytes. Additionally, it

would be interesting to know if other sarco-

meric mutations also decrease liver function

in end-stage disease, and if so, if a similar

mechanism is involved.

Other aspects of HCM can also be

explored in the R403Q HCM mouse model

within the framework of metabolic abnor-

malities. For example, what role does

calcium homeostasis play in the develop-

ment of cardiac and metabolic dysfunction?

Calcium is an important regulator of energy

metabolism and calcium levels and homeo-

stasis are altered in human HCM patients

(Wang et al, 2010). Perhaps restoring cal-

cium homeostasis in the heart could restore

metabolism in this mouse as well? More-

over, what is the basis for the phenotypic

gender differences in HCM? Is there likely a

protective role for estrogen at the level of

cardiac energetics as well as liver metabo-

lism in the HCM patient? Estrogen certainly

has a role both as it relates to AMPK and

hepatic lipid metabolism (D’Eon et al, 2005;

Bryzgalova et al, 2008), properties which

could be therapeutically exploited.

......................................................

“Certainly, the relationshipbetween the heart and liver isnot monogamous”......................................................

The studies of Magida and Leinwand

(2014) add to a growing number of exam-

ples in which the heart modulates energy

homeostasis and metabolism in non-cardiac

tissues. In this regard, the cardiac natriuretic

peptides, ANP and BNP, have been shown

to improve metabolic parameters by induc-

ing the “browning” of white adipocytes

(Bordicchia et al, 2012). While the thermo-

genic action by ANP is restricted to human,

but not rodent adipocytes (Bordicchia et al,

2012), ANP was shown to induce gluconeo-

genesis in rat hepatocytes (Rashed et al,

1992). Therefore, it is curious that ANP

expression is dramatically enhanced in

HCM, but this mechanism for hepatic glu-

cose output was left unexplored in these

studies. Similarly, elevated expression of the

Mediator subunit MED13 in the heart con-

fers metabolic benefits in mice. MED13 is

negatively regulated by a cardiac specific

microRNA, miR-208, which plays a key role

in cardiac hypertrophy (Grueter et al, 2012).

Whether the miR-208/MED13 axis influ-

ences the metabolic consequences associ-

ated with HCM is an interesting question for

the future. Perhaps a miR-208 inhibitor can

remedy the metabolic defects observed in

HCM by activating cardiac MED13, thus

enhancing systemic metabolism, and revers-

ing or preventing liver steatosis.

In summary, the work of Magida and

Leinwand (2014) highlights the inextricable

Lipid storage

Gluconeogenesis

Plasma lipids Blood glucose

Other organs

Other organs

Lipid uptake

R403Q

HCM

?

?

Figure 1. Crosstalk between the heart and liver is altered in the setting of hypertrophiccardiomyopathy. The HCM-causing mutation in myosin (R403Q) decreases cardiac lipid uptake resulting inincreased plasma lipid content. Consequently, lipid storage is increased in liver, leading to increasedgluconeogenesis, increased blood glucose, ultimately exacerbating cardiac disease. It is still unclear whetherother organs are involved in this crosstalk in HCM (denoted in the figure as ‘?’).

ª 2014 The Authors EMBO Molecular Medicine Vol 6 | No 4 | 2014

Kedryn K Baskin et al The heart-liver metabolic axis EMBO Molecular Medicine

437

Closeup

Salvaging hope: Is increasing NAD+ a keyto treating mitochondrial myopathy?Robert N Lightowlers & Zofia MA Chrzanowska-Lightowlers

Mitochondrial diseases can arise frommutations either in mitochondrial DNA orin nuclear DNA encoding mitochondriallydestined proteins. Currently, there is nocure for these diseases although treat-ments to ameliorate a subset of the symp-toms are being developed. In this issue ofEMBO Molecular Medicine, Khan et al(2014) use a mouse model to test the effi-cacy of a simple dietary supplement ofnicotinamide riboside to treat and preventmitochondrial myopathies.

See also: NA Khan et al (June 2014)

G etting the right levels of vitamins is

essential for health. Those of us of a

certain age will remember in junior

school being taught about the consequences

of vitamin deficiency and having to memo-

rise those consequences. For example, one

deficiency, exotically named pellagra,

resulted in a combination of dermatitis, diar-

rhoea and dementia. The underlying cause

was identified as a lack of nicotinic acid or

nicotinamide (vitamin B3). Indeed, the

defect was exacerbated by a dietary lack of

tryptophan. This is now understood, as all

three components are important building

blocks for the production of nicotinamide

adenine dinucleotide, NAD, a redox-active

coenzyme and enzyme substrate. This

molecule is well known as a key player in

metabolism, being the primary electron

donor in the mitochondrial respiratory

chain. It is also utilised and broken-down by

a variety of proteins in other subcellular

compartments, such as the family of protein

deacetylases (sirtuins), the poly (ADP

ribose)-phosphorylases (PARPs) and NAD

glycohydrolases. De novo synthesis from

tryptophan is a complex 8-step enzymatic

process, so there are likely to be recycling

pathways that utilise NAD synthesis inter-

mediates as substrates. This is where nico-

tinamide and nicotinic acid feature. Both are

intermediates in NAD biosynthesis, requir-

ing enzymatic pathways of only 2 or 3 steps

respectively to generate NAD (Bogan &

Brenner, 2008). An additional salvage path-

way has been identified in eubacteria and

eukaryotes that is distinct from these nico-

tinic acid or nicotinamide recycling (or

salvaging) pathways; in a two-step process,

nicotinamide riboside (NR) can be first

phosphorylated and then adenylylated to

form NAD+ (Bieganowski & Brenner, 2004;

see Fig 1). Those of us who remember

memorising those vitamin deficiency

diseases at school, probably also remember

the compulsory bottle of milk to be drunk at

break time. Although we did not realise it

then, this was a good source of nicotinamide

riboside, which in addition to being a

normal metabolite in the body is also pres-

ent in cow’s milk.

NR can protect against mitochondrialmyopathy in mice

Defects of the mitochondrial (mt) respiratory

chain constitute one of the most common

forms of heritable metabolic disease. Clinical

presentation varies widely, and significantly,

there is no effective cure. Khan et al hypoth-

esised that under conditions of respiratory

chain deficiency, NADH utilisation is

partially blocked leading to a decrease in the

NAD+/NADH ratio. This constitutes a signal

in the cell that is translated as indicating

high nutrient availability, a condition

completely at odds with the defective mito-

chondrial function. Therefore, by repleting

levels of NAD+, the authors surmise that

mitochondrial dysfunction could be amelio-

rated. To challenge their hypothesis, the

authors have used their mt-Deletor mouse, a

model of mitochondrial myopathy, and

administered the NAD+ precursor, NR. The

Deletor mouse carries a dominant patho-

genic mutation in the major mitochondrial

DNA (mtDNA) replicative helicase, Twinkle,

that corresponds to a mutation found in

patients (Tyynismaa et al, 2005). In Deletor

mice, this causes increased levels of deleted

mtDNA and a subtle but chronically progres-

sive mitochondrial myopathy. Control mice

and pre- and post-symptomatic Deletor mice

were dosed with large (400 mg/kg/day)

amounts of NR for up to 4 months, a regime

previously documented to result in increased

levels of NAD+ in skeletal muscle of wild-

type mice (Canto et al, 2012). Khan et al

show that this treatment resulted in a

marked increase in mitochondrial biogenesis

in skeletal muscle and brown adipose tissue

compared to undosed controls. A similar

increase had been shown in the previous

experiments following NR treatment, both of

cultured cells and in various mice tissue

(Canto et al, 2012). Crucially, however, for

these new NR supplement experiments, the

mt-biogenesis was concomitant with a

decrease in markers of disease progression

in Deletor mice, which were also protected

from ultrastructural abnormalities of mito-

chondria. NR invoked a minor increase in

overall mtDNA levels in both control and

Deletor mice, but intriguingly caused a

decrease in the levels of deleted mtDNA that

accumulated in skeletal muscle of the Dele-

tors. Thus, data were consistent with NR

treatment and increasing NAD+ levels

protecting against mitochondrial disease in

the Deletor mice. In addition to promoting

mt-biogenesis, NR also appeared to enhance

Wellcome Trust Centre for Mitochondrial Research, Institute for Cell and Molecular Biosciences, Medical School, Newcastle University, Newcastle upon Tyne, UK.E-mail: Robert.Lightowlers@ncl.ac.ukDOI 10.15252/emmm.201404179

ª 2014 The Authors. Published under the terms of the CC BY 4.0 license EMBO Molecular Medicine Vol 6 | No 6 | 2014 705

the mitochondrial unfolded protein

response. This increase in a subset of mito-

chondrial chaperones and proteases is

believed to be beneficial to health and

promote an increased lifespan (Pellegrino

et al, 2013).

Why does HR treatment promotemitochondrial biogenesis?

Previous reports have implicated increased

NAD+ levels with increased sirtuin activity,

most notably SIRT1 and SIRT3 (Lagouge

et al, 2006; Hirschey et al, 2011). The conse-

quence is an activation of key transcription

factors including SIRT1 and SIRT3 (Canto

et al, 2012), which upregulate gene products

that are central to mt-biogenesis (Feige et al,

2008). In addition to enhancing oxidative

metabolism in a range of tissues, SIRT1 acti-

vation has also been reported to protect

against diet-induced metabolic disorders by

enhancing fatty acid oxidation (Feige et al,

2008). Consistent with this, Khan et al pres-

ent data to show an NR-mediated increase in

skeletal muscle mRNA levels encoding

proteins that are involved in fatty acid trans-

port or oxidation, namely CD36, ACOX1 and

MCAD. Increasing mitochondrial biogenesis

as a way of treating mitochondrial dysfunc-

tion is encouraging and has been previously

shown to be efficacious for mouse models of

mitochondrial disease (Wenz et al, 2008).

However, it has been well described that

mitochondrial proliferation can occur as a

consequence of mtDNA disease in man. It

will certainly be interesting to discover

whether drug-induced mitochondrial biogen-

esis can also be beneficial to patients with

mitochondrial dysfunction.

Why are these results so encouraging?

To date, there is no effective therapy for

patients with mitochondrial myopathy. Vita-

min cocktails including vitamin B3

(although at far lower doses than used here)

have often been used to treat such patients

for many years, with only sporadic reports

of efficacy. The rationale for increasing

NAD+ levels in order to increase mitochon-

drial mass is reasonable, and the results

reported here are compelling. What is partic-

ularly exciting is that NAD+ intermediates

such as NR are readily available and rela-

tively simple drugs. If the efficacy of NR is

entirely due to its effects as an NAD+

precursor, it is not absolutely clear why

neither nicotinamide nor nicotinic acid

themselves could not be used. Perhaps

because there is evidence that the former is

hepatotoxic at high concentrations and its

efficacy in increasing NAD+ levels in skele-

tal muscle is unclear (Bogan & Brenner,

2008)? Nicotinic acid, however, has been

used for many years to treat patients with

high serum cholesterol levels but can cause

irritating vasodilation (flushing). To counter

this, slow release formulations have been

available for some time. Of these NAD+

precursors, NR or its phosphorylated NAD+

precursor nicotinamide mononucleotide

(NMN) might be the therapeutic molecule of

choice by virtue of being able to access mito-

chondria and be converted to NAD+ by

mitochondrial-specific enzymes. Isoforms of

the NR kinase and NMN adenylyltransferase

are known, but there is conflicting evidence

on their mitochondrial location (Felici et al,

2013). Finally, side effects following admin-

istration of other NAD+ precursors supple-

ments have been reported (Sauve, 2008). It

will of course be necessary to evaluate the

NR dosage used by Khan et al, as it appears

O

O–

N+

Pribo

NaMN

NH+

COOH

NH2Trp

O

O–

N+

ADPribo

NaAD+

O

O–

NH+

Na

O

N+

NH2

Pribo

NMN

O

NH+

NH2

Nam

1a1c

1b2

O

N+

NH2

Ribo

NR

Nicotinamide ribosidesupplement

O

N+

ADPribo

NAD+

NH2

Figure 1. The salvage/recycling pathway for NAD+ biosynthesis from nicotinamide riboside (NR) in man.NR, taken in to the body, can be converted to nicotinamide mononucleotide (NMN) by one of two highly conserved NR kinases in the cytoplasm (pathway 1a). NAM(nicotinamide) can also be converted by NMN synthetase to NMN (pathway 1b). NMN is further converted to NAD+ by the action of one of three adenylyltransferases(NMNAT1-3) that also acts on NaMN (nicotinic acid mononucleotide) to produce NaAD+ (nicotinic acid adenine dinucleotide). The latter is subsequently converted by NADsynthase to NAD+. Nicotinic acid (Na) feeds into the pathway through conversion to NaMN by Na phosphoribosyltransferase (pathway 1c). Tryptophan is the de novoprecursor of NAD+ that also feeds into NaMN synthesis via a multistep pathway (2) described in Bogan and Brenner (2008).

EMBO Molecular Medicine Vol 6 | No 6 | 2014 ª 2014 The Authors

EMBO Molecular Medicine NAD+ treatment for mitochondrial myopathy? Robert N Lightowlers and Zofia M A Chrzanowska-Lightowlers

706

Closeup

Salvaging hope: Is increasing NAD+ a keyto treating mitochondrial myopathy?Robert N Lightowlers & Zofia MA Chrzanowska-Lightowlers

Mitochondrial diseases can arise frommutations either in mitochondrial DNA orin nuclear DNA encoding mitochondriallydestined proteins. Currently, there is nocure for these diseases although treat-ments to ameliorate a subset of the symp-toms are being developed. In this issue ofEMBO Molecular Medicine, Khan et al(2014) use a mouse model to test the effi-cacy of a simple dietary supplement ofnicotinamide riboside to treat and preventmitochondrial myopathies.

See also: NA Khan et al (June 2014)

G etting the right levels of vitamins is

essential for health. Those of us of a

certain age will remember in junior

school being taught about the consequences

of vitamin deficiency and having to memo-

rise those consequences. For example, one

deficiency, exotically named pellagra,

resulted in a combination of dermatitis, diar-

rhoea and dementia. The underlying cause

was identified as a lack of nicotinic acid or

nicotinamide (vitamin B3). Indeed, the

defect was exacerbated by a dietary lack of

tryptophan. This is now understood, as all

three components are important building

blocks for the production of nicotinamide

adenine dinucleotide, NAD, a redox-active

coenzyme and enzyme substrate. This

molecule is well known as a key player in

metabolism, being the primary electron

donor in the mitochondrial respiratory

chain. It is also utilised and broken-down by

a variety of proteins in other subcellular

compartments, such as the family of protein

deacetylases (sirtuins), the poly (ADP

ribose)-phosphorylases (PARPs) and NAD

glycohydrolases. De novo synthesis from

tryptophan is a complex 8-step enzymatic

process, so there are likely to be recycling

pathways that utilise NAD synthesis inter-

mediates as substrates. This is where nico-

tinamide and nicotinic acid feature. Both are

intermediates in NAD biosynthesis, requir-

ing enzymatic pathways of only 2 or 3 steps

respectively to generate NAD (Bogan &

Brenner, 2008). An additional salvage path-

way has been identified in eubacteria and

eukaryotes that is distinct from these nico-

tinic acid or nicotinamide recycling (or

salvaging) pathways; in a two-step process,

nicotinamide riboside (NR) can be first

phosphorylated and then adenylylated to

form NAD+ (Bieganowski & Brenner, 2004;

see Fig 1). Those of us who remember

memorising those vitamin deficiency

diseases at school, probably also remember

the compulsory bottle of milk to be drunk at

break time. Although we did not realise it

then, this was a good source of nicotinamide

riboside, which in addition to being a

normal metabolite in the body is also pres-

ent in cow’s milk.

NR can protect against mitochondrialmyopathy in mice

Defects of the mitochondrial (mt) respiratory

chain constitute one of the most common

forms of heritable metabolic disease. Clinical

presentation varies widely, and significantly,

there is no effective cure. Khan et al hypoth-

esised that under conditions of respiratory

chain deficiency, NADH utilisation is

partially blocked leading to a decrease in the

NAD+/NADH ratio. This constitutes a signal

in the cell that is translated as indicating

high nutrient availability, a condition

completely at odds with the defective mito-

chondrial function. Therefore, by repleting

levels of NAD+, the authors surmise that

mitochondrial dysfunction could be amelio-

rated. To challenge their hypothesis, the

authors have used their mt-Deletor mouse, a

model of mitochondrial myopathy, and

administered the NAD+ precursor, NR. The

Deletor mouse carries a dominant patho-

genic mutation in the major mitochondrial

DNA (mtDNA) replicative helicase, Twinkle,

that corresponds to a mutation found in

patients (Tyynismaa et al, 2005). In Deletor

mice, this causes increased levels of deleted

mtDNA and a subtle but chronically progres-

sive mitochondrial myopathy. Control mice

and pre- and post-symptomatic Deletor mice

were dosed with large (400 mg/kg/day)

amounts of NR for up to 4 months, a regime

previously documented to result in increased

levels of NAD+ in skeletal muscle of wild-

type mice (Canto et al, 2012). Khan et al

show that this treatment resulted in a

marked increase in mitochondrial biogenesis

in skeletal muscle and brown adipose tissue

compared to undosed controls. A similar

increase had been shown in the previous

experiments following NR treatment, both of

cultured cells and in various mice tissue

(Canto et al, 2012). Crucially, however, for

these new NR supplement experiments, the

mt-biogenesis was concomitant with a

decrease in markers of disease progression

in Deletor mice, which were also protected

from ultrastructural abnormalities of mito-

chondria. NR invoked a minor increase in

overall mtDNA levels in both control and

Deletor mice, but intriguingly caused a

decrease in the levels of deleted mtDNA that

accumulated in skeletal muscle of the Dele-

tors. Thus, data were consistent with NR

treatment and increasing NAD+ levels

protecting against mitochondrial disease in

the Deletor mice. In addition to promoting

mt-biogenesis, NR also appeared to enhance

Wellcome Trust Centre for Mitochondrial Research, Institute for Cell and Molecular Biosciences, Medical School, Newcastle University, Newcastle upon Tyne, UK.E-mail: Robert.Lightowlers@ncl.ac.ukDOI 10.15252/emmm.201404179

ª 2014 The Authors. Published under the terms of the CC BY 4.0 license EMBO Molecular Medicine Vol 6 | No 6 | 2014 705

the mitochondrial unfolded protein

response. This increase in a subset of mito-

chondrial chaperones and proteases is

believed to be beneficial to health and

promote an increased lifespan (Pellegrino

et al, 2013).

Why does HR treatment promotemitochondrial biogenesis?

Previous reports have implicated increased

NAD+ levels with increased sirtuin activity,

most notably SIRT1 and SIRT3 (Lagouge

et al, 2006; Hirschey et al, 2011). The conse-

quence is an activation of key transcription

factors including SIRT1 and SIRT3 (Canto

et al, 2012), which upregulate gene products

that are central to mt-biogenesis (Feige et al,

2008). In addition to enhancing oxidative

metabolism in a range of tissues, SIRT1 acti-

vation has also been reported to protect

against diet-induced metabolic disorders by

enhancing fatty acid oxidation (Feige et al,

2008). Consistent with this, Khan et al pres-

ent data to show an NR-mediated increase in

skeletal muscle mRNA levels encoding

proteins that are involved in fatty acid trans-

port or oxidation, namely CD36, ACOX1 and

MCAD. Increasing mitochondrial biogenesis

as a way of treating mitochondrial dysfunc-

tion is encouraging and has been previously

shown to be efficacious for mouse models of

mitochondrial disease (Wenz et al, 2008).

However, it has been well described that

mitochondrial proliferation can occur as a

consequence of mtDNA disease in man. It

will certainly be interesting to discover

whether drug-induced mitochondrial biogen-

esis can also be beneficial to patients with

mitochondrial dysfunction.

Why are these results so encouraging?

To date, there is no effective therapy for

patients with mitochondrial myopathy. Vita-

min cocktails including vitamin B3

(although at far lower doses than used here)

have often been used to treat such patients

for many years, with only sporadic reports

of efficacy. The rationale for increasing

NAD+ levels in order to increase mitochon-

drial mass is reasonable, and the results

reported here are compelling. What is partic-

ularly exciting is that NAD+ intermediates

such as NR are readily available and rela-

tively simple drugs. If the efficacy of NR is

entirely due to its effects as an NAD+

precursor, it is not absolutely clear why

neither nicotinamide nor nicotinic acid

themselves could not be used. Perhaps

because there is evidence that the former is

hepatotoxic at high concentrations and its

efficacy in increasing NAD+ levels in skele-

tal muscle is unclear (Bogan & Brenner,

2008)? Nicotinic acid, however, has been

used for many years to treat patients with

high serum cholesterol levels but can cause

irritating vasodilation (flushing). To counter

this, slow release formulations have been

available for some time. Of these NAD+

precursors, NR or its phosphorylated NAD+

precursor nicotinamide mononucleotide

(NMN) might be the therapeutic molecule of

choice by virtue of being able to access mito-

chondria and be converted to NAD+ by

mitochondrial-specific enzymes. Isoforms of

the NR kinase and NMN adenylyltransferase

are known, but there is conflicting evidence

on their mitochondrial location (Felici et al,

2013). Finally, side effects following admin-

istration of other NAD+ precursors supple-

ments have been reported (Sauve, 2008). It

will of course be necessary to evaluate the

NR dosage used by Khan et al, as it appears

O

O–

N+

Pribo

NaMN

NH+

COOH

NH2Trp

O

O–

N+

ADPribo

NaAD+

O

O–

NH+

Na

O

N+

NH2

Pribo

NMN

O

NH+

NH2

Nam

1a1c

1b2

O

N+

NH2

Ribo

NR

Nicotinamide ribosidesupplement

O

N+

ADPribo

NAD+

NH2

Figure 1. The salvage/recycling pathway for NAD+ biosynthesis from nicotinamide riboside (NR) in man.NR, taken in to the body, can be converted to nicotinamide mononucleotide (NMN) by one of two highly conserved NR kinases in the cytoplasm (pathway 1a). NAM(nicotinamide) can also be converted by NMN synthetase to NMN (pathway 1b). NMN is further converted to NAD+ by the action of one of three adenylyltransferases(NMNAT1-3) that also acts on NaMN (nicotinic acid mononucleotide) to produce NaAD+ (nicotinic acid adenine dinucleotide). The latter is subsequently converted by NADsynthase to NAD+. Nicotinic acid (Na) feeds into the pathway through conversion to NaMN by Na phosphoribosyltransferase (pathway 1c). Tryptophan is the de novoprecursor of NAD+ that also feeds into NaMN synthesis via a multistep pathway (2) described in Bogan and Brenner (2008).

EMBO Molecular Medicine Vol 6 | No 6 | 2014 ª 2014 The Authors

EMBO Molecular Medicine NAD+ treatment for mitochondrial myopathy? Robert N Lightowlers and Zofia M A Chrzanowska-Lightowlers

706

strikingly high (400 mg/kg/day) compared

to most commercially available supplements

(60–500 mg/person/day). Whether such a

large dosage is viable as a supplement needs

to be established; however, it will be excit-

ing to follow new pharmacokinetic data for

this potentially therapeutic nucleoside deriv-

ative.

AcknowledgementsRNL and ZCL would like to thank The Wellcome

Trust [096919/Z/11/Z] for continuing support.

Conflict of interestThe authors declare that they have no conflict of

interest.

ReferencesBieganowski P, Brenner C (2004) Discoveries of

nicotinamide riboside as a nutrient and

conserved NRK genes establish a

Preiss-Handler independent route to NAD+ in

fungi and humans. Cell 117: 495 – 502

Bogan KL, Brenner C (2008) Nicotinic acid,

nicotinamide, and nicotinamide riboside: a

molecular evaluation of NAD+ precursor

vitamins in human nutrition. Annu Rev Nutr 28:

115 – 130

Canto C, Houtkooper RH, Pirinen E, Youn DY,

Oosterveer MH, Cen Y, Fernandez-Marcos PJ,

Yamamoto H, Andreux PA, Cettour-Rose P et al

(2012) The NAD(+) precursor nicotinamide

riboside enhances oxidative metabolism and

protects against high-fat diet-induced obesity.

Cell Metab 15: 838 – 847

Feige JN, Lagouge M, Canto C, Strehle A, Houten

SM, Milne JC, Lambert PD, Mataki C, Elliott PJ,

Auwerx J (2008) Specific SIRT1 activation

mimics low energy levels and protects against

diet-induced metabolic disorders by enhancing

fat oxidation. Cell Metab 8: 347 – 358

Felici R, Lapucci A, Ramazzotti M, Chiarugi A

(2013) Insight into molecular and functional

properties of NMNAT3 reveals new hints of

NAD homeostasis within human mitochondria.

PLoS ONE 8: e76938

Hirschey MD, Shimazu T, Jing E, Grueter CA, Collins

AM, Aouizerat B, Stancakova A, Goetzman E, Lam

MM, Schwer B et al (2011) SIRT3 deficiency and

mitochondrial protein hyperacetylation

accelerate the development of the metabolic

syndrome. Mol Cell 44: 177 – 190

Khan NA, Auranen M, Paetau I, Pirinen E, Euro L,

Forsström S, Pasila L, Velagapudi Carroll CJ,

Auwerx J, Suomalainen A (2014) Effective

treatment of mitochondrial myopathy by

nicotinamide riboside, a vitamin-B3. EMBO Mol

Med. 6: 721 – 731

Lagouge M, Argmann C, Gerhart-Hines Z, Meziane

H, Lerin C, Daussin F, Messadeq N, Milne J,

Lambert P, Elliott P et al (2006) Resveratrol

improves mitochondrial function and protects

against metabolic disease by activating SIRT1

and PGC-1alpha. Cell 127: 1109 – 1122

Pellegrino MW, Nargund AM, Haynes CM (2013)

Signaling the mitochondrial unfolded protein

response. Biochim Biophys Acta 1833: 410 – 416

Sauve AA (2008) NAD+ and vitamin B3: from

metabolism to therapies. J Pharmacol Exp Ther

324: 883 – 893

Tyynismaa H, Mjosund KP, Wanrooij S,

Lappalainen I, Ylikallio E, Jalanko A, Spelbrink

JN, Paetau A, Suomalainen A (2005) Mutant

mitochondrial helicase Twinkle causes multiple

mtDNA deletions and a late-onset

mitochondrial disease in mice. Proc Natl Acad

Sci USA 102: 17687 – 17692

Wenz T, Diaz F, Spiegelman BM, Moraes CT (2008)

Activation of the PPAR/PGC-1alpha pathway

prevents a bioenergetic deficit and effectively

improves a mitochondrial myopathy

phenotype. Cell Metab 8: 249 – 256

License: This is an open access article under the

terms of the Creative Commons Attribution 4.0

License, which permits use, distribution and

reproduction in any medium, provided the original

work is properly cited.

ª 2014 The Authors EMBO Molecular Medicine Vol 6 | No 6 | 2014

Robert N Lightowlers and Zofia M A Chrzanowska-Lightowlers NAD+ treatment for mitochondrial myopathy? EMBO Molecular Medicine

707

SynopsisAlthough short‑term anti‑miR‑33 therapy was reported to increase circulating HDL‑cholesterol and reduce atherosclerosis, long‑term adverse effects are here shown for the first time in mice fed a high‑fat diet to result in hypertriglyceridemia and moderate hepatic steatosis.• Theeffectoflong‑terminhibitionofmiR‑33wasdeterminedinmicefedachowdietand

high‑fatdiet.• ChronictherapeuticsilencingofmiR‑33increasedcirculatingtriglyceridesandlipid

accumulationintheliversofmicefedahigh‑fatdiet.• miR‑33inhibitionraisedtheexpressionofgenesinvolvedinfattyacidsynthesisandlipid

metabolism.• Furtherstudiesarewarrantedtounderstandthecomplexgeneregulatorynetwork

controlledbymiR‑33.

Long-term therapeutic silencing of miR-33 increases circulating triglyceride levels and hepatic lipid accumulation in miceLeighGoedeke1,2,3,4,†,AlessandroSalerno3,4,†,CristinaMRamírez1,2,3,4,LiangGuo3,4,RyanMAllen5,XiaokeYin6,SarahRLangley6,ChristineEsau7,AmarylisWanschel3,4,EdwardAFisher3,4,YajairaSuárez1,2,3,4,AngelBaldán5,ManuelMayr6andCarlosFernández‑Hernando*,1,2,3,4

1VascularBiologyandTherapeuticsProgram,YaleUniversitySchoolofMedicine,NewHaven,CT,USA,2IntegrativeCellSignalingandNeurobiologyofMetabolismProgram,SectionofComparativeMedicineYaleUniversitySchoolofMedicine,NewHaven,CT,USA,3LeonH.CharneyDivisionofCardiology,DepartmentofMedicine,NewYorkUniversitySchoolofMedicine,NewYork,NY,USA,4MarcandRutiBellVascularBiologyandDiseaseProgram,NewYorkUniversitySchoolofMedicine,NewYork,NY,USA,5EdwardA.DoisyDepartmentofBiochemistryandMolecularBiology,CenterforCardiovascularResearch,SaintLouisUniversitySchoolofMedicine,SaintLouis,MO,USA,6King'sBritishHeartFoundationCentre,King'sCollegeLondon,London,UK,7RegulusTherapeutics,SanDiego,CA,USA

*Correspondingauthor.Tel:+12037374615;Fax:+12037372290;E‑mail:carlos.fernandez@yale.edu

DOI: 10.15252/emmm.201404046EMBO Molecular Medicine (2014) 6 (9):1133-1141

Plasma high‑density lipoprotein (HDL) levels show a strong inverse correlation with atherosclerotic vascular disease. Previous studies have demonstrated that antagonism of miR‑33 in vivo increases circulating HDL and reverse cholesterol transport (RCT), thereby reducing the progression and enhancing the regression of athero‑sclerosis. While the efficacy of short‑term anti‑miR‑33 treatment has been previously studied, the long‑term effect of miR‑33 antagonism in vivo remains to be elucidated. Here, we show that long‑term therapeutic silencing of miR‑33 increases circulating triglyceride (TG) levels and lipid accumulation in the liver. These adverse effects were only found when mice were fed a high‑fat diet (HFD). Mechanistically, we demonstrate that chronic inhibition of miR‑33 increases the expression of genes involved in fatty acid synthesis such as acetyl‑CoA carboxylase (ACC) and fatty acid synthase (FAS) in the livers of mice treated with miR‑33 antisense oligonucleotides. We also report that anti‑miR‑33 therapy enhances the expression of nuclear transcription Y subunit gamma (NFYC), a transcriptional regulator required for DNA binding and full transcriptional activation of SREBP‑responsive genes, including ACC and FAS. Taken together, these results suggest that persistent inhibition of miR‑33 when mice are fed a high‑fat diet (HFD) might cause deleterious effects such as moderate hepatic steatosis and hypertriglyceridemia. These unexpected findings highlight the importance of assessing the effect of chronic inhibition of miR‑33 in non‑human primates before we can translate this therapy to humans.

Report SOURCEDATA

TRANSPARENTPROCESS

OPENACCESS

SynopsisUpon nutrient change, a homogeneous E. coli population can split into a growing and a non‑growing persister phenotype. Stochastic variation in metabolic flux is responsible for this responsive diversification.

• Responsivediversificationoffersanexplanationforlagphasesinbacterialcultures• Flux‑inducedphenotypicbistabilitygeneralizestocentralmetabolism• Conditionalbet‑hedgingbalancesfastglycolyticgrowthandabilityforgluconeogenic

growth• Limitedcarboninfluxisamajortriggerforpersistence

Phenotypic bistability in Escherichia coli's central carbon metabolismOliverKotte1,†,BenjaminVolkmer1,†,JakubLRadzikowski2andMatthiasHeinemann*,1,2

1InstituteofMolecularSystemsBiology,ETHZurich,Zurich,Switzerland,2MolecularSystemsBiology,GroningenBiomolecularSciencesandBiotechnologyInstitute,UniversityofGroningen,Groningen,TheNetherlands

*Correspondingauthor.Tel:+31503638146;E‑mail:m.heinemann@rug.nl†Theseauthorscontributedequallytothiswork

DOI: 10.15252/msb.20135022Molecular Systems Biology (2014) 10:736

Fluctuations in intracellular molecule abundance can lead to distinct, coexisting phenotypes in isogenic popula‑tions. Although metabolism continuously adapts to unpredictable environmental changes, and although bistabil‑ity was found in certain substrate‑uptake pathways, central carbon metabolism is thought to operate deterministi‑cally. Here, we combine experiment and theory to demonstrate that a clonal Escherichia coli population splits into two stochastically generated phenotypic subpopulations after glucose‑gluconeogenic substrate shifts. Most cells refrain from growth, entering a dormant persister state that manifests as a lag phase in the population growth curve. The subpopulation‑generating mechanism resides at the metabolic core, overarches the metabolic and transcriptional networks, and only allows the growth of cells initially achieving sufficiently high gluconeogenic flux. Thus, central metabolism does not ensure the gluconeogenic growth of individual cells, but uses a popula‑tion‑level adaptation resulting in responsive diversification upon nutrient changes.

Article

Acetate(Fumarate)

PEP

flux FBPFbp

CraE

Time

Expression

Ø

Ø

SOURCEDATA

TRANSPARENTPROCESS

OPENACCESS

SynopsisThis study reveals the interaction between inflammatory macrophages and β‑cells leading to the recruitment of diabetogenic neutrophils in the pancreas of neonatal mice via CXCR2/CXCR2 ligands. Inhibition of CXCR2 reduces the diabetogenic T‑cell response, insulitis, and incidence of diabetes.

• InyoungNODmice,CXCR2+‑neutrophilsarerecruitedfromthebloodintothepancreaticisletsandnotinthetwonon‑diabetesproneC57BL/6andBALB/cmice.

• ThetwoCXCR2ligands,CXCL1andCXCL2,aresecretedinthepancreaticisletsfromtheyoungandnotfromtheadultNODmiceorthetwonon‑diabetespronemice.

• Thepancreaticβ‑cellsarethemainsourceofCXCL1andCXCL2inthepancreaticisletsofyoungNODmice.

• TheproductionofCXCL1andCXCL2bytheβ‑cellsisinducedbyIL‑1b‑producingmacrophagesinfiltratingthepancreaticisletsofyoungNODmice.

• TheearlyblockageofneutrophilrecruitmentusingCXCR2antagonistreducestheinsulitis,theeffectoractivityofdiabetogenicCD8Tcells,andthedevelopmentofautoimmunediabetesinNODmice.

Macrophages and β-cells are responsible for CXCR2-mediated neutrophil infiltration of the pancreas during autoimmune diabetesJulienDiana*,1,2andAgnèsLehuen2,3,4

1InstitutNationaldelaSantéetdelaRechercheMédicale(INSERM),U1151,Necker‑EnfantsMaladesInstitute(INEM)NeckerHospital,Paris,France,2SorbonneParisCité,UniversitéParisDescartes,Paris,France,3InstitutNationaldelaSantéetdelaRechercheMédicale(INSERM),U1016,CochinInstituteCochinHospital,Paris,France,4Laboratoired'ExcellenceINFLAMEX,Paris,France

*Correspondingauthor.Tel:+33144495069;E‑mail:julien.diana@inserm.fr

DOI: 10.15252/emmm.201404144EMBO Molecular Medicine (2014) 6 (8):1090-1104

Autoimmune type 1 diabetes (T1D) development results from the interaction between pancreatic β‑cells, and the innate and the adaptive immune systems culminating with the destruction of the insulin‑secreting β‑cells by auto‑reactive T cells. This diabetogenic course starts during the first postnatal weeks by the infiltration of the pancreatic islets by innate immune cells and particularly neutrophils. Here, we aim to determine the cellular and molecular mechanism leading to the recruitment of this neutrophils in the pancreatic islets of non‑obese diabetic (NOD) mice. Here, we show that neutrophil recruitment in the pancreatic islets is controlled by inflammatory macrophages and β‑cells themselves. Macrophages and β‑cells produce the chemokines CXCL1 and CXCL2, recruiting CXCR2‑express‑ing neutrophils from the blood to the pancreatic islets. We further show that pancreatic macrophages secrete IL‑1β‑inducing CXCR2 ligand production by the β‑cells. Finally, the blockade of neutrophil recruitment at early ages using CXCR2 antagonist dampens the diabetogenic T‑cell response and the later development of autoimmune diabetes, supporting the therapeutic potential of this approach.

Research article TRANSPARENTPROCESS

OPENACCESS

SynopsisMolecular bypass therapy with orally administered deoxythymidine monophosphate and deoxycytidine monophosphate produces clinical, molecular, biochemical, and histological improvements in a mitochondrial DNA depletion syndrome Tk2 knock‑in mouse model.

• DeoxypyrimidinemonophosphatescrossbiologicalbarriersinTk2‑deficientmice.• dCMP+dTMPtreatmentrestoresmtDNAlevelsandamelioratesthephenotypeof

Tk2‑mutantmice.• dCMP+dTMPhavedose‑relatedclinicalandbiochemicaleffectsinTk2‑deficientmice.

Deoxypyrimidine monophosphate bypass therapy for thymidine kinase 2 deficiencyCaterinaGarone1,2,BeatrizGarcia‑Diaz1,ValentinaEmmanuele1,3,LuisCLopez4,SabaTadesse1,HasanOAkman1,KurenaiTanji5,CatarinaMQuinzii1andMichioHirano*,1

1DepartmentofNeurology,ColumbiaUniversityMedicalCenter,NewYork,NY,USA,2HumanGeneticsJointPhDProgram,UniversityofBolognaandTurin,Turin,Italy,3PediatricClinicUniversityofGenoaIRCCSG.GasliniInstitute,Genoa,Italy,4InstitutodeBiotecnología,CentrodeInvestigaciónBiomédica,UniversidaddeGranadaParqueTecnológicodeCienciasdelaSalud,Armilla,Spain,5DepartmentofPathologyandCellBiology,ColumbiaUniversityMedicalCenter,NewYork,NY,USA

*Correspondingauthor.Tel:+12123051048;Fax:+12123053986;E‑mail:mh29@columbia.edu

DOI: 10.15252/emmm.201404092EMBO Molecular Medicine (2014) 6 (8):1016-1027

Autosomal recessive mutations in the thymidine kinase 2 gene (TK2) cause mitochondrial DNA depletion, multiple deletions, or both due to loss of TK2 enzyme activity and ensuing unbalanced deoxynucleotide triphos‑phate (dNTP) pools. To bypass Tk2 deficiency, we administered deoxycytidine and deoxythymidine monophos‑phates (dCMP+dTMP) to the Tk2 H126N (Tk2−/−) knock‑in mouse model from postnatal day 4, when mutant mice are phenotypically normal, but biochemically affected. Assessment of 13‑day‑old Tk2−/− mice treated with dCMP+dTMP 200 mg/kg/day each (Tk2−/−200dCMP/dTMP) demonstrated that in mutant animals, the compounds raise dTTP concentrations, increase levels of mtDNA, ameliorate defects of mitochondrial respiratory chain enzymes, and significantly prolong their lifespan (34 days with treatment versus 13 days untreated). A second trial of dCMP+dTMP each at 400 mg/kg/day showed even greater phenotypic and biochemical improvements. In conclusion, dCMP/dTMP supplementation is the first effective pharmacologic treatment for Tk2 deficiency.

Research article SOURCEDATA

TRANSPARENTPROCESS

OPENACCESS

SynopsisAn RNAi screen for genes needed in mtDNA copy number maintenance in Drosophila yielded 97 positives, including previously characterized mtDNA maintenance proteins, subunits of the cytoribosome, proteasome, and ATP synthase.

• AnRNAiscreenforgenesneededinmtDNAcopynumbermaintenanceinDrosophilayielded97positives.

• TheseincludedpreviouslycharacterizedcomponentsofthemtDNAmaintenancemachinery.• Othermajorclassesofpositiveswerethecytoribosome,proteasome,andATPsynthase.• ATPsynthasedeficiencyresultsinincreasedROSandactivationofmitochondrialturnoverby

pathway(s)distinctfromclassicalautophagy.

Screen for mitochondrial DNA copy number maintenance genes reveals essential role for ATP synthaseAtsushiFukuoh1,2,3,†,GiuseppeCannino1,†,MikeGerards1,SuzanneBuckley1,SelenaKazancioglu1,FilippoScialo1,EeroLihavainen4,AndreRibeiro4,EricDufour1andHowardTJacobs*,1,5

1BioMediTechandTampereUniversityHospital,UniversityofTampere,Tampere,Finland,2DepartmentofClinicalChemistryandLaboratoryMedicine,KyushuUniversityGraduateschoolofMedicalSciences,Fukuoka,Japan,3DepartmentofMedicalLaboratoryScience,JunshinGakuenUniversity,Fukuoka,Japan,4DepartmentofSignalProcessing,TampereUniversityofTechnology,Tampere,Finland,5ResearchProgramofMolecularNeurology,UniversityofHelsinki,Helsinki,Finland

*Correspondingauthor.Tel:+358335517731,+358503412894;E‑mail:howard.t.jacobs@uta.fi†Theseauthorsequallycontributedtothiswork.

DOI: 10.15252/msb.20145117Molecular Systems Biology (2014) 10:734

The machinery of mitochondrial DNA (mtDNA) maintenance is only partially characterized and is of wide interest due to its involvement in disease. To identify novel components of this machinery, plus other cellular pathways required for mtDNA viability, we implemented a genome‑wide RNAi screen in Drosophila S2 cells, assaying for loss of fluorescence of mtDNA nucleoids stained with the DNA‑intercalating agent PicoGreen. In addition to previ‑ously characterized components of the mtDNA replication and transcription machineries, positives included many proteins of the cytosolic proteasome and ribosome (but not the mitoribosome), three proteins involved in vesicle transport, some other factors involved in mitochondrial biogenesis or nuclear gene expression, > 30 mainly uncharacterized proteins and most subunits of ATP synthase (but no other OXPHOS complex). ATP synthase knockdown precipitated a burst of mitochondrial ROS production, followed by copy number depletion involving increased mitochondrial turnover, not dependent on the canonical autophagy machinery. Our findings will inform future studies of the apparatus and regulation of mtDNA maintenance, and the role of mitochondrial bioenergetics and signaling in modulating mtDNA copy number.

Article SOURCEDATA

TRANSPARENTPROCESS

OPENACCESS

SynopsisIt is well established that cognitive deficits go hand in hand with aging. Restoring cholesterol levels in the aged hippocampus to values found in the young can rescue learning and memory in the old, linking age‑dependent cholesterol decline with synaptic plasticity and neuronal function.

• Amildyetsignificantreductioninmembranecholesterolcharacterizestheagingrodenthippocampus.

• LowsynaptichippocampalcholesteroldeterminesreducedAktdephosphorylationafterNMDA‑inducedLTD,togetherwithreducedglutamate(AMPA)receptorlateraldiffusionandendocytosis.

• LowsynaptichippocampalcholesterolplaysaroleinthepoorLTDofoldmiceandrats,inex-vivoandin vivoparadigms.

• NormallevelsofpAktafterNMDA,properreceptorlateraldiffusion,andinternalizationandnormal(younganimals‑like)LTDintheoldcanberescuedbymembranecholesterolreplenishment.

• Cholesterolreplenishmentinlivingoldratsimproveslearningandmemory.

Constitutive hippocampal cholesterol loss underlies poor cognition in old rodentsMauricioGMartin*,1,2,TariqAhmed3,AlejandraKorovaichuk4,CesarVenero5,SilviaAMenchón2,6,IsabelSalas1,SebastianMunck2,OscarHerreras4,DetlefBalschun3andCarlosGDotti*,1,2

1CentroBiologíaMolecular“SeveroOchoa”CSIC‑UAM,Madrid,Spain,2VIBCenterfortheBiologyofDisease,CenterforHumanGenetics,UniversityofLeuven(KULeuven),Leuven,Belgium,3LaboratoryofBiologicalPsychology,FacultyofPsychologyandEducationalSciences,UniversityofLeuven(KULeuven),Leuven,Belgium,4DepartamentodeNeurobiologíaFuncionalydeSistemas,InstitutoCajal–CSIC,Madrid,Spain,5DepartamentodePsicobiología,FacultaddePsicología,UNED,Madrid,Spain,6IFEG‑CONICETandFaMAF,UniversidadNacionaldeCórdoba,Córdoba,Argentina

*Correspondingauthor.Tel:+543514681465;E‑mail:mauricio.m@immf.uncor.edu*Correspondingauthor.Tel:+34911964401;E‑mail:cdotti@cbm.csic.es

DOI: 10.15252/emmm.201303711EMBO Molecular Medicine (2014) 6 (7):902-917

Cognitive decline is one of the many characteristics of aging. Reduced long‑term potentiation (LTP) and long‑term depression (LTD) are thought to be responsible for this decline, although the precise mechanisms underlying LTP and LTD dampening in the old remain unclear. We previously showed that aging is accompanied by the loss of cholesterol from the hippocampus, which leads to PI3K/Akt phosphorylation. Given that Akt de‑phosphorylation is required for glutamate receptor internalization and LTD, we hypothesized that the decrease in cholesterol in neuronal membranes may contribute to the deficits in LTD typical of aging. Here, we show that cholesterol loss triggers p‑Akt accumulation, which in turn perturbs the normal cellular and molecular responses induced by LTD, such as impaired AMPA receptor internalization and its reduced lateral diffusion. Electrophysiology recordings in brain slices of old mice and in anesthetized elderly rats demonstrate that the reduced hippocampal LTD associ‑ated with age can be rescued by cholesterol perfusion. Accordingly, cholesterol replenishment in aging animals improves hippocampal‑dependent learning and memory in the water maze test.

Research article SOURCEDATA

TRANSPARENTPROCESS

OPENACCESS

SynopsisMouse liver metabolites were quantified by mass spectrometry and mapped by genome‑wide association. Genetic factors were shown to contribute substantially to metabolite levels and adenoviral overexpression validated several of the identified loci.

• Livermetabolitesexhibitawiderangeofvariation,indicatingstronggeneticinfluences.• Approximately40%ofmetabolitesareestimatedtoberegulatedbygeneticfactors.• Asignificantoverlapwasobservedbetweengeneticfactorsregulatingmouseliver

metabolitesandgeneticfactorsregulatinghumanserummetabolites.• Metabolitelevelscorrelatedsignificantlybothwitheachotherandwithotherphenotypes

suchastranscriptlevelsandphysiologicaltraits.

Genetic regulation of mouse liver metabolite levelsAnatoleGhazalpour1,†,BrianJBennett1,6,†,DianaShih1,NamChe1,LuzOrozco2,CalvinPan3,RaffiHagopian1,AiqingHe4,PaulKayne4,Wen‑pinYang4,ToddKirchgessner5,AldonsJLusis*1,3.

1DivisionofCardiology,DepartmentofMedicine,UCLA,LosAngeles,CA,USA,2DepartmentofMolecularCellandDevelopmentalBiology,UCLA,LosAngeles,CA,USA,3DepartmentofHumanGenetics,UCLA,LosAngeles,CA,USA,4DepartmentofAppliedGenomics,Bristol‑MyersSquibb,Princeton,NJ,USA,5DepartmentofAtherosclerosisDrugDiscovery,Bristol‑MyersSquibb,Princeton,NJ,USA,6DepartmentofGenetics,UniversityofNorthCarolinaatChapelHill,Kannapolis,NC,USA

*Correspondingauthor.Tel:+13108251359;E‑mail:jlusis@mednet.ucla.edu

DOI: 10.15252/msb.20135004Molecular Systems Biology (2014) 10:730

We profiled and analyzed 283 metabolites representing eight major classes of molecules including Lipids, Carbohy‑drates, Amino Acids, Peptides, Xenobiotics, Vitamins and Cofactors, Energy Metabolism, and Nucleotides in mouse liver of 104 inbred and recombinant inbred strains. We find that metabolites exhibit a wide range of variation, as has been previously observed with metabolites in blood serum. Using genome‑wide association analysis, we mapped 40% of the quantified metabolites to at least one locus in the genome and for 75% of the loci mapped we identified at least one candidate gene by local expression QTL analysis of the transcripts. Moreover, we validated 2 of 3 of the significant loci examined by adenoviral overexpression of the genes in mice. In our GWAS results, we find that at significant loci the peak markers explained on average between 20 and 40% of variation in the metabolites. Moreover, 39% of loci found to be regulating liver metabolites in mice were also found in human GWAS results for serum metabolites, providing support for similarity in genetic regulation of metabolites between mice and human. We also integrated the metabolomic data with transcriptomic and clinical phenotypic data to evaluate the extent of co‑variation across various biological scales.

Article TRANSPARENTPROCESS

OPENACCESS

SynopsisMore and more evidence implicates ER stress in diabetes. Thioredoxin‑interacting protein (Txnip) is here shown to interact with and regulate protein disulfide isomerases (PDIs) activity and ER stress. This study highlights new therapeutic targets for treating diabetes.

• Anunbiasedproteomicsapproachaswellasspecificpulldownassaysrevealedaninteractionofthioredoxin‑interactingprotein(Txnip)withproteindisulfideisomerases(PDIs).

• TxnipincreasesPDIactivity,andTxnipknockoutleadstoincreasedproteinubiquitinationandincreasedlevelsofXbp1s,amarkerofERstress.

• IncreasedlevelsofXbp1sinTxnip‑KOmiceisreversedbytreatmentwithchemicalchaperones.

Thioredoxin-interacting protein regulates protein disulfide isomerases and endoplasmic reticulum stressSamuelLee1,2,3,SooMinKim1,2,JamesDotimas1,2,LetitiaLi1,2,EdwardPFeener4,StephanBaldus3,RonaldBMyers1,2,WilliamAChutkow2,ParthPatwari2,JunYoshioka2andRichardTLee*,1,2

1HarvardDepartmentofStemCellandRegenerativeBiology,HarvardStemCellInstituteHarvardMedicalSchoolBrighamandWomen'sHospital,Cambridge,MA,USA,2TheCardiovascularDivision,DepartmentofMedicine,HarvardMedicalSchoolBrighamandWomen'sHospital,Cambridge,MA,USA,3DepartmentIIIofInternalMedicine,UniversityHospitalofCologne,Cologne,Germany,4TheJoslinDiabetesCenter,HarvardMedicalSchool,Boston,MA,USA

*Correspondingauthor.Tel:+16177688282;Fax:+16177688280;E‑mail:RLee@partners.org

DOI: 10.15252/emmm.201302561EMBO Molecular Medicine (2014) 6 (6):732-743

The endoplasmic reticulum (ER) is responsible for protein folding, modification, and trafficking. Accumulation of unfolded or misfolded proteins represents the condition of ER stress and triggers the unfolded protein response (UPR), a key mechanism linking supply of excess nutrients to insulin resistance and type 2 diabetes in obesity. The ER harbors proteins that participate in protein folding including protein disulfide isomerases (PDIs). Chang‑es in PDI activity are associated with protein misfolding and ER stress. Here, we show that thioredoxin‑interact‑ing protein (Txnip), a member of the arrestin protein superfamily and one of the most strongly induced proteins in diabetic patients, regulates PDI activity and UPR signaling. We found that Txnip binds to PDIs and increases their enzymatic activity. Genetic deletion of Txnip in cells and mice led to increased protein ubiquitination and splicing of the UPR regulated transcription factor X‑box‑binding protein 1 (Xbp1s) at baseline as well as under ER stress. Our results reveal Txnip as a novel direct regulator of PDI activity and a feedback mechanism of UPR signaling to decrease ER stress.

Research article SOURCEDATA

TRANSPARENTPROCESS

OPENACCESS

SynopsisGlutamine plays an important role in cellular growth in several cancers. In this study, a further link between glutamine dependency and tumor invasiveness is established in ovarian cancer. Glutamine maintains the high‑invasive phenotype by regulating STAT3 signaling.

• High‑invasiveovariancancer(OVCA)cellsareglutaminedependentincontrasttolow‑invasivecellsthatareglutamineindependent.

• GlutamineregulatesSTAT3activationinhigh‑invasivecancercells.• Glutamine'sentryintoTCAcyclemodulatestheinvasivepotentialofhigh‑invasivecancercells.• Theratioofglutaminecatabolismoverglutamineanabolismisassociatedwithworseoverall

survivalinOVCApatients.

Metabolic shifts toward glutamine regulate tumor growth, invasion and bioenergetics in ovarian cancerLifengYang1,2,TylerMoss3,LingegowdaSMangala4,5,JuanMarini6,HongyunZhao1,2,StephenWahlig1,7,GuillermoArmaiz‑Pena4,5,DahaiJiang4,5,AbhinavAchreja1,2,JuliaWin1,2,RajeshaRoopaimoole4,5,CristianRodriguez‑Aguayo5,7,ImeldaMercado‑Uribe8,GabrielLopez‑Berestein5,7,JinsongLiu8,TakashiTsukamoto9,AnilK.Sood4,5,PrahladTRam3andDeepakNagrath*,1,2,10

1LaboratoryforSystemsBiologyofHumanDiseases,RiceUniversity,Houston,TX,USA,2DepartmentofChemicalandBiomolecularEngineering,RiceUniversity,Houston,TX,USA,3DepartmentofSystemsBiology,UniversityofTexasMDAndersonCancerCenter,Houston,TX,USA,4DepartmentsofGynecologicalOncologyandCancerBiology,UniversityofTexasMDAndersonCancerCenter,Houston,TX,USA,5CenterforRNAInterferenceandNon‑CodingRNAUniversityofTexasMDAndersonCancerCenter,Houston,TX,USA,6BaylorCollegeofMedicine,Houston,TX,USA,7DepartmentofExperimentalTherapeutics,UniversityofTexasMDAndersonCancerCenter,Houston,TX,USA,8DepartmentofPathology,UniversityofTexasMDAndersonCancerCenter,Houston,TX,USA,9JohnsHopkinsUniversity,Baltimore,MD,USA,10DepartmentofBioengineering,RiceUniversity,Houston,TX,USA

*Correspondingauthor.Tel:+17133486408;Fax:+17133485478;E‑mail:deepak.nagrath@rice.edu

DOI: 10.1002/msb.20134892Molecular Systems Biology (2014) 10:728.

Glutamine can play a critical role in cellular growth in multiple cancers. Glutamine‑addicted cancer cells are dependent on glutamine for viability, and their metabolism is reprogrammed for glutamine utilization through the tricarboxylic acid (TCA) cycle. Here, we have uncovered a missing link between cancer invasiveness and glutamine dependence. Using isotope tracer and bioenergetic analysis, we found that low‑invasive ovarian cancer (OVCA) cells are glutamine independent, whereas high‑invasive OVCA cells are markedly glutamine dependent. Consis‑tent with our findings, OVCA patients’ microarray data suggest that glutaminolysis correlates with poor survival. Notably, the ratio of gene expression associated with glutamine anabolism versus catabolism has emerged as a novel biomarker for patient prognosis. Significantly, we found that glutamine regulates the activation of STAT3, a mediator of signaling pathways which regulates cancer hallmarks in invasive OVCA cells. Our findings suggest that a combined approach of targeting high‑invasive OVCA cells by blocking glutamine's entry into the TCA cycle, along with targeting low‑invasive OVCA cells by inhibiting glutamine synthesis and STAT3 may lead to potential therapeutic approaches for treating OVCAs.

TCA

gln

lac

glc

Article SOURCEDATA

TRANSPARENTPROCESS

OPENACCESS

SynopsisWeak mitochondrial uncouplers prevent neoangiogenesis in vitro and in vivo by depleting cellular energy reserves in proliferating but not normal quiescent endothelial cells (ECs).

• NewvesselformationduringtumorgrowthrequiresECproliferationandincreasedoxidativephosphorylationtomeetthegreaterenergydemandduringangiogenesis.

• WeakmitochondrialuncouplerspreventneoangiogenesisbydepletingcellularenergyreservesinproliferatingbutnotnormalquiescentECs.

• ProliferatingECsaresensitizedtomitochondrialuncouplersbyareductioninmembranepotentialandlowerrespiratoryreservecapacity.

• GeneticaccumulationofmitochondrialDNAmutationsinmitochondrialmutatormicehighlightsthelinkbetweenreducedOxPhosactivityandimpairedangiogenicresponse.

• Weakmitochondrialuncouplerscouldbeclinicallyvaluableincontrollingpathologicalneoangiogenesiswhilesparingnormalvasculatureandcomplementingcytostaticdrugsincancertreatment.

Embelin inhibits endothelial mitochondrial respiration and impairs neoangiogenesis during tumor growth and wound healingOliverCoutelle*,1,Hue‑TranHornig‑Do2,AxelWitt3,MariaAndree3,LarsMSchiffmann1,MichaelPiekarek4,KerstinBrinkmann3,JensMSeeger3,MaximLiwschitz1,SatomiMiwa5,MichaelHallek1,MartinKrönke3,6,7,AleksandraTrifunovic6,SabineAEming4,6,7,RudolfJWiesner2,6,7,UlrichTHacker1,†andHamidKashkar3,6,7,†

1DepartmentIforInternalMedicine,UniversityofCologne,Cologne,Germany,2InstituteforVegetativePhysiologyUniversityofCologne,Cologne,Germany,3InstituteforMedicalMicrobiology,ImmunologyandHygiene,MedicalFaculty,UniversityofCologne,Cologne,Germany,4DepartmentofDermatology,UniversityofCologne,Cologne,Germany,5InstituteforAgeingandHealthNewcastleUniversity,NewcastleuponTyne,UK,6CologneExcellenceClusteronCellularStressResponsesinAging‑AssociatedDiseases(CECAD),MedicalFacultyUniversityofCologne,Cologne,Germany,7CenterforMolecularMedicineCologne(CMMC),Cologne,Germany

*Correspondingauthor.Tel:+492214787285;Fax:+492214787288;E‑mail:oliver.coutelle@gmail.com

DOI: 10.1002/emmm.201303016EMBO Molecular Medicine (2014) 6 (5):624-639

In the normal quiescent vasculature, only 0.01% of endothelial cells (ECs) are proliferating. However, this propor‑tion increases dramatically following the angiogenic switch during tumor growth or wound healing. Recent evidence suggests that this angiogenic switch is accompanied by a metabolic switch. Here, we show that prolifer‑ating ECs increasingly depend on mitochondrial oxidative phosphorylation (OxPhos) for their increased energy demand. Under growth conditions, ECs consume three times more oxygen than quiescent ECs and work close to their respiratory limit. The increased utilization of the proton motif force leads to a reduced mitochondrial membrane potential in proliferating ECs and sensitizes to mitochondrial uncoupling. The benzoquinone embe‑lin is a weak mitochondrial uncoupler that prevents neoangiogenesis during tumor growth and wound healing by exhausting the low respiratory reserve of proliferating ECs without adversely affecting quiescent ECs. We demonstrate that this can be exploited therapeutically by attenuating tumor growth in syngenic and xenograft mouse models. This novel metabolic targeting approach might be clinically valuable in controlling pathological neoangiogenesis while sparing normal vasculature and complementing cytostatic drugs in cancer treatment.

Research article SOURCEDATA

TRANSPARENTPROCESS

OPENACCESS

SynopsisIn this review, the authors introduce the concept of understanding and further maintaining metabolic flexibility as a way to limit spreading of metabolic disorders. By modifying function or biogenesis, the mitochondria perfectly illustrates this point as they adapt to different stress situations.

Mitochondrial response to nutrient availability and its role in metabolic diseaseArwenWGao1,CarlesCantó*,2andRiekeltHHoutkooper*,1

1LaboratoryGeneticMetabolicDiseases,AcademicMedicalCenter,Amsterdam,TheNetherlands,2NestléInstituteofHealthSciences,Lausanne,Switzerland

*Correspondingauthor.Tel:+41216326116;Fax:+41216326499;E‑mail:carlos.cantoalvarez@rd.nestle.com*Correspondingauthor.Tel:+31205666039;Fax:+31206962596;E‑mail:r.h.houtkooper@amc.nl

DOI: 10.1002/emmm.201303782EMBO Molecular Medicine (2014) 6 (5):580-589

Metabolic inflexibility is defined as an impaired capacity to switch between different energy substrates and is a hall‑mark of insulin resistance and type 2 diabetes mellitus (T2DM). Hence, understanding the mechanisms underlying proper metabolic flexibility is key to prevent the development of metabolic disease and physiological deterioration. An important downstream player in the effects of metabolic flexibility is the mitochondrion. The objective of this review was to describe how mitochondrial metabolism adapts to limited nutrient situations or caloric excess by changes in mitochondrial function or biogenesis, as well as to define the mechanisms propelling these changes. Altogether, this should pinpoint key regulatory points by which metabolic flexibility might be ameliorated in situ‑ations of metabolic disease.

Review

Drp1

MFN1Fusion

MFN2

OPA1

OPA1

OPENACCESS

SynopsisModern cells possess a sophisticated metabolic network, but its origins remain largely unknown. Reconstructing scenarios of the Archean ocean, we observe chemical reactions reminiscent of modern metabolic sequences, indicating that metabolism could be of prebiotic origin.

• Metabolitesofglycolysisandthepentosephosphateundergonon‑enzymaticinterconversionreactions.

• MetalionsabundantlyfoundinsedimentsoftheprebioticArcheanocean,predominantlyFe(II),catalyseadditionalsugarphosphateinterconversionreactions.

• ReactionscatalysedbytheArcheanoceanmetalsresembleenzyme‑catalysedreactionsfoundinthemodernglycolyticandpentosephosphatepathways.

• TheobservedreactionsareacceleratedandgainspecificityinconditionssimulatingtheArcheanocean.

Non-enzymatic glycolysis and pentose phosphate pathway-like reactions in a plausible Archean oceanMarkusAKeller1,AlexandraVTurchyn2andMarkusRalser*,1,3

1DepartmentofBiochemistryandCambridgeSystemsBiologyCentre,UniversityofCambridge,Cambridge,UK,2DepartmentofEarthSciences,UniversityofCambridge,Cambridge,UK,3DivisionofPhysiologyandMetabolism,MRCNationalInstituteforMedicalResearch,MillHillLondon,UK

*Correspondingauthor.Tel:+441223761346;Fax:+441223766002;E‑mail:mr559@cam.ac.uk

DOI: 10.1002/msb.20145228Molecular Systems Biology (2014) 10:725

The reaction sequences of central metabolism, glycolysis and the pentose phosphate pathway provide essential precursors for nucleic acids, amino acids and lipids. However, their evolutionary origins are not yet understood. Here, we provide evidence that their structure could have been fundamentally shaped by the general chemical environments in earth's earliest oceans. We reconstructed potential scenarios for oceans of the prebiotic Archean based on the composition of early sediments. We report that the resultant reaction milieu catalyses the intercon‑version of metabolites that in modern organisms constitute glycolysis and the pentose phosphate pathway. The 29 observed reactions include the formation and/or interconversion of glucose, pyruvate, the nucleic acid precursor ribose‑5‑phosphate and the amino acid precursor erythrose‑4‑phosphate, antedating reactions sequences simi‑lar to that used by the metabolic pathways. Moreover, the Archean ocean mimetic increased the stability of the phosphorylated intermediates and accelerated the rate of intermediate reactions and pyruvate production. The catalytic capacity of the reconstructed ocean milieu was attributable to its metal content. The reactions were particularly sensitive to ferrous iron Fe(II), which is understood to have had high concentrations in the Archean oceans. These observations reveal that reaction sequences that constitute central carbon metabolism could have been constrained by the iron‑rich oceanic environment of the early Archean. The origin of metabolism could thus date back to the prebiotic world.

Article TRANSPARENTPROCESS

OPENACCESS

SynopsisA primary cardiac myocyte defect leads to aberrant lipid accumulation and signaling in the liver. The resulting hepatic phenotype impacts cardiac function. Normalization of heart lipid delivery or inhibition of gluconeogenesis improves ventricular function.

• Geneticheartdiseasecausesmetabolicabnormalitiesintheliver.• ThereisreducedtriglycerideclearancebytheHCMheart.• Accumulatingplasmatriglyceridesaresequesteredbyhepatocytes.• Activationofgluconeogenesisexacerbatescardiacpathology.

Metabolic crosstalk between the heart and liver impacts familial hypertrophic cardiomyopathyJasonAMagida1andLeslieALeinwand*,1

1DepartmentofMolecular,CellularandDevelopmentalBiology,BioFrontiersInstituteUniversityofColoradoatBoulder,Boulder,CO,USA

*Correspondingauthor.Tel:+13034927606;Fax:+13034928907;E‑mail:leinwand@colorado.edu

DOI: 10.1002/emmm.201302852EMBO Molecular Medicine (2014) 6 (4):482-495

Familial hypertrophic cardiomyopathy (HCM) is largely caused by dominant mutations in genes encoding cardi‑ac sarcomeric proteins, and it is etiologically distinct from secondary cardiomyopathies resulting from pressure/volume overload and neurohormonal or inflammatory stimuli. Here, we demonstrate that decreased left ventricular contractile function in male, but not female, HCM mice is associated with reduced fatty acid translocase (CD36) and AMP‑activated protein kinase (AMPK) activity. As a result, the levels of myocardial ATP and triglyceride (TG) content are reduced, while the levels of oleic acid and TG in circulating very low density lipoproteins (VLDLs) and liver are increased. With time, these metabolic changes culminate in enhanced glucose production in male HCM mice. Remarkably, restoration of ventricular TG and ATP deficits via AMPK agonism as well as inhibition of gluco‑neogenesis improves ventricular architecture and function. These data underscore the importance of the systemic effects of a primary genetic heart disease to other organs and provide insight into potentially novel therapeutic interventions for HCM.

Research article TRANSPARENTPROCESS

OPENACCESS

SynopsisNicotinamide riboside (vitamin B3) delays the progression of mitochondrial myopathy by preventing pathology‑associated mitochondrial ultrastructure, improving mitochondrial DNA stability and further stimulating mitochondrial unfolded protein response.

• Nicotinamideriboside,vitaminB3,delaystheprogressionofmitochondrialmyopathy.• Nicotinamideribosidecurespathology‑associatedmitochondrialultrastructure.• NicotinamideribosideimprovesmitochondrialDNAstability.• Mitochondrialdiseaseinducesmitochondrialunfoldedproteinresponse,furtherenhancedby

nicotinamideriboside.• Nicotinamideribosideisapromisingtreatmentforadult‑onsetmitochondrialmyopathy.

Effective treatment of mitochondrial myopathy by nicotinamide riboside, a vitamin B3NahidAKhan1,MariAuranen1,2,†,IlsePaetau1,†,EijaPirinen3,4,LiliyaEuro1,SaaraForsström1,LottaPasila1,VidyaVelagapudi5,ChristopherJCarroll1,JohanAuwerx3andAnuSuomalainen*,1,2,6

1MolecularNeurology,ResearchProgramsUnit,UniversityofHelsinki,Helsinki,Finland,2DepartmentofNeurology,HelsinkiUniversityCentralHospital,Helsinki,Finland,3LaboratoryofIntegrativeSystemsPhysiology,ÉcolePolytechniqueFédéraledeLausanne,Lausanne,Switzerland,4BiotechnologyandMolecularMedicine,A.I.VirtanenInstituteforMolecularSciencesBiocenterKuopioUniversityofEasternFinland,Kuopio,Finland,5MetabolomicsUnit,InstituteforMolecularMedicineFinlandFIMM,Helsinki,Finland,6NeuroscienceResearchCentreUniversityofHelsinki,Helsinki,Finland

*Correspondingauthor.Tel:+358947171965;Fax:+358947171964;E‑mail:anu.wartiovaara@helsinki.fi†Theseauthorscontributedequallytothemanuscript.

DOI: 10.1002/emmm.201403943EMBO Molecular Medicine (2014) 6 (6):721-731

Nutrient availability is the major regulator of life and reproduction, and a complex cellular signaling network has evolved to adapt organisms to fasting. These sensor pathways monitor cellular energy metabolism, especially mitochondrial ATP production and NAD+/NADH ratio, as major signals for nutritional state. We hypothesized that these signals would be modified by mitochondrial respiratory chain disease, because of inefficient NADH utilization and ATP production. Oral administration of nicotinamide riboside (NR), a vitamin B3 and NAD+ precursor, was previously shown to boost NAD+ levels in mice and to induce mitochondrial biogenesis. Here, we treated mitochondrial myopathy mice with NR. This vitamin effectively delayed early‑ and late‑stage disease progression, by robustly inducing mitochondrial biogenesis in skeletal muscle and brown adipose tissue, prevent‑ing mitochondrial ultrastructure abnormalities and mtDNA deletion formation. NR further stimulated mitochon‑drial unfolded protein response, suggesting its protective role in mitochondrial disease. These results indicate that NR and strategies boosting NAD+ levels are a promising treatment strategy for mitochondrial myopathy.

Research article SOURCEDATA

TRANSPARENTPROCESS

OPENACCESS

SynopsisPersonalized GEMs for six hepatocellular carcinoma patients are reconstructed using proteomics data and a task‑driven model reconstruction algorithm. These GEMs are used to predict antimetabolites preventing tumor growth in all patients or in individual patients.

• Thepresenceofproteinsencodedby15,841genesintumorsfrom27HCCpatientsisevaluatedbyimmunohistochemistry.

• PersonalizedGEMsforsixHCCpatientsandGEMsfor83healthycelltypesarereconstructedbasedonHMR2.0andthetINITalgorithmfortask‑drivenmodelreconstruction.

• 101antimetabolitesarepredictedtoinhibittumorgrowthinallpatients.Antimetabolitetoxicityistestedusingthe83celltype‑specificGEMs.

• Anl‑carnitineanaloginhibitstheproliferationofHepG2cells.

Identification of anticancer drugs for hepatocellular carcinoma through personalized genome-scale metabolic modelingRasmusAgren1,†,AdilMardinoglu1,†,AnnaAsplund2,CarolineKampf2,MathiasUhlen3,4,JensNielsen*1,3.

1DepartmentofChemicalandBiologicalEngineering,ChalmersUniversityofTechnology,Gothenburg,Sweden,2DepartmentofImmunology,GeneticsandPathologyScienceforLifeLaboratory,UppsalaUniversity,Uppsala,Sweden,3ScienceforLifeLaboratoryKTH–RoyalInstituteofTechnology,Stockholm,Sweden,4DepartmentofProteomicsKTH–RoyalInstituteofTechnology,Stockholm,Sweden

*Correspondingauthor.Tel:+46317723804;Fax:+46317723801;E‑mail:nielsenj@chalmers.se†Theseauthorscontributedequallytothiswork.

DOI: 10.1002/msb.145122Molecular Systems Biology (2014) 10:721

Genome‑scale metabolic models (GEMs) have proven useful as scaffolds for the integration of omics data for under‑standing the genotype–phenotype relationship in a mechanistic manner. Here, we evaluated the presence/absence of proteins encoded by 15,841 genes in 27 hepatocellular carcinoma (HCC) patients using immunohistochemistry. We used this information to reconstruct personalized GEMs for six HCC patients based on the proteomics data, HMR 2.0, and a task‑driven model reconstruction algorithm (tINIT). The personalized GEMs were employed to identify anticancer drugs using the concept of antimetabolites; i.e., drugs that are structural analogs to metabo‑lites. The toxicity of each antimetabolite was predicted by assessing the in silico functionality of 83 healthy cell type‑specific GEMs, which were also reconstructed with the tINIT algorithm. We predicted 101 antimetabolites that could be effective in preventing tumor growth in all HCC patients, and 46 antimetabolites which were specific to individual patients. Twenty‑two of the 101 predicted antimetabolites have already been used in different cancer treatment strategies, while the remaining antimetabolites represent new potential drugs. Finally, one of the identi‑fied targets was validated experimentally, and it was confirmed to attenuate growth of the HepG2 cell line.

Article

Personalized GEMs

6 HCC patients

TRANSPARENTPROCESS

OPENACCESS

SynopsisThe impact of oncogene activation and hypoxia on energy metabolism is analyzed by integrating quantitative measurements into a redox‑balanced metabolic flux model. Glutamine‑driven oxidative phosphorylation is found to be a major ATP source even in oncogene‑expressing or hypoxic cells.

• TheintegrationofoxygenuptakemeasurementsandLC‑MS‑basedisotopetraceranalysesinaredox‑balancedmetabolicfluxmodelenabledquantitativedeterminationofenergygenerationpathwaysinculturedcells.

• Intransformedmammaliancells,eveninhypoxia(1%oxygen),oxidativephosphorylationproducesthemajorityofATP.

• TheoncogeneRassimultaneouslyincreasesglycolysisanddecreasesoxidativephosphorylation,thusresultinginnonetincreaseinATPproduction.

• Glutamineisthemajorsourceofhigh‑energyelectronsforoxidativephosphorylation,especiallyuponRasactivation.

Glutamine-driven oxidative phosphorylation is a major ATP source in transformed mammalian cells in both normoxia and hypoxiaJingFan1,JurreJKamphorst1,RobinMathew2,3,MichelleKChung1,EileenWhite2,3,4,TomerShlomi5,†andJoshuaDRabinowitz*,1,2,6,†

1DepartmentofChemistryandLewis‑SiglerInstituteforIntegrativeGenomics,PrincetonUniversity,Princeton,NJ,USA,2TheCancerInstituteofNewJersey,NewBrunswick,NJ,USA,3UniversityofMedicineandDentistryofNewJersey,RobertWoodJohnsonMedicalSchool,Piscataway,NJ,USA,4DepartmentofMolecularBiologyandBiochemistry,RutgersUniversity,Piscataway,NJ,USA,5DepartmentofComputerScience,Technion,Haifa,Israel,6DepartmentofMolecularBiology,PrincetonUniversity,Princeton,NJ,USA

*Correspondingauthor.DepartmentsofChemistryandIntegrativeGenomics,PrincetonUniversity,241CarlIcahnLaboratory,Princeton,NJ08544,USA.Tel.:+16092588985;Fax:+16092583565;E‑mail:joshr@princeton.edu†Theseauthorscontributedequallytothiswork.

DOI: 10.1038/msb.2013.65Molecular Systems Biology (2013) 9:712

Mammalian cells can generate ATP via glycolysis or mitochondrial respiration. Oncogene activation and hypoxia promote glycolysis and lactate secretion. The significance of these metabolic changes to ATP production remains however ill defined. Here, we integrate LC‑MS‑based isotope tracer studies with oxygen uptake measurements in a quantitative redox‑balanced metabolic flux model of mammalian cellular metabolism. We then apply this approach to assess the impact of Ras and Akt activation and hypoxia on energy metabolism. Both oncogene activa‑tion and hypoxia induce roughly a twofold increase in glycolytic flux. Ras activation and hypoxia also strongly decrease glucose oxidation. Oxidative phosphorylation, powered substantially by glutamine‑driven TCA turning, however, persists and accounts for the majority of ATP production. Consistent with this, in all cases, pharmaco‑logical inhibition of oxidative phosphorylation markedly reduces energy charge, and glutamine but not glucose removal markedly lowers oxygen uptake. Thus, glutamine‑driven oxidative phosphorylation is a major means of ATP production even in hypoxic cancer cells.

Report TRANSPARENTPROCESS

OPENACCESS

Glucose

Lactate

ATP

Glutamine

NADH

NADH

O2

NOTES

NOTES

NOTES

EMBO Molecular Medicine

embor.embopress.org | emboj.embopress.org | embomolmed.embopress.org | msb.embopress.org