integrationofmetabolomicsandtranscriptomicsrevealeda fatty … · maryland medical center...

Human Cancer Biology

IntegrationofMetabolomics andTranscriptomicsRevealedaFatty Acid Network Exerting Growth Inhibitory Effects inHuman Pancreatic Cancer

Geng Zhang1, Peijun He1, Hanson Tan1, Anuradha Budhu1, Jochen Gaedcke6, B. Michael Ghadimi6,Thomas Ried2, Harris G. Yfantis3, Dong H. Lee3, Anirban Maitra4, Nader Hanna5,H. Richard Alexander5, and S. Perwez Hussain1

AbstractPurpose: To identify metabolic pathways that are perturbed in pancreatic ductal adenocarcinoma

(PDAC), we investigated gene-metabolite networks with integration of metabolomics and transcriptomics.

ExperimentalDesign:Weconducted globalmetabolite profiling analysis on two independent cohorts of

resected PDAC cases to identify critical metabolites alteration that may contribute to the progression of

pancreatic cancer. We then searched for gene surrogates that were significantly correlated with the key

metabolites, by integrating metabolite and gene expression profiles.

Results: Fifty-five metabolites were consistently altered in tumors as compared with adjacent nontumor

tissues in a test cohort (N¼ 33) and an independent validation cohort (N¼ 31).Weighted network analysis

revealed a unique set of free fatty acids (FFA) that were highly coregulated and decreased in PDAC. Pathway

analysis of 157 differentially expressed gene surrogates revealed a significantly altered lipid metabolism

network, including key lipolytic enzymes PNLIP, CLPS, PNLIPRP1, and PNLIPRP2. Gene expressions of

these lipases were significantly decreased in pancreatic tumors as compared with nontumor tissues, leading

to reduced FFAs. More importantly, a lower gene expression of PNLIP in tumors was associated with poorer

survival in two independent cohorts. We further showed that two saturated FFAs, palmitate and stearate,

significantly induced TRAIL expression, triggered apoptosis, and inhibited proliferation in pancreatic cancer

cells.

Conclusions:Our results suggest that impairment in a lipolytic pathway involving lipases, and a unique

set of FFAs,mayplay an important role in thedevelopment andprogressionof pancreatic cancer andprovide

potential targets for therapeutic intervention. Clin Cancer Res; 19(18); 4983–93. �2013 AACR.

IntroductionPancreatic cancer is the fourth leading cause of cancer-

related death in the United States with an estimated 44, 920new cases and 37, 390 deaths in 2013 (1). Pancreatic cancercases and deaths have been on the rise since 1998. The

median survival of all pancreatic ductal adenocarcinoma(PDAC) cases is less than 6months, and only 6%of patientssurvive 5 years after diagnosis. The mortality rate has notimproved significantly, due to late diagnosis and resistanceto available chemotherapy. Therefore, a better understand-ing of molecular mechanisms of disease progression anddiscovery of novel therapeutic targets are desperately need-ed to improve outcomes in patients with PDAC.

Genetic alterations have been extensively characterizedin pancreatic cancer. We and others have previously iden-tified gene markers of PDAC which have prognostic andtherapeutic significance (2, 3). However, the impact ofthose gene alterations on metabolism and gene-metabolitenetworks in PDAC has not been clearly defined. Severaloncogenes and tumor suppressors such as c-Myc and P53control the activity of different metabolic pathways tosupport the metabolic transformation of a cancer cell (4).A growing body of evidence showed a strong connectionbetween cancer and metabolism, including the discoverythat some key metabolic enzymes such as succinate dehy-drogenase, fumarate hydratase, isocitrate dehydrogenase,and phosphoglycerate dehydrogenase, if mutated, could

Authors' Affiliations: 1Pancreatic Cancer Unit, Laboratory of HumanCarcinogenesis, Center for Cancer Research, 2Genetics Branch, NationalCancer Institute, NIH, Bethesda; 3Pathology and Laboratory Medicine,Baltimore Veterans Affairs Medical Center, 4The Sol Goldman PancreaticCancer Research Center, Johns Hopkins University School of Medicine,5Division of Surgical Oncology, TheDepartment ofSurgery and theMarleneand Stewart GreenebaumCancer Center, University of Maryland School ofMedicine, Baltimore, Maryland; and 6Department of General and VisceralSurgery, University Medicine, G€ottingen, Germany

Note: Supplementary data for this article are available at Clinical CancerResearch Online (http://clincancerres.aacrjournals.org/).

CorrespondingAuthor:S.PerwezHussain, 37ConventDrive, Building 37,Room 3044B, Pancreatic Cancer Unit Laboratory of Human Carcinogen-esis, National Cancer Institute, NIH, Bethesda, MD 20892. Phone: 301-402-3431; Fax: 301-496-0497; E-mail: [email protected]

doi: 10.1158/1078-0432.CCR-13-0209

�2013 American Association for Cancer Research.

ClinicalCancer

Research

www.aacrjournals.org 4983

on October 29, 2020. © 2013 American Association for Cancer Research. clincancerres.aacrjournals.org Downloaded from

Published OnlineFirst August 5, 2013; DOI: 10.1158/1078-0432.CCR-13-0209

http://clincancerres.aacrjournals.org/

lead to different forms of cancer (5). The development ofpancreatic cancer has also been linked to abnormal glucosemetabolism, which is induced by long-term type-II diabe-tes, a known risk factor for pancreatic cancer (6). Metabo-lites are central in intermediary metabolism, and theyprovide substrates for biologic processes, and have an activerole in regulating cell cycle, proliferation, and apoptosis (7).Recently, there has been an increased interest in globalanalysis of metabolites for cancer biomarker discovery andidentification of potential novel therapeutic targets. Newtechnologies applying chromatography-mass spectrometry(MS) provide sensitive and reproducible detection of hun-dreds to thousands of metabolites in a single biofluid ortissue sample, and allows nontargeted high-throughputmetabolic profiling analysis (8). Furthermore, integrationof comprehensive gene expression profile with metabolicprofiling has been shown to be an innovative way to revealthe complex regulatory networks involving genes and met-abolic pathways in cancers (9, 10).

In the present study, we conducted global metaboliteprofiling analysis in two independent cohorts of PDACcases to identify critical metabolite alterations that maycontribute to the progression of pancreatic cancer. We thensearched for gene surrogates that were significantly corre-lated with the key metabolites using transcriptomic profil-ing data of the same samples in the test cohort from ourprevious study (2). The integrative analysis of metabolo-mics and transcriptomic data revealed a lipid metabolismnetwork involving four lipases and a unique set of free fattyacids (FFA) that may play an important role in pancreatictumor progression and could provide potential targets fortherapeutic intervention.

Materials and MethodsTissue collection

Primary pancreatic tumor and adjacent nontumor tissueswere collected from patients with PDAC at the University of

Medicine (G€ottingen,Germany), and from theUniversity ofMaryland Medical Center (Baltimore, MD) through theNCI-UMD resource contract. Tissues were flash frozenimmediately after surgery. Demographic and clinical infor-mation for each tissue donor, including age, sex, clinicalstaging, resection margin status, survival times from diag-nosis, and receipt of adjuvant chemotherapywere collected.Tumorhistopathologywas classified according to theWorldHealth Organization Classification of Tumor system (11).Use of these clinical specimens was reviewed and approvedby the NCI-Office of the Human Subject Research (OHSR,Exempt # 4678) at the NIH (Bethesda, MD).

Metabolic profiling of PDACMetabolic profiling of PDAC samples (tumor and adja-

cent nontumor tissues) was carried out at Metabolon Inc.using the general protocol as outlined earlier (refs. 12, 13;see more details in the Supplementary File). Metabolonanalytic platform incorporates two separate ultra-high per-formance liquid chromatography/tandem mass spectrom-etry (UHPLC/MS/MS2) injections and one gas chromatog-raphy/mass spectrometry injection per sample. TheUHPLCinjections are optimized for basic species and acidic species.This integrated platform enabled the high-throughput col-lection and relative quantitative analysis of analytic dataand identified a large number and broad spectrum ofmolecules with a high degree of confidence (12). A totalof 469 known metabolites were measured.

Weighted coexpression network analysisWeighted coexpression network analysis (WGCNA) has

been implemented in R, a free and open source statisticalprogramming language (14). We followed the protocols ofWGCNA to create metabolite networks. Briefly, for eachmetabolite profiling dataset, Pearson correlation coeffi-cients were calculated for all pairwise comparisons of meta-bolites across all tumor samples. The resulting Pearsoncorrelation matrix was transformed into an adjacencymatrix using a power function, which resulted in aweightednetwork (14). WGCNA defines modules as a group ofdensely interconnected molecules with high topologicaloverlap in weighted network analysis. For each dataset, weused average linkage hierarchical clustering with a dynamictree-cutting algorithm to identify modules on the basis ofthe topological overlap dissimilarity measure (15).

Ingenuity pathways analysisCanonical pathway analysis identified the pathways from

the Ingenuity Pathways Analysis library of canonical path-ways that were most significant to the dataset. The associ-ation between the dataset and the canonical pathway wasmeasured in 2 ways: (i) The statistical significance: Fischerexact test was used to calculate a P value, determining theprobability that the association between the genes in thedataset and the canonical pathway is explained by chancealone; (ii) a ratio of the number of genes from the datasetthat map to the given pathway divided by the total numberof genes in that canonical pathway (16).

Translational RelevancePancreatic ductal adenocarcinoma (PDAC) is one of

the most lethal malignancies worldwide. Altered metab-olism is considered as one of the hallmarks of cancer.Therefore, a better understanding of metabolic dysregu-lation in pancreatic cancer could lead to the discovery ofnovel therapeutic targets. In the present study, we inves-tigated metabolic dysregulation and gene-metabolitenetworks in pancreatic cancer, with integration of meta-bolomics and transcriptomics. Our results suggest thatthe impaired lipolytic network involving four lipases anda unique set of fatty acids may play an important role inthe development and progression of pancreatic cancer.Furthermore, our study showed that integration of omicsdata provide a systems level perspective of pancreaticcancer that could facilitate the development of noveltreatments for this disease.

Zhang et al.

Clin Cancer Res; 19(18) September 15, 2013 Clinical Cancer Research4984




RNA isolation and transcriptome profilingRNA from frozen tissue samples was extracted using

standard TRIzol (Invitrogen) protocol. RNA quality wasconfirmedwith the Agilent 2100 Bioanalyzer (Agilent Tech-nologies) before the microarray gene expression profiling.Tumors and paired nontumor tissues from the Germanycohort were profiled using the Affymetrix GeneChipHuman 1.0 ST arrays, according to the manufacturer’sprotocol at the LMT microarray core facility at NationalCancer Institute(Frederick, MD). All arrays were RMA nor-malized and gene expression summaries were created foreach gene by averaging all probe sets for each gene usingPartek Genomics Suite 6.5. All data analysis was conductedon gene-summarized data. The microarray gene expressiondata has been deposited in the National Center for Bio-technology Information’s (NCBI) Gene Expression Omni-bus (GEO; http://www.ncbi.nlm.nih.gov/geo) with acces-sion number GSE28735.

Quantitative RT-PCRTotal RNA was reverse transcribed using High-Capacity

cDNA Reverse Transcription Kit (Applied Biosystems).Quantitative Reverse Transcription PCR (qRT-PCR) reac-tions in 384-well plates were carried out using TaqManGene Expression Assays on an ABI Prism 7900HT SequenceDetection instrument from Applied Biosystems. Expressionlevels of GAPDHwere used as the endogenous controls. Allassays were conducted in triplicate. For quality control, anysamplewith a gene cycle valuemore than 36was consideredof poor quality and removed. All the primers for qRT-PCR inthe present study were purchased from Applied Biosystems(Supplementary Table S1).

Cell lines and culture conditionsHuman pancreatic carcinoma cell lines hTERT-HPNE

(CRL-4023), MIApaca2 (CRL-1420) and Panc1 (CRL-1469) were obtained from American Type Culture Col-lection. Cells were maintained in GIBCO RPMI Media1640 supplemented with GlutaMAX-I (Invitrogen), pen-icillin–streptomycin (50 IU/mL and 50 mg/mL, respec-tively), and 10% (v/v) fetal calf serum (FCS). Fatty acidswere purchased from Sigma. The stock solutions of fattyacids bound to bovine serum albumin (BSA) were pre-pared as described earlier (17). A 5% FFA-free BSA solu-tion was prepared in H2O and dissolved for 30 minutes at55�C in a water bath. The appropriate amount of 50mmol/L FFA stock solution was added to BSA solutionand incubated for 8 hours at 37�C under nitrogen atmo-sphere to prevent oxidization. The FFA/BSA stock solu-tion was then cooled to 25�C, filter sterilized, and storedat �20�C.

MTT assayCells were seeded in 96-well plates (3,000–5000 cells/

well) and incubated for 2 to10days. Then, theMTT solutionwas added and incubated for 4 hours. The solution wasaspirated and 100 mL DMSO was added to each well. Theabsorbance was measured at 570 and 650 nm.

Apoptosis assayPanc1 andMIApaca2 cells were seeded and incubated for

24 hours in standard medium. After 12 hours of serumstarvation, cells were incubated in serum-free mediumwithBSA-bound fatty acids or BSA control for 12 hours and 24hours. Caspase activity was measured by Apo-ONE Homo-geneous Caspase-3/7 Assay (Promega) and potency of cas-pase activation was calculated, compared with control cells.

Quantification of TRAIL protein level by ELISATRAIL protein level was directly quantified in pancre-

atic cancer cells treated with 0.1 mmol/L FFA or BSA for24 hours. Cells were lysed (5 � 105 cells/100 mL lysatebuffer) and TRAIL production in cell lysates was mea-sured using Quantikine Human TRAIL/TNFSF10 ELISAKIT (R&D Systems) according to the manufacturer’sinstructions. Optical density of each well was then deter-mined using a microplate reader set to 450 nm. TRAILconcentrations were calculated using a standard curve andlinear regression analysis.

Statistical analysisA student t test was used to compare metabolite or gene

expression between tumors and nontumor tissues usingGraphPad Prism 5.0 (GraphPad Software Inc). Correlationanalysis and Kaplan–Meier analysis were conducted withGraphPad Prism 5.0 (GraphPad Software Inc). Fisher exacttest, correlation analysis, and Cox Proportional-hazardsregression analysis were conducted using Stata 11 (Stata-Corp LP). Univariate Cox regression was conducted ongenes and clinical covariates to examine the influence ofeach on patient survival. For these analyses, resection mar-gin statuswas dichotomized as positive (R1) versus negative(R0); TNM staging was dichotomized on the basis of non-metastatic (I-IIA) versus metastatic (IIB-IV) disease; histo-logic grade was dichotomized on the basis of well andmoderately differentiated (G1 and 2) versus poorly differ-entiated (G3 and 4). All Cox regression models were testedfor proportional hazards assumptions based on Schoenfeldresiduals, and no model violated these assumptions. Thestatistical significance was defined as P < 0.05. All P valuesreported were two-sided.

ResultsMetabolomics of PDAC

The characteristics of the patients with PDAC in a testcohort (N¼33) and a validation cohort (N¼31) are shownin Supplementary Table S1. The two cohorts were similar inTNM staging, resection margin status, grade, and cancer-specificmortality (P¼ 0.76, Kaplan–Meier log rank)with 1-year survival rate of 50.8% for the test cohort and 51% forthe validation cohort.

Metabolite profiling was conducted using liquid and gaschromatography coupled with mass spectrometry to iden-tify and statistically compare the relative metabolite expres-sion levels between tumor and nuntumor tissues fromPDAC cases. We identified 55 metabolites that were differ-entially expressed in tumors as compared with nontumor

Integration of Metabolomics and Transcriptomics in PDAC

www.aacrjournals.org Clin Cancer Res; 19(18) September 15, 2013 4985




tissues (P < 0.01) in both test and validation cohorts(Supplementary Table S2).

Weighted network analysis identified a unique set offatty acids that are coregulated in PDAC

We constructed coexpression networks using the 55metabolite profiling data in two independent cohorts.According to recently described methodology (14), theconnectivity (k) was determined for all metabolites in the

network by taking the sum of their connection strengths(coexpression similarity) with all other nodes in the net-work. To identify modules of highly coregulated metabo-lites, we used average linkage hierarchical clustering togroup metabolites based on the topologic overlap of theirconnectivity (see Materials and Methods for details).WGCNA identified three modules of highly connectedmetabolites. Each module was assigned a unique coloridentifier (Fig. 1A), with the remaining, poorly connected

Figure 1. Integration of metabolomics and transcriptomics revealed altered lipid metabolism pathways in PDAC. A, weighted network analysis showedthat the turquoise-colored module was highly conserved in both test and validation cohort. 8 fatty acids (indicated by stars) with high connectivity(IMconn > 6, Table 1) in both cohorts represent the main hubs in the metabolite network that were chosen for integration analysis. B, canonical pathwaysanalysis identified 10 pathways from the Ingenuity Pathways Analysis (IPA) library of canonical pathways that were most significant (P < 0.05) to thedata set of 157 surrogate genes for the unique set of 8 fatty acids in PDAC. Glycerolipid metabolism was the most significantly enriched pathway. C,IPA analysis revealed a lipid metabolism network involving 4 lipases and fatty acids highlighted here, that may potentially regulate and interact with apoptosissignaling pathways. Lipolytic genes PNLIP, CLPS, PNLIRP1, and PNLIPRP2 from the surrogate gene dataset, play central roles in lipolysis.

Zhang et al.





metabolites colored gray. In a topological overlap matrix(TOM) plot, the increasing color intensity indicates higherconnectivity among metabolites in the network (Fig. 1A).Metabolites with the greatest connectivity index representnetwork "hubs" and are localized in the center of individualmodules. Highly connected and correlated hubmetabolitesare often sharing commonpathways and tightly coregulatedwithin the same metabolism networks (18).Because of the indicated importance of hubs in the

network, we ranked metabolites within turquoise module(Supplementary Table S3), based on their intramodularconnectivity to identify module hubs (Table 1). Eight meta-boliteswith high connectivity (IMconn>6) in the turquoisemodule represented the main hubs in both test and vali-dation cohorts. Interestingly, this set of highly coregulatedhub metabolites are all free fatty acids, suggesting that fattyacid metabolism may be significantly altered in PDAC.

Integration of metabolomics and transcriptomicsrevealed altered lipid metabolism pathway in PDACTo define genetic alterations and themolecular pathways

associated with 8 fatty acids identified from WGCNA, wefirst searched for gene surrogates that were significantlycorrelated with these 8 fatty acids within the same 33samples in test cohort. Using transcriptomic profiling datafrom our previous study (2), we conducted Pearson corre-lation analysis between gene expression and metaboliteprofile data of 8 fatty acids. These analyses identified 157gene surrogates that were highly correlated with the set of 8fatty acids (Pearson correlation P < 0.01) and differentiallyexpressed between tumors and nontumors (t test, FDR-corrected P < 0.01, |fold change|>1.5, Supplementary TableS4).Furthermore, Ingenuity Pathways Analysis showed that

157 gene surrogates of 8 fatty acids were highly enriched forglycerolipid metabolism, axonal guidance signaling, starchand sucrose metabolism, intrinsic prothrombin activationpathway, and other pathways. The glycerolipidmetabolismpathway is the top hit with Fisher exact test P value of 0.001,representing the most significant pathway that is associatedwith our dataset (Fig. 1B). Significant enrichmentswere alsoobserved for surrogate genes associated with apoptosis andWnt signaling, which are known to play critical roles intumorigenesis. IPA’s network analysis further identifiedinteracting modules involved in lipid metabolism andapoptosis signaling networks, which include fatty acids andtheir surrogate genes PNLIP (pancreatic lipase), CLPS (coli-pase, pancreatic), PNLIPRP1 (pancreatic lipase-related pro-tein 1), and PNLIPRP2 (pancreatic lipase-related protein2; Fig. 1C).

Lypolytic enzymes PNLIP, CLPS, PNLIPRP1, andPNLIPRP2 are significantly decreased in PDACIt should be noted that surrogate genes PNLIP, CLPS,

PNLIPRP1, and PNLIPRP2 which encode key lipolyticenzymes playing central roles in glycerolipid metabolism,were decreased in tumors as compared with adjacent non-tumor tissues in the test cohort (Fig. 2A), which is consistent

with the downregulation of FFAs in pancreatic tumors(Table 1).

To validate the association of these lipases with the set of8 fatty acids identified in metabolic profiling, we then usedthe validation cohort of PDAC cases (N ¼ 31) to examinethe gene expression of PNLIP, CLPS, PNLIPRP1, andPNLIPRP2 by qRT-PCR.Our data confirmed that the expres-sion of all 4 lipases were decreased in tumors, as comparedwith surrounding nontumor tissues in two independentcohorts (Fig. 2), and consistently all 8 fatty acids were alsodecreased in tumors from both cohorts (Table 2). Particu-larly, qRT-PCR data in the validation cohort showedapproximately 100- to 1,000-fold decrease in the geneexpression of these lipases in pancreatic tumors. Our dataalso showed that gene expression of PNLIP, CLPS,PNLIPRP1, andPNLIPR2 are lower inPDACcell lines Panc1and MIApaca2, as compared with nontumorigenic hTERT-HPNE cells (Supplementary Fig. S1). These data are con-sistent with other publicly available transcriptional profil-ing data in Oncomine database (Supplementary Table S5),suggesting that the gene expression of these lipases arepotential diagnostic markers for PDAC.

Pancreatic lipases directly hydrolyse triglycerolipids intofatty acid and glycerol, therefore regulating fatty acid turn-over and signaling (19). In both test and validation cohorts,gene expressions of these lipases are significantly correlatedwith the set of 8 FFAs (Supplementary Table S6), indicatingthat markedly decreased expression of PNLIP, CLPS,PNLIPRP1, and PNLIPRP2may have a role in reduced fattyacid level in pancreatic tumors.

PNLIP is a predictor of cancer-specific mortality inPDAC

There is no significant association between 8 fatty acidsand cancer-specificmortality in both cohorts of PDAC cases(data not shown). We then tested whether the 4 lipolyticgenes were associated with patient outcome. The associa-tion of gene expression with cancer-specific mortality wasevaluated in both test and validation cohort using Coxregression analysis, in which we dichotomized high andlow gene expression as values above and below themedian.In our study, univariateCox regression analysis (Table 2) forall cases showed that only PNLIP was associated withprognosis in both test cohort [ HR 0.36; 95% confidenceinterval (CI), 0.15–0.88; P ¼ 0.023] and validation cohort(HR 0.36; 95% CI, 0.15–0.87; P ¼ 0.02). Therefore, surro-gate gene PNLIP maybe a prognostic marker for pancreaticcancer.

Palmitate and stearate inhibits cell growth in vitroIPA Canonical Pathway analysis identified an enriched

set of interactions between the lipid metabolism and theapoptosis signaling pathway (Fig. 1B and C), leading tothe hypothesis that fatty acids may regulate apoptosisand cell growth of pancreatic tumors. Therefore, westudied the effects of this unique set of FFAs on theproliferation of human pancreatic cancer cell lines Panc1and MIApaca2. Palmitate, stearate, linoleate, and oleate






were chosen because they are the most abundant fattyacids in animals (17). MTT cell proliferation assaysshowed that palmitate and stearate significantly inhibitPanc1 and MIApaca2 cell growth in a dose-dependentmanner. In contrast, linoleate and oleate had little effecton cell proliferation in both cells (Fig. 3A). Cell countingand bromodeoxyuridine assays, in addition to MTTassay, also showed significant growth inhibitory effectof palmitate and stearate on Panc1 and MIApaca2 cells(Supplementary Fig. S2). To better assess the growth-inhibitory effect of palmitate and stearate, cell growthcurve was generated by incubating Panc1 and MIApaca2cells in the media containing either BSA alone (as acontrol) or 0.25 mmol/L FFAs over a 8-day period (Fig.3B). Our data showed that palmitate and stearate sub-stantially inhibit the growth of pancreatic cancer cells.

Palmitate and stearate induce TRAIL expression andpromote apoptosis in pancreatic cancer cells

To determine whether palmitate and stearate could affectapoptosis, we examined the caspase-3 activity in Panc1 and

MIApaca2 cells following treatment of FFAs for 24 hours.Our data showed that palmitate and stearate (0.25mmol/L)significantly increased caspase-3/7 activity by 2- to 3- fold ascomparedwith controls in both cell lines (P < 0.01, Fig. 4A).To elucidate the potential underlying mechanisms of apo-ptosis regulation by palmitate and stearate, we then ana-lyzed the gene expression of apoptosis related genes inresponse to FFA treatment using qRT-PCR and found thatpalmitate and stearate significantly induced the expressionof the pro-apoptotic gene TRAIL by 2- to 3- fold in pancre-atic cancer cell lines as comparedwith controls (P<0.01, Fig.4B). Consistent with the gene expression, the protein levelof TRAIL was also increased by about 3- to 4- fold following24-hour incubation with 0.25 mmol/L palmitate or stea-rate. Taken together, these results show that palmitate andstearate induce TRAIL expression and trigger apoptosis inpancreatic cancer cells.

DiscussionAlteredmetabolism is considered as one of the hallmarks

of cancer (20). Genetic alterations enable cancer cells to

Figure 2. Gene expressions ofPNLIP, CLPS, PNLIPRP1, andPNLIPRP2 in two independentcohorts of PDAC. A, PNLIP, CLPS,PNLIPRP1, and PNLIPRP2 aredecreased in tumors as comparedwith adjacent non-tumor tissues intest and validation cohorts. Boxplots represent gene expressionlevel with relative intensity (log2) ofmicroarray data in the test cohort orrelative threshold cycle value (Ct)normalized with endogenouscontrol gene GAPDH using qRT-PCR data in validation cohort. Barsindicate median value. Student ttests P value and tumor: nontumorratios (T:N) are indicated in thegraphs. B, Kaplan–Meier analysisof PNLIP in test and validationcohorts. Gene expression of PNLIPis dichotomized into high and lowgroups using amedian cutoff. Log-rank P value is indicated in thegraphs.

Zhang et al.





reprogram metabolism to meet increased energy demandsfor cell proliferation and to survive in hypoxic and nutrient-deprived tumor microenvironment (21). In this regard, abetter understanding of metabolic dysregulation in pancre-atic cancer could lead to the discovery of novel therapeutictargets (22). Integrative postgenomic studies and systemsbiology approaches have emerged with the aim of devel-oping a more comprehensive understanding of cellularphysiology and metabolism (23, 24). To the best of ourknowledge, here, we report for the first time, the imple-mentation of a systems biology approach to investigategene-metabolite networks and metabolic dysregulation inpancreatic cancer, with integration of metabolomics andtranscriptomics.Metabolomics allow for global assessment of a cellular

state within the context of the immediate environment,taking into account genetic regulation, altered kineticactivity of enzymes, and changes in metabolic reactions(25). Thus, compared with transcriptomics or proteo-mics, metabolomics reflects changes in phenotype andtherefore cellular function (26). Metabolomics strategies

have been applied to tissues, serum, and other body fluid,to develop novel early diagnostic biomarkers in humancancers (27–30).

In this study, we identified 55 differentially expressedmetabolites using metabolic profiling in two independentcohorts of PDACcases.Next,we appliedWGCNA to analyzemetabolic networks in PDAC.WGCNA is a systemsbiology-based network analysis that has been shown to be animportant alternative and a more meaningful tool fordiscovery of molecular interaction networks and candidatebiomarkers (31–35). Using WGCNA, we have identified 8highly connected fatty acid hubs in a conserved lipid mod-ule, which are decreased in tumors as compared withadjacent non-tumor tissues in two independent cohorts ofPDAC. Further data mining revealed significant involve-ment of these fatty acids in cell proliferation and tumori-genesis (36). FFAs play an important role in numerousbiologic functions. They serve as a source of energy and asprecursors of many signaling and cellular components. Theeffect of different types of FFAs on cell proliferation andapoptotic activity in pancreatic cancer remains unclear. In

Table 2. Univariate Cox regression analysis on test and validation cohorts

Test cohort Validation cohort

Variables (comparison/referent) HR (95% CI) P HR (95% CI) P

PNLIP (high/low)�a 0.36 (0.15–0.88) 0.023 0.36 (0.15–0.87) 0.020CLPS (high/low)a 0.34 (0.14–0.84) 0.027 0.46 (0.20–1.09) 0.077PNLIPRP1 (high/low)a 0.37 (0.15–0.89) 0.028 0.56 (0.24–1.31) 0.182PNLIPRP2 (high/low)a 0.35 (0.14–0.85) 0.020 0.50 (0.20–1.21) 0.122Grading (G3&4/1&2) 1.94 (0.89–4.26) 0.097 1.11 (0.47–2.62) 0.819Resection Margin (R1/R0) 1.18 (0.54–2.58) 0.677 0.80 (0.32–1.96) 0.622Tumor stage (IIB-IV/I-IIA) 1.07 (0.54–2.14) 0.844 0.95 (0.39–2.32) 0.914

aGene expression value was dichotomized into high and low groups using median. P-value was calculated using univariate coxregression analysis.

Table 1. A set of coregulated fatty acids are identified using WGCNA on two independent cohorts

Test cohort Validation cohort

Metabolite ID Pa Ratiob IMconnc Pa Ratiob IMconnc

linolenate (18:3n3 or 6) 2.3E-04 0.53 6.91 1.9E-06 0.42 9.83palmitate (16:0) 4.2E-04 0.53 6.00 1.1E-04 0.48 6.60margarate (17:0) 1.0E-03 0.56 6.79 5.5E-07 0.40 13.24stearate (18:0) 1.5E-03 0.58 6.85 1.3E-06 0.41 13.50linoleate (18:2n6) 1.9E-03 0.58 6.79 4.6E-06 0.43 12.20oleate (18:1n9) 3.1E-03 0.59 7.86 4.3E-06 0.43 13.62eicosenoate (20:1n9 or 11) 3.9E-03 0.59 6.43 6.7E-05 0.47 12.2210-nonadecenoate (19:1n9) 4.1E-03 0.59 7.50 2.6E-05 0.45 12.12

aP-value calculated using t test in each cohort.bRatio of tumor vs. non-tumor.cIntramodular connectivity represents the strength of co-expression for each metabolite in network analysis.






this study, we showed for the first time that two majorsaturated FFAs, palmitate and stearate, exert a strong growthinhibitory effect in pancreatic cancer cells. Our data alsoshowed that palmitate and stearate significantly induceapoptosis in pancreatic cancer cells. These findings areconsistent with previous reports on induction of apoptosisby palmitate in other cell types, including breast cancer cells(37), hematopoietic cells (38), pancreatic b-cells (39), andcardiomyocytes (40). In contrast, many studies reportedcontradictory findings with respect to the role of unsatu-rated fatty acids in tumor growth, particularly for oleic acid,in breast cancer cells and tumor xenograft models (41). Ourfunctional investigation in different pancreatic cancer celllines showed that unsaturated FFAs, linoleate and oleate,have no significant effect on the proliferation of pancreaticcancer cells.

It has been proposed that excess palmitate could inducecell death through increased intracellular concentrationof ceramide (38, 39), a metabolite exclusively producedfrom saturated FFAs. However, other studies suggestedthat apoptosis induced by palmitate could occur throughthe generation of reactive oxygen species (ROS) ratherthan ceramide synthesis (40). FFA-induced production ofmitochondrial ROS is linked to the activation of proteinkinase C (PKC) and the redox-sensitive transcriptionfactor NF-kB (42), which might be involved in the reg-ulation of apoptosis (43). However, the biochemicalpathways by which, fatty acids influence pancreatic cancercell growth and death have not been adequately defined.Our data show that palmitate and stearate can upregulateTRAIL expression in pancreatic cancer cells, which maycontribute to the apoptosis induced by these two fattyacids.

To further define the genetic alterations and molecularpathways associated with this unique set of fatty acidsidentified by metabolic profiling, we integrated transcrip-tomics and metabolomics, and identified 157 gene surro-gates for the fatty acid set that is associated with PDAC.Pathway and network analysis revealed that the expectedlipid metabolism, particularly in lipolytic pathway involv-ing gene surrogates PNLIP, CLPS, PNLIPRP1, andPNLIPRP2, is significantly altered in PDAC (Fig. 1B andC). Pancreatic lipase, also known as pancreatic triacylgly-cerol lipase, is encoded by PNLIP, and secreted by thepancreas, and is the primary lipase that hydrolyzes lipids,converting triglyceride substrates to monoglycerides andFFAs (44). Unlike some pancreatic enzymes that are acti-vated by proteolytic cleavage, pancreatic lipase is secreted inits final form. Colipase, encoded by the CLPS gene, is aprotein coenzyme required for optimal enzyme activity ofpancreatic lipase (45). PNLIPRP1 and PNLIPRP2 code fortwo novel human pancreatic lipase-related proteins inpancreatic juice, referred to as PNLIP-related proteins 1 and2, each showing an amino acid sequence identity of 68% toPNLIP. PNLIPRP2 shows a lipolytic activity that is onlymarginally dependent on the presence of colipase, whereasthe function of PNLIPRP1 remains unclear (44). Overall,these functionally related lipases play key roles in directregulation of fatty acid turnover and signaling. Therefore,decreased expressions of lipases eventually lead to reducedlevels of FFAs. Consistent with the function of lipases, ourdata showed positive correlations between FFA levels andgene expression of these lipases in tumor samples, indicat-ing that a profound dysregulation of the lipolytic networkexists in PDAC and may play an important role in tumorgrowth (Fig. 1C).

Figure 3. Palmitate and Stearateinhibit pancreatic cancer cellproliferation. A, dose dependenceof palmitate and stearate oninhibition of cell proliferation inPanc1 and MIApaca2 cells. After16 h of starvation, cells are treatedwith BSA control or BSA-boundFFAs for 72 h. Saturated fattyacids, palmitate and stearate,show significant inhibition of cellproliferation at the concentration of0.031–0.5 mmol/L. Unsaturatedfatty acids linolate and oleatehave little or no effect on cellproliferation. B, there aresignificant decreases in cell growthof Panc1 and MIApaca2 treatedwith 0.25 mmol/L palmitateand stearate as compared withcontrol (CTRL) cells over a 8-dayperiod. Data are presented asmeans � S.D. from 3 independentexperiments. ��, t test P < 0.01,ANOVA P < 0.01.

Zhang et al.





Excessive production of pancreatic lipase may indicatethe presence of certain disorders, most notably inflamma-tion of the pancreas or pancreatitis (46). Elevated levels ofpancreatic lipases also occur in bowel obstruction or kidneydisease (47, 48).On the other hand, individualswithCrohndisease, cystic fibrosis, and celiac disease suffer from lipasedeficiency, in which the cells of the pancreas responsible forproducing this enzyme may be irreversibly damaged (49).The most common symptoms associated with lipase defi-ciency are muscle spasms, acne, arthritis, gallbladder stressand formation of gallstones, bladder problems, and cystitis.Therefore, pancreatic lipase supplements are used to treatPNLIP deficiency diseases (49). However, the role of PNLIPin pancreatic cancer remains unknown. In this study, wehave shown striking decreases (>100-fold) in the geneexpression of all four lipases including PNLIP in pancreatictumors, and consistently, immunohistochemical stainingon paraffin sections also showed that PNLIP protein level islower in PDACs as compared with normal ducts (Supple-mentary Fig. S3). In addition, our study also provided thefirst evidence that a lower expression of PNLIP is associatedwith poor outcome in PDAC.In summary, we have used a systems biology approach

and identified an altered lipolytic network involvinglipase genes and a unique set of fatty acids that areassociated with pancreatic cancer. This approach movesbeyond single gene investigation to provide a systems-

level perspective on the potential relationships amongmembers of a biologic network in pancreatic cancer. Ourresults suggest that the impaired lipolytic pathway maycontribute to the development and progression of PDAC.In conclusion, our study showed significant decrease infatty acids and their surrogate lipase genes in PDAC andshowed tumor inhibitory roles of palmitate and stearatein pancreatic cancer. Furthermore, to the best of ourknowledge, this is the first report indicating the potentialprognostic significance of PNLIP in pancreatic cancer.Further studies are needed to determine the mechanismunderlying the altered expression of these lipase genes,and to explore their potential clinical application inPDAC. These studies will lead to a better understandingof the aggressiveness of pancreatic cancer and may facil-itate therapeutic target discovery.

Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

Authors' ContributionsConception and design: G. Zhang, A. Maitra, S.P. HussainDevelopment of methodology: G. Zhang, P. He, T. RiedAcquisitionofdata (provided animals, acquired andmanagedpatients,provided facilities, etc.): G. Zhang, H. Tan, J. Gaedcke, B.M. Ghadimi,H.G. Yfantis, N. Hanna, H.R. AlexanderAnalysis and interpretation of data (e.g., statistical analysis, biosta-tistics, computational analysis): G. Zhang, P. He, A. Budhu, J. Gaedcke,B.M. Ghadimi, T. Ried, S. P. Hussain

Figure 4. Palmitate and stearateinduce TRAIL expression andpromote apoptosis. A, thereare significant increases incaspase-3/7 activity in Panc1and MIApaca2 cells with palmitateand stearate treatment. RelativeCaspase-3/7 activity representsthe effect of FFAs on apoptosiscompared with control cells after24-hour incubation. B, palmitateand stearate upregulate TRAILexpression. Cell lysates werecollected after 24-hour ofincubation with 0.25 mmol/L FFAsor BSA control. Real-time PCRwasconducted to determine TRAILmRNA levels. TRAIL protein level incell lysates wasmeasured using anELISA kit (R&DSystems) accordingto the manufacturer's instructions.Data represent means � S.D.from 3 independent experiments.��, t test P < 0.01, ANOVA P < 0.01.






Writing, review, and/or revision of themanuscript:G. Zhang, A. Budhu,J. Gaedcke, B.M. Ghadimi, T. Ried, D.H. Lee, A. Maitra, H.R. Alexander, S. P.HussainAdministrative, technical, or material support (i.e., reporting or orga-nizing data, constructing databases):G. Zhang, P. He, J. Gaedcke, T. Ried,N. HannaStudy supervision: S. P. Hussain

AcknowledgmentsThe authors thank Drs. Xinwei Wang and Stefan Ambs (National Cancer

Institute) for helpful discussions, Dr. Matthias Gaida for histopathologicevaluation of pancreatic tissue samples, and Ms. Elise Bowman for handlingof clinical samples and maintenance of the frozen tissue database. Theauthors also thank Drs. Jeff Pfohl and Ryan Michalek (Metabolon) for

helpful discussions and the personnel at UMMS for their contribution tothe collection of clinical samples under NCI-UMD contract.

Grant SupportThis work was supported by the Intramural Research Program of the

National Cancer Institute, Center for Cancer Research, NIH.The costs of publication of this article were defrayed in part by the

payment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely to indicatethis fact.

Received January 23, 2013; revised June 14, 2013; accepted July 14, 2013;published OnlineFirst August 5, 2013.

References1. Siegel R, Naishadham D, Jemal A. Cancer statistics, 2013. CA Cancer

J Clin 2013;63:11–30.2. ZhangG, Schetter A, HeP, FunamizuN,Gaedcke J,Ghadimi BM, et al.

DPEP1 inhibits tumor cell invasiveness, enhances chemosensitivityand predicts clinical outcome in pancreatic ductal adenocarcinoma.PLoS ONE 2012;7:e31507.

3. Yeh JJ. Prognostic signature for pancreatic cancer: are we close?Future Oncol 2009;5:313–21.

4. Griffin JL, Shockcor JP. Metabolic profiles of cancer cells. Nat RevCancer 2004;4:551–61.

5. Dang CV. Links between metabolism and cancer. Genes Dev 2012;26:877–90.

6. Wang F, HerringtonM, Larsson J, Permert J. The relationship betweendiabetes and pancreatic cancer. Mol Cancer 2003;2:4.

7. Hsu PP, Sabatini DM. Cancer cell metabolism: warburg and beyond.Cell 2008;134:703–7.

8. Dunn WB, Broadhurst D, Begley P, Zelena E, Francis-McIntyre S,Anderson N, et al. Procedures for large-scale metabolic profiling ofserum and plasma using gas chromatography and liquid chroma-tography coupled to mass spectrometry. Nat Protoc 2011;6:1060–83.

9. HiraiMY, YanoM,GoodenoweDB,KanayaS,Kimura T, AwazuharaM,et al. Integration of transcriptomics andmetabolomics for understand-ing of global responses to nutritional stresses in Arabidopsis thaliana.Proc Natl Acad Sci U S A 2004;101:10205–10.

10. Ferrara CT, Wang P, Neto EC, Stevens RD, Bain JR, Wenner BR, et al.Genetic networks of liver metabolism revealed by integration of met-abolic and transcriptional profiling. PLoS Genet 2008;4:e1000034.

11. Hamilton SR, Aaltonen LA. Pathology and genetics of tumours of thedigestive system. In:Hamilton SR, Aaltonen LA, editors. WHO Classi-fication of tumors. Lyon, France: IARC Press; 2000:204.

12. Evans AM, DeHaven CD, Barrett T, Mitchell M, Milgram E. Integrated,nontargeted ultrahigh performance liquid chromatography/electro-spray ionization tandem mass spectrometry platform for the identifi-cation and relative quantification of the small-molecule complement ofbiological systems. Anal Chem 2009;81:6656–67.

13. Budhu A, Roessler S, Zhao X, Yu Z, Forgues M, Ji J, et al. Integratedmetabolite and gene expression profiles identify lipid biomarkersassociated with progression of hepatocellular carcinoma and patientoutcomes. Gastroenterology 2013;144:1066–75.

14. Zhang B, Horvath S. A general framework for weighted gene co-expression network analysis. Stat Appl Genet Mol Biol 2005;4:Article17.

15. Langfelder P, ZhangB, Horvath S. Defining clusters from a hierarchicalcluster tree: the Dynamic Tree Cut package for R. Bioinformatics2008;24:719–20.

16. Connor SC, Hansen MK, Corner A, Smith RF, Ryan TE. Integration ofmetabolomics and transcriptomics data to aid biomarker discovery intype 2 diabetes. Mol Biosyst 2010;6:909–21.

17. Hardy S, Langelier Y, Prentki M. Oleate activates phosphatidylinositol3-kinase and promotes proliferation and reduces apoptosis of MDA-MB-231 breast cancer cells, whereas palmitate has opposite effects.Cancer Res 2000;60:6353–8.

18. Han JD, Bertin N, Hao T, Goldberg DS, Berriz GF, Zhang LV, et al.Evidence for dynamically organized modularity in the yeast protein-protein interaction network. Nature 2004;430:88–93.

19. Mead JF. Lipid Metabolism. Annu Rev Biochem 1963;32:241–68.20. Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation.

Cell 2011;144:646–74.21. Le A, Rajeshkumar NV, Maitra A, Dang CV. Conceptual framework for

cutting the pancreatic cancer fuel supply. Clin Cancer Res 2012;18:4285–90.

22. Svensson RU, Shaw RJ. Cancer metabolism: tumour friend or foe.Nature 2012;485:590–1.

23. Kitano H. Systems biology: a brief overview. Science 2002;295:1662–4.

24. Tomita M, Kami K. Cancer. Systems biology, metabolomics, andcancer metabolism. Science 2012;336:990–1.

25. Spratlin JL, Serkova NJ, Eckhardt SG. Clinical applications of meta-bolomics in oncology: a review. Clin Cancer Res 2009;15:431–40.

26. Fan TW, LaneAN,Higashi RM. The promise ofmetabolomics in cancermolecular therapeutics. Curr Opin Mol Ther 2004;6:584–92.

27. Claudino WM, Quattrone A, Biganzoli L, Pestrin M, Bertini I, Di Leo A.Metabolomics: available results, current research projects in breastcancer, and future applications. J Clin Oncol 2007;25:2840–6.

28. Wang H, Tso VK, Slupsky CM, Fedorak RN. Metabolomics anddetection of colorectal cancer in humans: a systematic review. FutureOncol 2010;6:1395–406.

29. Bathe OF, Shaykhutdinov R, Kopciuk K, Weljie AM, McKay A, Suther-land FR, et al. Feasibility of identifying pancreatic cancer based onserum metabolomics. Cancer Epidemiol Biomarkers Prev 2010;20:140–7.

30. Clyne M. Kidney cancer: metabolomics for targeted therapy. Nat RevUrol 2012;9:355.

31. Seligson DB, Horvath S, Shi T, Yu H, Tze S, Grunstein M, et al. Globalhistone modification patterns predict risk of prostate cancer recur-rence. Nature 2005;435:1262–6.

32. ChenY, Zhu J, LumPY, Yang X, PintoS,MacNeil DJ, et al. Variations inDNA elucidate molecular networks that cause disease. Nature2008;452:429–35.

33. Voineagu I, Wang X, Johnston P, Lowe JK, Tian Y, Horvath S, et al.Transcriptomic analysis of autistic brain reveals convergent molecularpathology. Nature 2011;474:380–4.

34. Horvath S, Zhang B, Carlson M, Lu KV, Zhu S, Felciano RM, et al.Analysis of oncogenic signaling networks in glioblastoma identifiesASPM as a molecular target. Proc Natl Acad Sci U S A 2006;103:17402–7.

35. Gargalovic PS, Imura M, Zhang B, Gharavi NM, Clark MJ, Pagnon J,et al. Identification of inflammatory gene modules based on variationsof human endothelial cell responses to oxidized lipids. Proc Natl AcadSci U S A 2006;103:12741–6.

36. Fritz V, Fajas L.Metabolism andproliferation share common regulatorypathways in cancer cells. Oncogene 2010;29:4369–77.

37. Hardy S, El-AssaadW, Przybytkowski E, Joly E, PrentkiM, Langelier Y.Saturated fatty acid-induced apoptosis inMDA-MB-231breast cancercells. A role for cardiolipin. J Biol Chem 2003;278:31861–70.

Zhang et al.





38. PaumenMB, IshidaY,MuramatsuM, YamamotoM,Honjo T. Inhibitionof carnitine palmitoyltransferase I augments sphingolipid synthesisand palmitate-induced apoptosis. J Biol Chem 1997;272:3324–9.

39. Shimabukuro M, Zhou YT, Levi M, Unger RH. Fatty acid-induced betacell apoptosis: a link between obesity and diabetes. ProcNatl AcadSciU S A 1998;95:2498–502.

40. Listenberger LL, Ory DS, Schaffer JE. Palmitate-induced apoptosiscan occur through a ceramide-independent pathway. J Biol Chem2001;276:14890–5.

41. Fay MP, Freedman LS, Clifford CK, Midthune DN. Effect of differenttypes and amounts of fat on the development of mammary tumors inrodents: a review. Cancer Res 1997;57:3979–88.

42. Evans JL, Goldfine ID, Maddux BA, Grodsky GM. Are oxidative stress-activated signaling pathwaysmediators of insulin resistance and beta-cell dysfunction? Diabetes 2003;52:1–8.

43. Romeo G, Liu WH, Asnaghi V, Kern TS, Lorenzi M. Activation ofnuclear factor-kappaB induced by diabetes and high glucose reg-ulates a proapoptotic program in retinal pericytes. Diabetes 2002;51:2241–8.

44. Reboul E, Berton A, Moussa M, Kreuzer C, Crenon I, Borel P. Pan-creatic lipase and pancreatic lipase-related protein 2, but not pancre-atic lipase-relatedprotein 1, hydrolyze retinyl palmitate in physiologicalconditions. Biochim Biophys Acta 2006;1761:4–10.

45. van Tilbeurgh H, Bezzine S, Cambillau C, Verger R, Carriere F. Coli-pase: structure and interaction with pancreatic lipase. Biochim Bio-phys Acta 1999;1441:173–84.

46. Yadav D, Ng B, Saul M, Kennard ED. Relationship of serum pancreaticenzyme testing trends with the diagnosis of acute pancreatitis. Pan-creas 2011;40:383–9.

47. Brooks S, Phelan MP, Chand B, Hatem S. Markedly elevated lipase asa clue to diagnosis of small bowel obstruction after gastric bypass. AmJ Emerg Med 2009;27:1167.

48. Ozkok A, Elcioglu OC, Cukadar T, Bakan A, Sasak G, Atilgan KG, et al.Low serum pancreatic enzyme levels predict mortality and are asso-ciated with malnutrition-inflammation-atherosclerosis syndrome inpatientswith chronic kidneydisease. IntUrol Nephrol 2012;45:477–84.

49. Dominguez-Munoz JE. Pancreatic enzyme therapy for pancreaticexocrine insufficiency. Curr Gastroenterol Rep 2007;9:116–22.






2013;19:4983-4993. Published OnlineFirst August 5, 2013.Clin Cancer Res Geng Zhang, Peijun He, Hanson Tan, et al. Pancreatic CancerAcid Network Exerting Growth Inhibitory Effects in Human Integration of Metabolomics and Transcriptomics Revealed a Fatty

Updated version

10.1158/1078-0432.CCR-13-0209doi:

Access the most recent version of this article at:

Material

Supplementary

http://clincancerres.aacrjournals.org/content/suppl/2013/08/02/1078-0432.CCR-13-0209.DC1

Access the most recent supplemental material at:

Cited articles

http://clincancerres.aacrjournals.org/content/19/18/4983.full#ref-list-1

This article cites 45 articles, 16 of which you can access for free at:

Citing articles

http://clincancerres.aacrjournals.org/content/19/18/4983.full#related-urls

This article has been cited by 16 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://clincancerres.aacrjournals.org/content/19/18/4983To request permission to re-use all or part of this article, use this link



http://clincancerres.aacrjournals.org/lookup/doi/10.1158/1078-0432.CCR-13-0209

http://clincancerres.aacrjournals.org/content/suppl/2013/08/02/1078-0432.CCR-13-0209.DC1

http://clincancerres.aacrjournals.org/content/19/18/4983.full#ref-list-1

http://clincancerres.aacrjournals.org/content/19/18/4983.full#related-urls

http://clincancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://clincancerres.aacrjournals.org/content/19/18/4983


integrationofmetabolomicsandtranscriptomicsrevealeda fatty … · maryland medical center...

Documents