Proteomic and meta-transcriptomic study on lymph node metastasis in gastric cancer

(1)

eupa open proteomics 3 (2014)183–194

Available

online

at

www.sciencedirect.com

ScienceDirect

j o u r n a l ho me p ag e :h t t p : / / w w w . e l s e v i e r . c o m / l o c a t e / e u p r o t

Proteomic

and

meta-transcriptomic

study

on

lymph

node

metastasis

in

gastric

cancer

Hiroshi

Ichikawa

a,b

,

Tatsuo

Kanda

c

,

Shin-ichi

Kosugi

b

,

Yasuyuki

Kawachi

d

,

Toshifumi

Wakai

b

,

Tadashi

Kondo

a,∗

a_Division_of_{Pharmacoproteomics,}_National_Cancer_Center_Research_Institute,_5-1-1_Tsukiji,_Chuo-ku,_Tokyo

104-0045,Japan

b_Division_of_Digestive_and_General_Surgery,_Niigata_University_Graduate_School_of_Medical_and_Dental_Sciences,

1-757,Asahimachidori,Chuo-ku,Niigata-shi,Niigata951-8510,Japan

c_Tsubame_Rosai_Hospital,_633,_Sado,_Tsubame-shi,_Niigata_959-1228,_Japan

d_Nagaoka_Chuo_General_Hospital,_2041,_{Kawasakimachi,}_Nagaoka-shi,_Niigata_940-8653,_Japan

a

r

t

i

c

l

e

i

n

f

o

Articlehistory:

Received28November2013 Receivedinrevisedform 21February2014 Accepted10March2014 Availableonline15March2014

Keywords:

Gastriccancer

Lymphnodemetastasis Label-freeproteomics Beta-3integrin Transcriptome

a

b

s

t

r

a

c

t

Toexaminetheproteomicbackgroundoflymphnodemetastasis(LNM)ingastriccancer, weperformedproteinexpressionprofilingofpairednon-tumor,primarytumor,andLNM tissues.Usingalabel-freeproteomicapproach,wegeneratedproteinexpressionprofilesof 3894uniqueproteinsandidentified109differentiallyexpressedproteins.Functional path-wayanalysisofthedifferentiallyexpressedproteinsshowedthatmembersofthebeta-3 integrin(ITGB3)pathwayweresignificantlyenriched.AberrationsofITGB3werereported invariousmalignancies;however,ITGB3inLNMtissueshasnotbeenexaminedtodate. DifferentlevelofITGB3expressionwasconfirmedin20gastriccancercasesbyWestern blotting.WeanalyzedthemRNAlevelsofthedifferentiallyexpressedproteinsbyusinga publicmRNAexpressiondatabase;38.8%oftheproteinsexamined,includingthoseinvolved inoxidationandreduction,showedcorrelationbetweenproteinandmRNAlevels.Proteins withoutsuchcorrelationincludedfactorsrelatedtocelladhesion.Ourstudysuggestsanovel rolefortheintegrinpathwayinthedevelopmentofLNMingastriccancerandindicated possiblebenefitsofobservationaltranscriptomicanalysisforproteomicstudies.

1.

Introduction

Gastriccancer isthe fourth mostcommon cancerand the second leading cause of cancer-related deaths worldwide [1]. Lymph node metastasis (LNM) is an early event of

∗ _{Corresponding}_author_at_:_Division_of_{Pharmacoproteomics,}_National_Cancer_Center_Research_Institute,_5-1-1_Tsukiji,_Chuo-ku,_Tokyo 104-0045,Japan.Tel.:+81335422511x3004;fax:+8135475298.

E-mailaddress:takondo@ncc.go.jp(T.Kondo).

the metastatic process andis mostcommonlyobservedin metastaticgastriccancer[2,3].TheLNMstatusisoneofthe most important prognostic factors for gastric cancer, and the 5-year survival rate for LNM-positive gastric cancer is 33.2%[4,5].Theseobservationssuggestthatthe understand-ingofthemolecularmechanismsofLNMcanleadtoanovel

http://dx.doi.org/10.1016/j.euprot.2014.03.001

(2)

prognosticindicatorortherapeuticapproachforthetreatment ofgastriccancer.

Metastasis of tumorcells consistsof multiple selection steps. Subpopulationsof cancercells withhigh metastatic potentialmaymigratefromtheprimarysitetodistanttissues, wherecells adaptivetoforeignmicroenvironmentsclonally proliferateinmetastaticsites[6].Toelucidatethemolecular mechanisms underlying these multistage events, genome-wide screening has been performed to identify genetic differences between the primarytumors and the LNM tis-suesingastriccancer[7,8].Thesestudieshavesuggestedthe existenceofcomplexgeneticabnormalitiesrelatedtogastric cancermetastasis.

Theproteomeisafunctionaltranslation ofthegenome. Thegenomic aberrations ofcancercells are transcribed to the transcriptome, translated tothe proteome, then deter-mining cancer phenotypes. In this sense, the proteome is a functional translation ofthe genome, directly regulating tumorbehavior.Proteomicstudiescangenerateuniquedata about the final productsof genome information. The pro-teomicstudyhasbeenemployedtoelucidatethemechanisms underlying LNM development in several types of cancers, andvariousintriguingfindingshavebeenreported[9]. There-fore, the investigations by the proteomic approach will be importantclues tounderstand the molecular mechanisms underlyingLNMingastriccancer.Theseapproacheswill fur-thertheunderstanding ofbiologicalmechanismsingastric cancerprogressionandwilleventuallybenefitcancerpatients. Inthisstudy,weaimedtoelucidatetheproteomic back-groundofLNMingastriccancer.Proteinexpressionprofiles consisting of 3894 unique proteins were generated using sodium dodecyl sulfate (SDS)-PAGE-based protein separa-tionfollowedbyLC–MS/MS.Usingalabel-freequantification method,weidentified109differentiallyexpressedproteinsin theLNMtissues.Functionalpathwayanalysisdemonstrated thatproteinsinvolvedinthebeta-3integrin(ITGB3)pathway weresignificantlyenrichedwithinLNM,andtheexpression patternofITGB3ingastriccancerprogressionandmetastasis wasexaminedbyWesternblotting.Theapplicationof meta-transcriptomicsrevealedthepossibletrendsofproteinswith concordantanddiscordantexpressionbetweenproteinsand mRNA.

2.

Material

and

methods

2.1. Clinicalsamples

Thisstudyincluded20patientswithgastriccancer.Matched and unmatched pairs of tumors, non-tumor and LNM tis-sueswereused;pairedtissuesfromtwocaseswereusedfor proteomicstudies, and all sampleswere usedfor Western blotting.ThepatientsunderwentsurgeryatNiigata Univer-sityMedicalandDentalHospitalandtheafﬁliatedinstitutions inNiigataprefecture(NagaokaChuoGeneralHospital, Nan-bugoGeneralHospital,NagaokaRedCrossHospital,Nippon DentalUniversityMedicalHospital,NiigataCityGeneral Hos-pital,TachikawaGeneralHospital,ShibataHospital,Saiseikai NiigataDainiHospital,SakamachiHospital,Kashiwazaki Gen-eral Hospital and Medical Center). All patients underwent

surgicalresectionin2010andhadnohistoryofcancer treat-ments involving systemic therapy. At the time of surgery, tissue fragments were grossly resected from the primary tumors,andthematchednon-tumormucosatissueslocated atleast5cmawayfrom thetumormargins werecollected. Allthetissueswereimmediatelysnap-frozeninliquid nitro-gen and storedat −80◦C untiluse. TheLNM tissues were cutinto twopieces; onewasembeddedinOCT compound (Sakura Finetechnical, Tokyo, Japan) for histological iden-tification of tumor metastasis, and another was used for proteomicstudyandWesternblotting.Bystainingfrozen sec-tionswithhematoxylin–eosin,weconfirmedthatthelymph node metastatic tissueshad at least50% tumorcells.The tumorswereclassifiedonthebasisofthehistologicalstage, accordingtotheInternationalUnionagainstCancer tumor-node-metastasis (TNM) classification, 7th edition [10]. The clinicopathologicaldatafor20casesarepresentedinTable1. ThisstudywasapprovedbytheEthicalReviewBoardofthe NiigataUniversityFacultyofMedicine,theaffiliated institu-tionsofNiigataUniversityMedicalandDentalHospital,and NationalCancerCenter.Informedconsentwasobtainedfrom allthepatientsateachinstitution.

2.2. Proteinexpressionproﬁling

Frozensampleswere crushed topowder inliquidnitrogen witha Multi-beadsshocker(YasuiKikai,Osaka, Japan)and treated withlysis buffercontaining 6M urea,2M thiourea, 3% CHAPS, and 1% Triton X-100. After the mixtures were centrifugedat15,000rpmfor30min,thesupernatantswere recoveredandusedinsubsequentproteinexpressionstudies. Protein expression profiles were generated by the LC–MS/MSmethod.Sixtymicrogramsofeachproteinsample wasseparatedonaready-made12.5%SDS-PAGEgel(ATTO, Tokyo,Japan).Eachgellaneswerecutinto48piecesofequal size by GridCutter (Gel Company, San Francisco, CA) and subjectedtoin-geltrypticdigestionasdescribedpreviously [11].Trypsindigestsweresubjectedtoliquidchromatography coupledwithanalysiswithaLTQ-OrbitrapXLmass spectrom-eter(ThermoFisherScientific,SanJose,CA).AcquiredMSand MS/MS spectra (Thermoraw files) were analyzed with the ProgenesisLC-MSsoftware,version3.4(Nonlinear,Dynamics, Newcastle, UK).Profile dataofMS scanswere transformed to peak lists with respective peak m/z values, intensities, abundances(areasunderthepeaks),andm/zwidth.MS/MS spectraweretreatedsimilarly.

After selecting one sample as the reference, the reten-tion timesofall theothersamplesusedintheexperiment were alignedbymanuallycreatingthreetofivelandmarks, followedbyautomaticalignmentofallretentiontimesto max-imaloverlayofthe2Dfeaturemaps.Featureswithonlytwo or threecharges wereincluded infurtheranalyses.MS/MS spectrawereexportedfromtheProgenesissoftwareas Mas-cotgenericfiles(mgf)andwereusedforpeptideidentification bysearchingtheSWISS-PROTdatabase(Homosapiens,471,472 sequencesintheSprot57.5.fastafile)byusingtheMascot soft-ware(version2.2;MatrixScience,London,UK).Thefollowing searchparameterswereused:toleranceoftwomissedtrypsin cleavages,variablemodificationsonthemethionineresidue (oxidation,+16Da),maximumprecursorionmasstolerance

(3)

e u p a o p e n p r o t e o m i c s 3 ( 2 0 1 4 ) 183–194

185

Table1–Clinicalandpathologicaldataofthegastriccancerpatientswhoparticipatedinthisstudy. Case numbera Patient age Patient sex Histological differentiation

Tstatusb _N_statusb _M_statusb _Number_of

metastaticnodes Lymphatic invasion Vascular invasion TNM stageb

1 76 Male WD T4 N3 M1 9 Presence Presence IV

2 65 Male MD T4 N1 M1 2 Presence Presence IV

3 81 Male WD T1 N0 M0 0 Absence Absence I

5 70 Male MD T2 N0 M0 0 Absence Absence I

6 63 Male MD T2 N0 M0 0 Absence Absence I

8 72 Male MD T3 N0 M0 0 Absence Absence II

9 63 Male MD T3 N0 M0 0 Absence Absence II

10 72 Male MD T2 N1 M0 1 Presence Absence II

11 81 Male MD T3 N1 M0 1 Presence Absence II

12 76 Male WD T3 N1 M0 1 Presence Presence II

13 64 Male MD T2 N1 M0 1 Presence Presence II

14 56 Male MD T3 N3 M0 18 Presence Presence III

16 55 Male MD T3 N2 M0 3 Absence Absence III

18 75 Male MD T3 N2 M0 3 Absence Presence III

19 72 Male MD T4 N3 M1 11 Presence Presence IV

20 70 Male WD T4 N3 M1 12 Presence Presence IV

WD,welldifferentiated;MD,moderatelydifferentiated.

a _The_samples_in_case₁_and₂_were_used_to_create_the_protein_expression_proﬁles,_and_those_in_case₁_to20_were_used_for_Western_blotting_analysis.

(4)

of±10ppm,andfragmentionmasstoleranceof±0.8Da. Pro-teinswithaMascotscoreof34ormorewereconsideredas positivelyidentiﬁed.Peptidesearchresultswerere-imported totheProgenesissoftware.

For protein quantification, only nonconflicting peptides (i.e.,peptides occurringinonlyoneprotein) were selected. Aftersumminguptheabundancesofallthepeptidesallocated toeachprotein,theresultsof48fractionswerecombinedin atotalanalysisset.Allthe experimentswere performedin duplicates,andeachnumberrepresentedanaverageoftwo replicates. Theprocessed rawdata on protein abundances werenormalizedbyfixingthemeanintensityofeach sam-pleatoneandwerepresentedasascatterplot;theratiosof meanabundancesintwocasesbetweentissuetypeswere cal-culatedusingtheExpressionistanalystsoftware(GeneData, Basel,Switzerland). Wedefinedthe different expressionas morethan2.0-folddifferenceofthemeanproteinabundance betweentissuetypes.Proteinsthatshowedhigherexpression intumorthannon-tumortissues,andalsohigherexpression inLNMthaninthe tumortissues,orproteinsthatshowed lowerexpressionintumorsthaninthe non-tumortissues, andevenlowerexpressionintheLNMtissueswereselected ascandidateproteinsassociatedwithLNM.

2.3. Pathwayanalysis

Pathway analysis of differentially expressed proteins was performedusingtheCytoscapesoftware[12]withthe Reac-tomefunctionalinteraction(FI)plug-in[13].TheReactomeFI datasetunitesinteractionsfromReactomeandthosederived from the pathwaydatabases,including KEGG,BioCye, Pan-ther, TheCancer Cell Map (http://cancer.cellmap.org/), and NCI-PID, withpair-wise interactions gleaned from physical protein-proteininteractionsinhumanandmodelorganisms, genecoexpressiondata,proteindomain–domaininteractions, protein interactions generated from text mining, and GO annotations[13].Toinvestigatethefunctionsofthenetwork created,pathwayenrichmentanalysiswasperformedusing theReactomeFIplug-in.Pathwayswithafalsediscoveryrate (FDR)<0.05wereconsideredtobesigniﬁcantlyenriched.

2.4. Westernblotting

Protein samples were separated by SDS-PAGE and blotted onto aPVDF membrane. Primary antibodies against ITGB3 (1:1000;BDBioscience,SanJose,CA)andsecondary antibod-ies against mouse IgG (1:2000, GE Healthcare Biosciences, Uppsala,Sweden) were used, and immunocomplexes were detectedusinganenhancedchemiluminescencesystem(ECL Prime Western Blotting DetectionReagents; GE Healthcare Biosciences)and LAS-3000 (Fujiﬁlm, Tokyo, Japan). To nor-malizetheamountofproteinloadedineachlane,thesame membraneswere stained with PonceauS. Theintensity of WesternBlotsignalsandthedensityofthetotallanesstained withPonceauSweremeasuredusingImageQuant software (GEHealthcareBiosciences),andtherelativeintensitywas cal-culatedbydividingtheintensityofWesternBlotsignalsbythe opticaldensityofthetotallanes.

2.5. Geneexpressionanalysis

The microarrayand clinical and pathological dataset of89 gastriccancercaseswasobtainedfromtheGeneExpression Omnibus(GEO)database(GSE4007).Thisdatasetconsistsof atotalof44,500probescorrespondingtoabout30,300unique genes.Geneexpressiondataof14casesthatincluded14 pri-mary tumor tissues, 14 corresponding LNM tissues, and 5 matchednon-tumor mucosae, were selectedforthe analy-sis.TocompareproteinandmRNAexpression,weselected probescorrespondingtothedifferentiallyexpressed109genes identiﬁedinproteomicanalysis.Wecouldobtainthe expres-siondataof67genes.Weselectedaprobewiththehighest expressionvaluewhenmultipleonesexistfortheidentical gene.Additionaldataprocessingsuchasnormalizationwas notperformed. Geneexpressiondatawere importedtothe Expressionistanalystsoftware(GeneData,Basel,Switzerland), and mean expressionratios between differenttissuetypes werecalculated.Wedeﬁnedthedifferentexpressionasmore than1.0folddifferenceofthemeanexpressionvaluebetween thetissuetypes.

2.6. Geneontologyenrichmentanalysis

Geneontology(GO)enrichmentanalysiswasperformedusing DAVID,version6.7[14].Datarelatedtogeneswithconcordant anddiscordantmRNAandproteinexpressionwereseparately uploadedtoDAVID,andthebackgroundwasdefinedas“Homo sapiens”. Functional annotation charts were created using “GOTERMBPFAT”(collectionofbroadest GOtermscurated from GOannotationsdataset).Thresholdswerechangedto a gene count of two and an EASE score of one (modified FisherexactP-value).TheP-valueforeachGOtermreflects enrichmentinthefrequencyofthatGOtermintheuploaded gene list, relative to all the genes in the background list. The GO terms were filtered to include enrichments with Benjamini–HochbergcorrectedP-valueslessthan0.20. 2.7. Statisticalanalysis

TheKruskal–Wallistest wasperformedtocompare contin-uous variables in multiple groups by using the SPSS 11.5 statistical package(SPSS,Chicago,IL),and differenceswith

P<0.05wereconsideredstatisticallysigniﬁcant.

3.

Results

AnoverviewoftheexperimentworkflowisshowninFig.1. First,wegeneratedproteinexpressionprofilesofpaired non-tumor, tumor, and LNMtissues fromcases 1and 2(Fig. 1, Table1).EachproteinsamplewasseparatedbySDS-PAGE,and thegellaneswerecuttoobtain48pieces.Trypsindigestswere extracted fromeach gelpiece,and subjectedtoLC–MS/MS. Weidentifiedandquantified109,949uniquepeptides corre-spondingto3894nonredundantproteinsfromsixsamplesin technicalduplicatesbytheProgenesisLC-MSsoftware (Sup-plementaryTables1and2).Tomonitorthereproducibilityof thewholeprocess(SDS-PAGEseparation,in-geltryptic diges-tions,LC–MS/MS,andproteinquantification),wegenerateda

(5)

187

Fig.1–Experimentalworkflowofalabel-freeproteomics approachwithpathwayenrichmentandmRNAexpression meta-analysis.Non-tumor(NT),tumor(T)andlymphnode metastasis(LNM)tissueswereexamined.Thelabel-free quantificationmethodquantifiedtheexpressionofatotal of3894proteins;109differentiallyexpressedproteinswere furtheranalyzedforfunctionalinteractionsandbiological enrichedpathways.Thebeta-3integrin(ITGB3)pathway wasthemostenrichedpathwayinthefunctional interactionnetworkcreated.Theassociationbetween ITGB3proteinexpressionandgastriccancerprogression wasconfirmedbyWesternblotting.Amongthe109genes, 67weregroupedaccordingtosimilaritybetweenthe proteinandmRNAexpressionpatterns,suchas

“Concordantgenes”or“Discordantgenes”.Geneontology analysisshowedtheenrichedfunctionalcharacteristicsin theeachgenegroups.

scatterplotoftechnicalduplicatesfromidenticalsamplesof theprimarytumortissuefromcase2(Table1)andcalculated thePearson’scorrelationcoefﬁcient.Thescatterplotshowed thatthenormalizedabundanceof83.7%and94.5%ofthe pro-teinsrangedwithin2.0–4.0-folddifference,respectively,and

thatthecorrelationcoefﬁcientwas0.92(SupplementaryFig. 1).Theseresultsdemonstratedhighreproducibilityof label-freeproteinexpressionproﬁling.

SupplementaryTables1and2andFig.1canbefound,in theonlineversion,atdoi:10.1016/j.euprot.2014.03.001.

We compare protein expression profiles between the non-tumor,tumor,andLNMtissuesintwocases(Case1and 2, Table 1). Among the 3894 proteins, 55 proteins showed significantly (>2-fold ratio of means in two cases) higher expressionintumorthannon-tumortissuesandalso signif-icantlyhigherexpressioninLNMthan inthetumortissues (Fig.2).Incontrast,54proteinsshowedsignificantly(<2-fold ratio of means in two cases) lower expression in tumors thaninthenon-tumortissues,andevensignificantlylower expressioninthe LNMtissues(Fig.3).Weselectedthe 109 differentiallyexpressedproteinsascandidateproteins asso-ciatedwithLNM.Detailsoftheseproteinsare presentedin SupplementaryTable3.

SupplementaryTable3canbefound,intheonlineversion, atdoi:10.1016/j.euprot.2014.03.001.

Westudiedthefunctional interactionsofthese109 pro-teinsbyusingCytoscapewiththeReactomeFIplug-in.We foundthat 11(10.1%)ofthe109 proteinswere functionally linked toeach other and thatthe ITGB3cell surface inter-actionspathwaywasthemostsignificantlyenrichedinthe identifiednetworkofthe11proteins(FDR<1.00E-4,Fig.4Aand Table2).ITGB3wasidentifiedandexaminedinourproteomic analysis. ITGB3protein expressioninthe LNM tissueswas lower thanthatofthetumortissues(ratioofmeans=3.96), and that ofthe tumortissueswashigher than that ofthe non-tumor tissues(ratio ofmeans=1.76) (Fig.4B).We vali-datedtherelationshipbetweenITGB3expressionandgastric cancer progression in 20 cases byWestern blotting. ITGB3 expressionwas foundto significantlyincrease withcancer stageprogressionbutdecreaseinLNMtissues(Fig.4C).These observationsconfirmedthe resultsofthe label-freeprotein expressionprofiling.

We analyzedthe mRNAlevels for the 109 differentially expressed proteins by using a public mRNA expression database.WeselectedtheGSE4007datasetintheNCBIGene ExpressionOmnibus(GEO) becauseit wasthe onlydataset that included mRNAexpressiondata obtainedfrom paired non-tumortissues,primarytumortissues,andLNMtissues, alongwithsufﬁcientclinicalinformation.Fromthe109genes, themRNAexpressiondatafor67geneswereincludedinthe GSE4007 dataset.Clinical andpathological informationand dataonmRNAexpressionhavebeenpresentedin Supplemen-taryTables4and5,respectively.

SupplementaryTables4and5canbefound,intheonline version,atdoi:10.1016/j.euprot.2014.03.001.

To characterize the genes according to the correlation between mRNA and proteinexpression, wedivided the 67 genesintotwogroups:26genes(38.8%)thatshowedthesame protein and mRNA expression patterns were grouped into “Concordantgenes”(Table3),whereas41genesthatshowed differentpatternsweregroupedinto“Discordantgenes”.We analyzedtheenrichedGOtermsineachgroupbyusingDAVID, version6.7.GOtermsrelatedtooxidation,reduction,or hor-monemetabolismwereenrichedinthegroupof“Concordant genes”,whilethoseinvolvedinphagocytosisorcelladhesion

(6)

Fig.2–Fifty-fivesignificantlyupregulatedproteinsidentifiedbyalabel-freequantificationmethod.Theproteinabundance ineachcasewasz-transformedacrossallsixsamplesandhasbeenpresentedinaheatmapformat.Tissuetypesare markedasnon-tumor(NT),tumor(T),andlymphnodemetastasis(LNM).CasenumberscorrespondtothoseinTable1and Fig.4.RefertoSupplementaryTable3forthegenedescriptionandvaluesoftheratioofmeans.

wereover-representedinthe“Discordantgenes”group(Fig.5 andSupplementaryTable6).

SupplementaryTable6canbefound,intheonlineversion, atdoi:10.1016/j.euprot.2014.03.001.

4.

Discussion

LNMisoneofthemostimportantprognosticfactorsin gas-triccancer,andinvestigationofthemolecularbackgroundof LNMformationmayleadtonoveltherapeuticmodalities[4]. Metastasishasbeenassumedtooccurthroughclonalgenomic andepigeneticevolution[5].Molecularmechanismsrelatedto tumormetastasisareupregulatedinmetastatictumorcells comparedto primary tumors,and exploring differences in

the molecularbackgroundofprimarytumorsandLNM tis-suesmaybethemostdirectandcrediblewaytoelucidatethe molecularmechanismsunderlyingthemetastasisprocess.

We employed SDS-PAGE-based size fractionation before massspectrometryanalysis.Pre-fractionationofproteinsby SDS-PAGE hassigniﬁcant advantagesover other separation techniques in that it detects a greater number ofproteins andiswidelyusedtocharacterizeproteincomplexesin can-cerproteomicstudies[15–17].Gel-basedsizefractionationfor theidentiﬁcationofproteinsissuperiortoseparationby liq-uidchromatographyasapre-fractionationmethod[18].Jafari et al. compared gel-based protein-separation techniques, including SDS-PAGE, isoelectric focusing with immobilized pH gradientgelstrips (IEF-IPG),and two-dimensionalPAGE (2D-PAGE), on the basis of their ability to serve as a

(7)

189

Fig.3–Fifty-foursignificantlydownregulatedproteinsidentifiedbyalabel-freequantificationmethod.Theprotein abundanceofeachgenewasz-transformedacrossall6samplesandhasbeenpresentedinaheatmapformat.Thetissue typesaremarkedasshowninFig.2.CasenumberscorrespondtothoseinTable1andFig.4.RefertoSupplementaryTable 3forthegenedescriptionandvaluesoftheratioofmeans.

fractionationtechniqueforMSanalysisofacomplexprotein sample[19].TheyreportedthatSDS-PAGEyieldedthehighest numberofidentiﬁableproteins.

In addition, we quantified the identified proteins by a label-free method with the Progenesis LC-MS software. Label-freequantificationiswidelyusedbecauseitisasimple techniquethatallowssimultaneousidentificationand quan-tificationofproteins.Alternatively,isotopiclabelingmethods suchasICAT[20],iTRAQ[21],andSILAC[22]arealsousedfor MS-basedproteinquantification;theseisotopiclabeling meth-odsallowforthecomparisonofmultiplesamplesinasingle LC–MS/MSrun,therebyprovidingincreasedaccuracy, preci-sion,andreproducibility[23].However,suchexperimentsare complicatedbecauseoftheadditionallabelingreactions,and SILACisavailableonlyintissueculturesystems.Alabel-free

methodbasedonProgenesisLC-MScanquantifyand iden-tify thousands of proteins from complex samples with a simpletechniqueandhighreproducibility.ProgenesisLC-MS alignstheionchromatogramstocompensateforvariationsin retentiontimesamongmultiplesamplesbeforequantification based on ion intensities. Ithas been reported that analy-sisofretentiontimealignmentshowssimilarquantification precisionandreproducibilitybuthigheridentification capac-itythanisotopelabelingmethodssuchasSILAC[24].Inthis study,weidentifiedandquantified3894non-redundant pro-teinsfromcomplexclinicalsampleswithhighreproducibility (SupplementaryFig.1)andthenemployedalabel-free quan-tificationmethodfortheproteomicstudies.

Thedevelopmentoftumorinvasionand metastasisisa very complicatedandcontinuous processwithmultisteps.

(8)

Fig.4–Beta-3integrin(ITGB3)pathwayandassociationof theITGB3proteinexpressionwithprogressionofgastric cancer.CasenumberscorrespondedtothoseinTable1. FunctionalinteractionnetworkanalysisbyCytoscapewith theReactomeFIplug-inisdemonstrated(A).TheITGB3 proteinexpressionlevelquantiﬁedbyalabel-freemethod (B).ITGB3proteinexpressionlevelwasz-transformed acrossallsixsamplesexaminedandwasrepresentedasa heat-mapformatandlinechart.ITGB3proteinexpression in20gastriccancercases,evaluatedbyWesternblotting(C).

Investigationofthemolecularbackgroundoftumor metas-tasis through“omics” studiesrevealed that multiplegenes aberrationswerecontributed tothe tumormetastasis [7,8]. Therefore, we tried toexamine the overall features ofthe expressedproteins,andidentified109aberrantly expressed proteins.Next,weidentifiedtheITGB3pathwayasthemost enrichedproteininteractionnetworkthoroughthepathway enrichmentanalysisinthe109proteins(Fig.4A).Inaddition, wemonitored the ITGB3protein expressionusing the cre-atedproteinexpressionprofile.Integrins are heterodimeric cellsurfaceglycoproteinswithalphaandbetasubunits.ITGB3 expressionismainlyassociatedwiththemigrationand inva-sionoftumorcells[25],andoverexpressionofITGB3isrelated tothemetastatic potentialofmelanoma[26],breastcancer [27], colorectal cancer [28], and bone metastasis in breast cancer[29]. However,in ovariancancer, ITGB3reducesthe

metastaticpotentialoftumorcells[30–32].Inhepatocellular carcinomacells,ITGB3hasbeenshowntohaveaproapoptotic function,and downregulationofITGB3isrelatedto aggres-sivetumorgrowth [33].Therefore, theassociation ofITGB3 expressionwithtumormetastasisandprogressiondepends onthecancertype.Inthecurrentstudy,ITGB3expressionwas increasedascancerstageadvanced,therebylinkingitto gas-triccancerprogression.Theseobservationsaresupportedby previousfindings[34].However,downregulationofITGB3in theLNMtissueshasnotbeenreporteduntilourstudy. Neo-plasticinvadingcellsmustovercometheintegrin-mediated death (IMD) induced by the interaction of ITGB3 with the extracellular matrixofhosttissuestoestablish metastases [25].Whenatumorcellmigratesthroughamicroenvironment wheretheextracellularmatrix(ECM)doesnotcontaina suit-able ligandforITGB3,the integrin cytoplasmictailrecruits caspase-8to thecell membrane,and apoptosis isinduced. Stupacketal.[35]showedthatITGB3-expressingmelanoma cellsundergoapoptosis incollagengelsthatlackITGB3 lig-ands.ITGB3specificallybindsawiderangeofECMmolecules, includingfibronectin,fibrinogen,vonWillebrandfactor,and vitronectin[36].Thepresentproteomicanalysisdemonstrated thatseveralcollagen-relatedgeneswereupregulatedinLNM tissues,whilefibronectin,fibrinogen,andvitronectinwerenot identifiedasupregulatedproteinsinLNMtissues(Fig.2and SupplementaryTable3).Therefore,ITGB3downregulationin gastriccancercellsthatshowedLNMmightreflectasurvival mechanism related tothe preventionofIMD inmetastatic tissues.

The problems of the proteomic biomarker studies are lack of the validation studies in the independent clinical samplesfromthemulti-institutions.Selectionofbiomarker candidates from the differentially expressed proteins and acquisition of the independent clinical samples are criti-calstepsinmulti-institutionalbiomarkervalidationstudies. Transcriptomedata are deposited inpublicdatabases with clinicopathologicaldataofthesamples,andarefreely avail-ableforindependentvalidationstudies;asclinicalmaterials aregenerallynotveryaccessible,suchinfrastructureishighly desirableforproteomicbiomarkerstudies.However,presently, thereisnoanalogousproteomicdatabase.Forproteinswhose levels demonstrate concordance with mRNA expression, a transcriptomedatabaseshouldbeausefultoolforbiomarker selection and validation. Several studies have compared mRNAandproteinlevelsintissueculturecells.Chenetal.[37] comparedmRNAandproteinexpressionbetweentwogastric cancercelllinesandestimatedanoverallcorrelation coefﬁ-cientof0.29.PreviousstudiescomparingmRNAandprotein expressioninhumanandmousecelllinesconcludedthatthe mRNAlevelsexplainapproximately40%ofthevariabilityin proteinlevels[38].Althoughthesestudiessuggestedpossible beneﬁtsofusingtranscriptomedatabasesforvalidation stud-ies,thecorrelationbetweenproteinandmRNAexpressionin surgicallyresectedtissuesremainunclear.Weanalyzedthe mRNAlevelsforthe109differentiallyexpressedproteinsby usingapublicmRNAexpressiondatabase.Ofthese109 pro-teins,67werelistedinthegastriccancerdatasetoftheGEO database. We foundthat 26 of 67 genes (38.8%) showed a concordantexpressionpatternforproteinandmRNA.These concordant genes were enriched in GO terms related to

(9)

191

Table2–Pathwayenrichmentanalysisfortheidentiﬁedproteinnetworkconsistingof11proteins.

Pathwaya Sourceb FDRc Numberof

proteinsd Includedproteins

e

Beta3integrincellsurfaceinteractions NCI-PID 1.00E−04 3 COL1A2,COL1A1,THBS1

Amoebiasis KEGG 1.11E−04 4 COL1A2,COL2A1,COL1A1,COL5A2

Integrincellsurfaceinteractions Reactome 1.25E−04 4 COL1A2,COL2A1,COL1A1,THBS1

Proteindigestionandabsorption KEGG 1.43E−04 4 COL1A2,COL2A1,COL1A1,COL5A2

SignalingbyPDGF Reactome 1.67E−04 5 COL1A2,COL2A1,COL1A1,THBS1,COL5A2

Integrinsignalingpathway Panther 1.82E−04 4 COL1A2,COL2A1,COL1A1,COL5A2

Focaladhesion KEGG 2.00E−04 5 COL1A2,COL2A1,COL1A1,THBS1,COL5A2

Plateletadhesiontoexposedcollagen Reactome 2.50E−04 3 COL1A2,FCER1G,COL1A1

ECM-receptorinteraction KEGG 3.33E−04 5 COL1A2,COL2A1,COL1A1,THBS1,COL5A2

Beta1integrincellsurfaceinteractions NCI-PID 5.00E−04 5 COL1A2,COL2A1,COL1A1,THBS1,COL5A2

Extracellularmatrixorganization Reactome 1.00E−03 7 PLOD2,MMP9,COL22A1,COL1A2,COL2A1,

COL1A1,COL5A2

Cellsurfaceinteractionsatthevascularwall Reactome 1.31E−03 3 COL1A2,FCER1G,COL1A1

Axonguidance Reactome 1.42E−03 4 COL1A2,COL2A1,COL1A1,COL5A2

Plateletactivation,signalingandaggregation Reactome 1.93E−03 4 COL1A2,FCER1G,COL1A1,THBS1

VEGFR3signalinginlymphaticendothelium NCI-PID 4.00E−03 2 COL1A2,COL1A1

Syndecan-4-mediatedsignalingevents NCI-PID 6.06E−03 2 MMP9,THBS1

ValidatedtranscriptionaltargetsofAP1family

membersFra1andFra2

NCI-PID 7.76E−03 2 MMP9,COL1A2

Bladdercancer KEGG 9.39E−03 2 MMP9,THBS1

a _Pathway_enrichment_analysis_was_performed_using_Cytoscape_with_Reactome_FI_plug-in_[_{http://wiki.reactome.org/index.php/Reactome}_FI

CytoscapePlugin].

b _Reactome_FI_data_set_unites_interactions_from_Reactome_and_those_derived_from_other_pathway_databases,_including_KEGG,_NCI-PID,_BioCyc,

PantherandTheCancerCellMap.

c _FDR_was_calculated_by_Reactome_FI_plug-in,_and_the_signiﬁcant_threshold_was_set_at_less_than_0.05.

d _The_numbers_of_proteins_which_were_examined_in_this_study_involved_in_the_pathway.

e _Gene_names_which_were_identiﬁed_to_be_{signiﬁcantly}_up-_or_{down-regulated}_by_proteomic_experiment_were_shown.

Fig.5–FunctionalcharacteristicsofgeneswithconcordantanddiscordantproteinandmRNAexpression.Geneswere groupedaccordingtotheirproteinandmRNAexpressionpatternsandanalyzedforenrichedgeneontologytermsby DAVID,version6.7(http://david.abcc.ncifcrf.gov).GOtermswithenrichmentofP<0.20werelisted,andP-valueswere z-transformedandhavebeenpresentedinaheatmapformat.

(10)

e u p a o p e n p r o t e o m i c s 3 ( 2 0 1 4 ) 183–194

Table3–Listof26geneswithcorrelationbetweenproteinandmRNAexpression.

Genesymbola _Gene_IDb _Gene_description _Protein _mRNA

Accession numberc Ratio (T/NT)d Ratio (LNM/T)e ProbeIDf _Accession numberg Ratio (T/NT)d Ratio (LNM/T)e Up-regulatedgenesatbothproteinandmRNAlevel

MMP9 4318 Matrixmetalloproteinase-9 P14780 3.46 2.63 10805 T64837 2.56 1.03

AGMAT 79814 Agmatinase,mitochondrial Q9BSE5 2.91 2.51 15950 AA934764 1.11 1.24

CTSH 1512 CathepsinH P09668 2.78 3.33 953 AA487231 1.04 1.05

TGOLN2 10618 Trans-Golginetworkintegralmembraneprotein2 O43493 2.84 3.73 15417 T81338 1.21 1.05

SLC26A2 1836 Sulfatetransporter P50443 2.42 2.07 10913 W15263 1.13 1.02

XRN2 22803 5_–3_{exoribonuclease}₂ _Q9H0D6 _13.38 _2.36 ₈₁₆₂ _AA028164 _1.41 _1.13

FAM103A1 83640 ProteinFAM103A1 Q9BTL3 2.09 2.79 12871 AA432100 1.14 1.01

TAPBP 6892 Tapasin O15533 102.06 726.07 18291 AA704775 2.20 1.09

TCL1A 8115 T-cellleukemia/lymphomaprotein1A P56279 88.57 3.65 12786 R97095 1.07 2.62

GDAP1 54332 Ganglioside-induceddifferentiation-associatedprotein1 Q8TB36 3.99 3.09 17310 H15302 1.45 1.06

NCF2 4688 Neutrophilcytosolfactor2 P19878 4.97 6.98 6201 AA872098 3.26 1.42

CHD4 1108 Chromodomain-helicase-DNA-bindingprotein4 Q14839 2.07 2.49 5588 N34372 1.16 1.20

Down-regulatedgenesatbothproteinandmRNAlevel

CA2 760 Carbonicanhydrase2 P00918 4.61E−01 2.99E−01 15331 H23187 1.21E−01 9.92E−01

GKN2 200504 Gastrokine-2 Q86XP6 3.76E−01 4.97E−01 8090 AI732254 1.45E−02 3.96E−01

MUC2 4583 Mucin-2 Q02817 4.04E−01 3.47E−01 25161 AA534503 4.62E−01 5.30E−01

CLDN18 51208 Claudin-18 P56856 3.91E−01 2.98E−01 11694 AI820565 3.45E−01 6.81E−01

DHRS4 10901 Dehydrogenase/reductaseSDRfamilymember4 Q9BTZ2 4.87E−01 2.76E−01 18212 AA429946 6.46E−01 9.40E−01

DHRS3 9249 Short-chaindehydrogenase/reductase3 O75911 3.98E−01 1.68E−01 1133 AA171606 7.77E−01 8.00E−01

HSD11B2 3291 Corticosteroid11-beta-dehydrogenaseisozyme2 P80365 3.70E−01 3.27E−01 174 W95082 6.95E−01 8.90E−01

PGC 5225 Gastricsin P20142 4.34E−01 3.00E−01 19526 AI674972 3.04E−02 3.25E−01

REG4 83998 Regeneratingislet-derivedprotein4 Q9BYZ8 4.92E−01 4.95E−01 3562 AA535703 6.40E−01 7.55E−01

PRR15 222171 ProteinPRR15 Q8IV56 4.33E−01 4.87E−01 17960 AA515032 8.49E−01 6.52E−01

GMPR 2766 GMPreductase1 P36959 4.07E−01 3.27E−01 19 AA406242 5.21E−01 9.33E−01

CLMN 79789 Calmin Q96JQ2 2.91E−01 4.71E−01 23849 AA775028 7.31E−01 8.14E−01

ABCC3 8714 Canalicularmultispeciﬁcorganicaniontransporter2 O15438 4.72E−01 2.75E−01 15480 AA429895 7.51E−01 6.54E−01

a _Gene_symbols_were_derived_from_UniGene. b _Gene_IDs_were_derived_from_Entrez_Gene_database.

c_Accession_numbers_of_proteins_were_derived_from_SWISS-PROT_and_NCBI_nonredundant_databases.

d_Ratios_were_calculated_by_dividing_the_mean_expression_value_of_primary_tumor_samples_(T)_by_that_of_non-tumor_samples_(NT).

e _Ratios_were_calculated_by_dividing_the_mean_expression_value_of_lymph_node_metastasis_samples_(LNM)_by_that_of_primary_tumor_samples_(T). f _Probe_IDs_were_derived_from_NCBI_GEO_platform_(GPL1283).

(11)

193

oxidationreduction,orhormonemetabolism,whilethegenes withdiscordantexpressionpatternwereenrichedinGOterms relatedtophagocytosisandcelladhesion.Theseresultsare consistent with the ﬁndings of Schwänhausser et al. [38], whoreportedthatgenesrelatedtooxidationreduction,and metabolismwereenrichedinthegroupwithstableand con-cordantmRNAandproteinexpression.Theyalsofoundthat theGOtermsofcelladhesionwereenrichedinthegroupof genesthatwas characterizedbystablemRNAbut unstable proteinandwasexpectedtohavediscordantexpression[38]. Unfortunately,thereliabilityofouranalysisdidnotreachtheir onebecausethenumberofsamplesinourproteomicstudy waslimitedandmRNAexpressiondatawasobtainedfrom differentpatientcohorts.Furtherinvestigationofadditional sampleswillberequiredtogeneralizeourobservations.The useofpublictranscriptomedatabaseswillsolvetheproblems ofproteomicbiomarkerstudies,andtherefore,these investi-gationsshouldbecontinued.

Inourstudy,highlycomprehensiveandreproducible pro-teomicanalysisperformed usingalabel-free quantification method showed downregulation of ITGB3 gene expression inthe LNM tissues. Downregulation ofITGB3 represents a prosurvival response for overcoming apoptotic IMD at the metastaticsite.Furtherstudiesontheprognosticand biolog-icalsignificanceofITGB3mayleadtonovelrisk-stratification approachesforgastriccancer.

Acknowledgements

WewouldgreatlyappreciateDr.MSasagawa(Nanbugo Gen-eralHospital),Dr.NShimakage(NagaokaRedCrossHospital), Dr.TSuda (NipponDentalUniversityMedicalHospital),Dr. N Katayanagi (Niigata City General Hospital), Dr. T Tada (TachikawaGeneral Hospital), Dr. SShimoda (Shibata Hos-pital), Dr. N Musha (Saiseikai Niigata Daini Hospital), Dr. H Tomita (Sakamachi Hospital) and K Ueki (Kashiwazaki GeneralHospitaland MedicalCenter)tocollect theclinical samplesandtheclinicopathologicaldata.Thisworkwas sup-portedbytheNationalCancerCenterResearchCoreFacility andtheNationalCancerCenterResearchandDevelopment Fund (23-A-8 and 23-A-10). Hiroshi Ichikawa was awardee of Research Resident Fellowship from the Foundation for PromotionofCancerResearch(Japan)forthe3rdTerm Com-prehensive10-YearStrategyforCancerControl.

r

e

f

e

r

e

n

c

e

s

[1] FerlayJ,ShinHR,BrayF,FormanD,MathersC,ParkinDM. Estimatesofworldwideburdenofcancerin2008: GLOBOCAN2008.IntJCancer2010;127:2893–917. [2] IsobeY,NashimotoA,AkazawaK,OdaI,HayashiK,

MiyashiroI,etal.GastriccancertreatmentinJapan:2008 annualreportoftheJGCAnationwideregistry.Gastric Cancer2011;14:301–16.

[3] WangW,LiYF,SunXW,ChenYB,LiW,XuDZ,etal. Prognosisof980patientswithgastriccanceraftersurgical resection.ChinJCancer2010;29:923–30.

[4] ParkSR,KimMJ,RyuKW,LeeJH,LeeJS,NamBH,etal. Prognosticvalueofpreoperativeclinicalstagingassessedby

computedtomographyinresectablegastriccancerpatients: aviewpointintheeraofpreoperativetreatment.AnnSurg 2010;251:428–35.

[5] DengJ,LiangH,SunD,ZhangR,ZhanH,WangX.Prognosis ofgastriccancerpatientswithnode-negativemetastasis followingcurativeresection:outcomesofthesurvivaland recurrence.CanJGastroenterol2008;22:835–9.

[6] ChafferCL,WeinbergRA.Aperspectiveoncancercell metastasis.Science2011;331:1559–64.

[7] XieHL,LiZY,GanRL,LiXJ,ZhangQL,HuiM,etal. Differentialgeneandproteinexpressioninprimarygastric carcinomasandtheirlymphnodemetastasesasrevealedby combinedcDNAmicroarrayandtissuemicroarrayanalysis.J DigDis2010;11:167–75.

[8] LiuX-P,LiD-Y,LiuX-L,XuJ-D,FuruyaT,KawauchiS,etal. Comparisonofchromosomalaberrationsbetweenprimary tumorsandtheirsynchronouslymph-nodemetastasesin intestinal-typegastriccarcinoma.PatholResPract 2009;205:105–11.

[9] SilvestriA,CalvertV,BellucoC,LipskyM,DeMariaR,DengJ, etal.Proteinpathwayactivationmappingofcolorectal metastaticprogressionrevealsmetastasis-speciﬁcnetwork alterations.ClinExpMetastasis2013;30:309–16.

[10] RuschVW,RiceTW,CrowleyJ,BlackstoneEH,Rami-PortaR, GoldstrawP.TheseventheditionoftheAmericanJoint CommitteeonCancer/InternationalUnionAgainstCancer StagingManuals:theneweraofdata-drivenrevisions.J ThoracCardiovascSurg2010;139:819–21.

[11] KondoT,HirohashiS.Applicationofhighlysensitive ﬂuorescentdyes(CyDyeDIGEFluorsaturationdyes)tolaser microdissectionandtwo-dimensionaldifferencegel electrophoresis(2D-DIGE)forcancerproteomics.NatProtoc 2007;1:2940–56.

[12] ShannonP,MarkielA,OzierO,BaligaNS,WangJT,Ramage D,etal.Cytoscape:asoftwareenvironmentforintegrated modelsofbiomolecularinteractionnetworks.GenomeRes 2003;13:2498–504.

[13] CroftD,O’KellyG,WuG,HawR,GillespieM,MatthewsL, etal.Reactome:adatabaseofreactions,pathwaysand biologicalprocesses.NucleicAcidsRes2011;39:D691–7. [14] HuangdaW,ShermanBT,LempickiRA.Systematicand

integrativeanalysisoflargegenelistsusingDAVID bioinformaticsresources.NatProtoc2009;4:44–57. [15] XieLQ,ZhaoC,CaiSJ,XuY,HuangLY,BianJS,etal.Novel

proteomicstrategyrevealcombinedalpha1antitrypsinand cathepsinDasbiomarkersforcolorectalcancerearly screening.JProteomeRes2010;9:4701–9.

[16] ThakurD,RejtarT,WangD,BonesJ,ChaS,Clodfelder-Miller B,etal.Microproteomicanalysisof10,000lasercaptured microdissectedbreasttumorcellsusingshort-rangesodium dodecylsulfate-polyacrylamidegelelectrophoresisand porouslayeropentubularliquidchromatographytandem massspectrometry.JChromatogrA2011;1218:8168–74. [17] ZhangY,XuB,LiuY,YaoH,LuN,LiB,etal.Theovarian

cancer-derivedsecretory/releasingproteome:arepertoireof tumormarkers.Proteomics2012;12:1883–91.

[18] PernemalmM,OrreLM,LengqvistJ,WikstromP,Lewensohn R,LehtioJ.Evaluationofthreeprincipallydifferentintact proteinprefractionationmethodsforplasmabiomarker discovery.JProteomeRes2008;7:2712–22.

[19] JafariM,PrimoV,SmejkalGB,MoskovetsEV,KuoWP,Ivanov AR.Comparisonofin-gelproteinseparationtechniques commonlyusedforfractionationinmass

spectrometry-basedproteomicproﬁling.Electrophoresis 2012;33:2516–26.

[20] GygiSP,RistB,GerberSA,TurecekF,GelbMH,AebersoldR. Quantitativeanalysisofcomplexproteinmixtures usingisotope-codedafﬁnitytags.NatBiotechnol 1999;17:994–9.

(12)

[21] RossPL,HuangYN,MarcheseJN,WilliamsonB,ParkerK, HattanS,etal.Multiplexedproteinquantitationin

Saccharomycescerevisiaeusingamine-reactiveisobaric taggingreagents.MolCellProteomics2004;3:1154–69. [22] OngSE,BlagoevB,KratchmarovaI,KristensenDB,SteenH,

PandeyA,etal.Stableisotopelabelingbyaminoacidsincell culture,SILAC,asasimpleandaccurateapproachto expressionproteomics.MolCellProteomics2002;1:376–86. [23] LiZ,AdamsRM,ChoureyK,HurstGB,HettichRL,PanC.

Systematiccomparisonoflabel-free,metaboliclabeling,and isobaricchemicallabelingforquantitativeproteomicson LTQOrbitrapVelos.JProteomeRes2012;11:1582–90. [24] MerlJ,UefﬁngM,HauckSM,vonToerneC.Direct

comparisonofMS-basedlabel-freeandSILACquantitative proteomeproﬁlingstrategiesinprimaryretinalMullercells. Proteomics2012;12:1902–11.

[25] HoodJD,ChereshDA.Roleofintegrinsincellinvasionand migration.NatRevCancer2002;2:91–100.

[26] FilardoEJ,BrooksPC,DemingSL,DamskyC,ChereshDA. RequirementoftheNPXYmotifintheintegrinbeta3 subunitcytoplasmictailformelanomacellmigration invitroandinvivo.JCellBiol1995;130:441–50.

[27] GalliherAJ,SchiemannWP.Beta3integrinandSrcfacilitate transforminggrowthfactor-betamediatedinductionof epithelial-mesenchymaltransitioninmammaryepithelial cells.BreastCancerRes2006;8:R42.

[28] LeiY,HuangK,GaoC,LauQC,PanH,XieK,etal.Proteomics identiﬁcationofITGB3asakeyregulatorinreactiveoxygen species-inducedmigrationandinvasionofcolorectalcancer cells.MolCellProteomics2011;10.M110.005397.

[29] LiapisH,FlathA,KitazawaS.IntegrinalphaVbeta3 expressionbybone-residingbreastcancermetastases. DiagnMolPathol1996;5:127–35.

[30] ChenJ,ZhangJ,ZhaoY,LiJ,FuM.Integrinbeta3

down-regulatesinvasivefeaturesofovariancancercellsin SKOV3cellsubclones.JCancerResClinOncol

2009;135:909–17.

[31] PartheenK,LevanK,OsterbergL,ClaessonI,FalleniusG, SundfeldtK,etal.Fourpotentialbiomarkersasprognostic factorsinstageIIIserousovarianadenocarcinomas.IntJ Cancer2008;123:2130–7.

[32] PartheenK,LevanK,OsterbergL,ClaessonI,SundfeldtK, HorvathG.ExternalvalidationsuggestsIntegrinbeta3as prognosticbiomarkerinserousovarianadenocarcinomas. BMCCancer2009;9:336.

[33] WuY,ZuoJ,JiG,SaiyinH,LiuX,YinF,etal.Proapoptotic functionofintegrinbeta(3)inhumanhepatocellular carcinomacells.ClinCancerRes2009;15:60–9.

[34] ChuYQ,YeZY,TaoHQ,WangYY,ZhaoZS.Relationship betweencelladhesionmoleculesexpressionandthe biologicalbehaviorofgastriccarcinoma.WorldJ Gastroenterol2008;14:1990–6.

[35] StupackDG,PuenteXS,BoutsaboualoyS,StorgardCM, ChereshDA.Apoptosisofadherentcellsbyrecruitment ofcaspase-8tounligatedintegrins.JCellBiol2001;155: 459–70.

[36] vanderFlierA,SonnenbergA.Functionandinteractionsof integrins.CellTissueRes2001;305:285–98.

[37] ChenYR,JuanHF,HuangHC,HuangHH,LeeYJ, LiaoMY,etal.Quantitativeproteomicandgenomic proﬁlingrevealsmetastasis-relatedproteinexpression patternsingastriccancercells.JProteomeRes 2006;5:2727–42.

[38] SchwanhausserB,BusseD,LiN,DittmarG,SchuchhardtJ, WolfJ,etal.Globalquantiﬁcationofmammaliangene expressioncontrol.Nature2011;473:337–42.

www.sciencedirect.com

h t t p : / / w w w . e l s e v i e r . c o m / l o c a t e / e u p r o t

(http://creativecommons.org/licenses/by-nc-nd/3.0/

http://cancer.cellmap.org/

doi:10.1016/j.euprot.2014.03.001.

http://wiki.reactome.org/index.php/Reactome

2010;127:2893–917.

2011;14:301–16.

2010;29:923–30.

Park

2008;22:835–9.

2011;331:1559–64.

2010;11:167–75.

Liu

2013;30:309–16.

2010;139:819–21.

Kondo

Shannon

2011;39:D691–7.

2009;4:44–57.

2010;9:4701–9.

2011;1218:8168–74.

2012;12:1883–91.

2008;7:2712–22.

Jafari

Gygi

2004;3:1154–69.

2002;1:376–86.

2012;11:1582–90.

2012;12:1902–11.

2002;2:91–100.

1995;130:441–50.

2006;8:R42.

2011;10.

1996;5:127–35.

Chen

2008;123:2130–7.

2009;9:336.

2009;15:60–9.

2008;14:1990–6.

2001;155:459–70.

2001;305:285–98.

Chen

2011;473:337–42.