therapeutic discovery molecular cancer therapeutics from ... · connected subgraphs, in the...

Ther

FroTre

Ming-Chun-

Abst

Intro

Nascy incommcestry

AuthorOncoloInstitut5DeparUniver7DeparSciencInformaof Otol11DepaKainanInstitute

Note: SCancer

Correscine, N228267Jin-ChiGeneraext. 56

doi: 10

©2010

www.a

D

Published OnlineFirst August 17, 2010; DOI: 10.1158/1535-7163.MCT-09-0966
Published OnlineFirst on August 17, 2010 as 10.1158/1535-7163.MCT-09-0966
Molecular
Cancer
apeutics

apeutic Discovery

m NPC Therapeutic Target Identification to Potential

Ther

atment Strategy

Ying Lan1,3,10, Chi-Long Chen6,7, Kuan-Ting Lin4,8, Sheng-An Lee11, Wu-Lung R. Yang9,
Nan Hsu8,12, Jaw-Ching Wu3, Ching-Yin Ho5,10, Jin-Ching Lin2,3, and Chi-Ying F. Huang3,4
ractNas

southetheraplectedmaryinterainvolvconnethreefurthegenesinvolvtlenecthe intarget

on cance. The etio

s' Affiliatiogy, Taichues of 3Clintment of Msity; 6Depatment of Pae, Academtion Enginearyngologyrtment ofUniversity, University

upplementTherapeuti

ponding Auational Yan904; Fax:ng Lin, Depl Hospital13; Fax: 886

.1158/1535-

American A

acrjourna

ownloade

opharyngeal carcinoma (NPC) is relatively rare in Western countries but is a common cancer inrn Asia. Many differentially expressed genes have been linked to NPC; however, how to prioritizeeutic targets and potential drugs from unsorted gene lists remains largely unknown. We first col-558 upregulated and 993 downregulated NPC genes from published microarray data and the pri-literatures. We then postulated that conversion of gene signatures into the protein-proteinction network and analyzing the network topologically could provide insight into key regulatorsed in tumorigenesis of NPC. Of particular interest was the presence of cliques, called fullycted subgraphs, in the inferred NPC networks. These clique-based hubs, connecting with more thanqueries and ranked higher than other nodes in the NPC protein-protein interaction network, werer narrowed down by pathway analysis to retrieve 24 upregulated and 6 downregulated bottleneckfor predicting NPC carcinogenesis. Moreover, additional oncogenes, tumor suppressor genes, genesed in protein complexes, and genes obtained after functional profiling were merged with the bot-k genes to form the final gene signature of 38 upregulated and 10 downregulated genes. We useditial and final NPC gene signatures to query the Connectivity Map, respectively, and found thatreduction through our pipeline could efficiently uncover potential drugs with cytotoxicity to

cancer cells. An integrative Web site (http://140.109.23.188:8080/NPC) was established to facilitate
NPCfuture NPC research. This in silico approach, from target prioritization to potential drugs identification,might be an effective method for various cancer researches. Mol Cancer Ther; 9(9); OF1–13. ©2010 AACR.
withenvirocogenexpre

duction

opharyngeal carcinoma (NPC) is a rare malignan-most parts of the world but is one of the most

rs among those of Chinese or Asian an-logy of NPC is thought to be associated

NPC,NPCbe eluPro

tuallyhavindefine11). Tthe idthe netherapintegrmaticin thiIn a

refinetant. AsensitadvanmolecEBV-bare in

ns: Departments of 1Otolaryngology and 2Radiationng Veterans General Hospital, Taichung, Taiwan;ical Medicine and 4Biomedical Informatics, andedicine, School of Medicine, National Yang-Mingrtment of Pathology, Taipei Medical University;thology, WanFang Hospital; 8Institute of Informationia Sinica; 9Department of Computer Science andering, National Taiwan University; and 10Department, Taipei Veterans General Hospital, Taipei, Taiwan;Computer Science and Information Engineering,, Taoyuan, Taiwan; and 12Information Sciencesof Southern California, Marina del Rey, California

ary material for this article is available at Molecularcs Online (http://mct.aacrjournals.org/).

thors: Chi-Ying F. Huang, Institute of Clinical Medi-g-Ming University, Taipei 112, Taiwan. Phone: 886-886-228224045. E-mail: [email protected] andartment of Radiation Oncology, Taichung Veterans, Taichung 114, Taiwan. Phone: 886-423592525-423741316. E-mail: [email protected]

7163.MCT-09-0966

ssociation for Cancer Research.

ls.org

on March 28, 2020. © 201mct.aacrjournals.org d from

a complex interaction of genetic, EBV exposure,nmental, and dietary factors. Although some on-es, tumor suppressor genes, and microarrayssion data have been previously reported ina complete understanding of the pathogenesis ofin the context of global gene expression remains tocidated (1–9).tein-protein interactions (PPI) are important for vir-every biological process. In a PPI network, nodesg more than one connection with another node ared as hubs and are more likely to be essential (10,he key challenge facing a disease PPI network isentification of a node or combination of nodes intwork whose perturbation might result in a desiredeutic outcome. We have previously constructed anated PPI web service, POINeT (12), as a bioinfor-s tool to construct and to analyze the NPC networks study.ddition to elucidating the pathogenesis of NPC, thement of current treatment modalities is also impor-lthough NPC is highly radiosensitive and chemo-

ive, the treatment of patients with locoregionallyced disease remains problematic. Many specificular-targeted therapies, epigenetic therapies, and
ased immunotherapy have been developed andclinical trials (Supplementary Table S1; ref. 1).
OF1

0 American Association for Cancer Research.

http://mct.aacrjournals.org/

It istify pquest(http:that tcouldtargetpliedand Ctial drsmallture inegatin cMtreatlowedthan tquire

Mate

CompAcq

NPCthe NstudymicroTablesult onasopThe

text mof 200taininnot hafurthefirst eIntelliThe Ggenethe ofwe mmentistractthe gegenessite (htools ury TabgenesPPI inEva

netwofrom tfrom Gall itsotherof thr

we hainto PnetwocountPPI nthe NWe

matiomainDatabtabaselian pcliquein proproteiRan

relativthrouness cis thesentsworknodeity caand fiThe

overre(23) tsets oregulcliqucliquerankeby Cotwo pin thegenesfollow

Wevalueeachfor YeThe

downtionaltion,ref. 24pathwKEGGTION

Lan et al.

Mol COF2

D


not clear how to elucidate key regulators and iden-otential drugs for NPC treatment. To address theseions, we first collected NPC-associated genes//140.109.23.188:8080/NPC) and hypothesizedhe PPI network, derived from the gene signature,be analyzed topologically to prioritize potentials. We further performed pathway analysis and ap-gene signatures to drug-gene interaction databasesonnectivity Map (cMap; refs. 13, 14) to find poten-ugs for the treatment of NPC. It is supposed that amolecule may potentially reverse the disease signa-f the molecule-induced signature is significantlyively correlated with the disease-induced signatureap (15–17). In short, identifying potential drugs toNPC by using an in silico screening approach fol-by empirical validation might be easier and fasterhose traditional drug discovery pipelines that re-tremendous effort and time.

rials and Methods

utational methodsuiring NPC-related gene sets and constructingPPI network. Two major components constitutedPC-related gene expression signatures in this. One component included the collection of thearray profiles from three studies (SupplementaryS2; refs. 4, 5, 7). All microarray data were the re-f nontreated NPC tissues compared with normalharyngeal tissues.second part of the gene collections consisted of theining of NPC-related PubMed abstracts. By March8, we extracted 4,939 abstracts from PubMed con-g the keyword “Nasopharyngeal carcinoma” butving the keywords “SNP” or “polymorphism.” Tor extract the genes mentioned in the abstracts, wentered all these abstracts into the Adaptive Internetgent Agents laboratory's Gene Mention Tagger (18).ene Name Service (19) was used to translate thesenames into corresponding gene identifiers, such asficial gene symbol and the Entrez gene ID. Then,anually read the top 10 abstracts with most genesoned from the above method and another 150 ab-s published from 2007 to 2008 to further annotatenes as upregulated or downregulated genes. Theseand annotations can be accessed through our Webttp://140.109.23.188:8080/NPC), and the Web sitesed in this study are summarized in Supplementa-le S3. We inputted the above-collected NPC-relatedas query terms into the POINeT (12) to detect theNPC.luation of cliques and complexes from the PPIrk. The cliques of the PPI network were calculatedhe following definition of cliques, a term borrowedraph Theory. A clique is a part of a graph in whichnodes are completely connected to each other. In
words, a 3-clique is a completely connected graphee nodes, which is a triangle. From this definition,
stringther i

ancer Ther; 9(9) September 2010

on March 28, 2020. © 201mct.aacrjournals.org ownloaded from

ve developed CliquePOINT, which was embeddedOINeT, to calculate these cliques in the NPC PPIrk. Expanding the definition of the 3-clique, we alsoed the number of 4-cliques and 5-cliques in the NPCetwork, and there is no clique larger than 5-cliques inPC PPI network.further collected and integrated the complex infor-n to obtain an abundant data set from public do-databases, including the Human Protein Referencease (20), the Protein Interacting in the Nucleus da-(21), and the Comprehensive Resource of Mamma-rotein complexes (22), and asked whether thes identified from the PPI network were involvedtein complexes. The cliques having more than threens involved in complexes were reported.king the hubs in the PPI network. To elucidate thee roles of each node, we analyzed node centralitygh POINeT, including degree centrality (DC), close-entrality (CC), and eccentricity centrality (EC). DCnumber of link incident upon a node. CC repre-the closeness between nodes in the biological net-. EC is the longest distance required for a givento reach the entire network. By conducting central-lculation, nodes in global networks can be rankedltered using various network analysis formulas.enriched pathways from the ConsensusPathDBpresentation analysis. We used ConsensusPathDBo perform overrepresentation analysis on the fourf gene lists: (a) upregulated genes in NPC, (b) down-ated genes in NPC, (c), upregulated genes aftere analysis, and (d) downregulated genes afteranalysis. The significant pathway results were

d by using an F score instead of the P value givennsensusPathDB. The F score was used to normalizearameters: (a) the percentage of overlapping genespathway and (b) the percentage of overlappingin the input list. To normalize these, we used theing formula:

Fscore ¼ 2ðA� BÞðAþ BÞ

compared the P values to evaluate whether the Ps degrade after clique analysis and thereby givepathway a score of degradation (0 for No and 1s).final NPC gene signature. The 98 up-clique and 51-clique genes were used as queries to perform func-annotation clustering on the Database for Annota-Visualization, and Integrated Discovery (DAVID;), respectively. The clustering was done on sevenay resources: BBID, BIOCARTA, EC_NUMBER,_COMPOUND, KEGG_PATHWAY, KEGG_REAC-, and PANTHER_PATHWAY. The classification
ency was set to “Medium.” For each cluster, we fur-ntersected the genes of the pathways to obtain the
Molecular Cancer Therapeutics



“bottldownAm

supprfoundbottleNPC.genesHie

KEGGquerieof DAscript0.05) itern tFor uare 1 (sure fto “PeFoo

To cotion (fromServiccorrestrez gwas dmappupregtargetApp

connegeneploredgenestag filfiles wsults silaritiegenescMapthreesists odowndown399 uwhichThe thgenesscores

BioloImm

fixed,casespressi(mous1:400)

immuviewtionsfor anoven.withgoat sfor 30PBS twith bdin- broomagaindiamminutstainiand mlar denormand np53 afor thBCL2Cel

TW01(NatiderivepatienautheDr. S.was dtient (10% fand sviabilnitro-inneraccorfour h103 cemicro(Sigmthe sexposat diffwith mof thewas mMax2

Resu

NPCTo

tureshave

Identify Targets and Drugs for NPC

www.a

D


eneck” genes. Twenty-four upregulated and sixregulated bottleneck genes were obtained.ong cliques, those, including oncogenes, tumoressor genes, genes involved in complex, and genesby group functional profiling, were added into theneck genes list to obtain the final gene signature ofIt includes 38 upregulated and 10 downregulated(Supplementary Table S4).rarchical clustering the final gene signature inpathways. We used the final gene signature as

s to conduct the functional annotation clusteringVID against the KEGG pathway database. A perlwas written to convert the pathway records (P <nto a gct file, which can be uploaded onto GenePat-o perform hierarchical clustering and visualization.pregulated and downregulated genes, the valuesred) and −1 (green), respectively. The distance mea-or both genes (row) and pathways (column) was setarson correlation, absolute value.”d and Drug Administration–approved drug targets.llect target genes of Food and Drug Administra-FDA)–approved drugs, the chemical-protein linksSTITCH (25) was downloaded. Then, Gene Namee (19) was used to translate the protein ID to itsponding HUGO-approved gene symbol and En-ene ID. The DrugCard file from Drug Bank (26)ownloaded. We selected FDA-approved drugs,ed the drugs' corresponding genes with the NPCulated genes, and finally identified known drugs in the NPC upregulated PPI network.lying NPC gene signatures to cMap. Functionalctions between various NPC gene signatures andsignatures induced by small molecules were ex-using the cMap database (13, 14). The upregulatedwere grouped and their probe sets formed the upe and so did the downregulated genes. These twoere used to query the cMap database, and the re-howed the most significant similarities and dissim-s to the database profiles. The 558 up and 993 downwould convert to >1,000 probe sets. Because thecould only take up to 1,000 probe sets per input,groups of NPC genes were used. The first group con-f 100 randomly chosen sets of 100 upregulated/regulated probe sets from whole 558 up and 993NPC gene signature. The second group consists ofpregulated and 443 downregulated probe sets,represent first 70% ranked queries served as hubs.ird group, the final gene signature, consists of 38 upand 10 down genes. Only drugs with negativeand a P value of <0.05 were retained.

gical methodsunohistochemical analysis in NPC. Formalin-paraffin-embedded biopsy specimens of 143 NPCwere collected and analyzed for detection of the ex-on of p53 (mouse anti-human p53, 1:50) and BCL2
e anti-BCL2, 1:80; DAKO), BAX (mouse anti-BAX,, and MYC (mouse anti-MYC, 1:50; Santa Cruz) by
the NPubM

acrjournals.org


nohistochemistry (IHC) with the institutional re-board approval. Briefly, 5 to 6 μm of paraffin sec-were deparaffinized and placed into citrate buffertigen retrieval once placed inside a microwaveAfter the sections were cooled down and rinsedPBS, the sections were incubated with 5% normalerum followed by reaction with primary antibodyminutes at room temperature, then washed with

hrice at 3 minutes each. The sections were reactediotinylated second antibody followed by streptavi-iotin complex in the LsAB detection kit (DAKO) attemperature for 10 minutes and washed with PBS. The sections were colorized using freshly preparedinobenzidine solution containing H2O2 for 2 to 5es. After washing with running water and counter-ng with hematoxylin, the sections were dehydratedounted. Positive staining showed brownish granu-posits in the nuclei of cells. Adenocarcinoma andal mucosa gland of the colon were used as positiveegative controls, respectively, for the expression ofnd MYC, whereas follicular lymphoma was usede positive and negative control of the expression ofand BAX (Supplementary Fig. S1).l culture and cell viability test. NPC cell lines,, TW03, and TW04 provided by Dr. C.T. Linonal Taiwan University, Taoyuan, Taiwan), wered from primary nasopharyngeal tumors of Chinesets with de novo NPC and had been tested andnticated (27). NPC cell line BM1, provided byK. Liao (Chang Gung University, Taoyuan, Taiwan),erived from bone metastatic lesions of an NPC pa-28). NPC cell lines were maintained in DMEM withetal bovine serum containing penicillin (100 U/mL)treptomycin (100 μg/mL) in 5% CO2 at 37°C. Cellity was determined using the 2,3-bis[2-methoxy-4-5-sulfophenyl]-2H-tetrazolium-5-carboxanilidesalt (XTT) cell viability assay kit (Sigma-Aldrich),ding to the manufacturer's instructions. Twenty-ours after seeding cells at a concentration of 2 ×lls per well in 100 μL culture medium in a 96-wellplate, cells were then treated with Trichostatin Aa-Aldrich) and Trifluoperazine (Sigma-Aldrich),elected small molecules from cMap. Cells wereed with or without small molecules for 72 hourserent concentrations. Then, the cells were incubatededium containing XTT in an amount equal to 20%culture medium volume for 2 hours. Absorbanceeasured using a microplate reader (Spectral

50) at 450 nm.

lts

gene collectionssystematically analyze the gene expression signa-of NPC and identify potential drugs for NPC, weset up in silico approaches (Fig. 1). We collected
PC gene sets from two sources: one gene set fromed with 70 upregulated and 78 downregulated
Mol Cancer Ther; 9(9) September 2010 OF3



genes(SupplatedtheseNPCregula

InferTo

seemilatedtect tSupplinteraNeT, tconnewereNPCturalgenesPPIs,hand,query

The iinteraOf

is theconneactionquerycliquetively3-cliquS6; FimuchnetwovolveassocMYCparticThe

vide inode.morethe noquerynetwolevel oteinsrankinThosehigheBec

of thesponstionalfor prDatablian pthe Nare fiprote

FigureNPC geextractesignatuderivedfrom thto quer6 downcancer-under cand geDrugBain cMapA totalfinal ge

Lan et al.

Mol COF4

D


, and the other from three major microarray studieslementary Table S2; refs. 4, 5, 7) with 512 upregu-genes and 936 downregulated genes. By mergingtwo data sets, the gene expression signature of

ne signature for querying cMap to identify potential dugs.

contained 558 upregulated genes and 993 down-ted genes (http://140.109.23.188:8080/NPC).

thesoconsi



red NPC PPI networkuncover the potential interaction networks of thesengly unrelated NPC upregulated and downregu-genes, the Web site tool, POINeT, was used to de-he PPI in NPC, and results were summarized inementary Table S5. Despite many queries withoutcting proteins, based on our PPI collections in POI-he queries of NPC-related proteins formed a highlycted interactome. A total of 8,231 and 7,728 PPIsidentified in the upregulated and downregulatedPPI networks, respectively. The fundamental struc-details revealed that 257 of 558 NPC upregulatedinteract with each other and form 492 query-queryconstituting the interaction networks. On the other324 of 993 NPC downregulated queries form 395-query PPIs.

nferred NPC network consists of highlyctive cliques and complexes

particular interests in the inferred NPC PPI networkpresence of cliques (29), which refer to completelycted subgraphs. Nodes within a clique have inter-s with all the others. In our analysis, the NPC-query network contains 198 and 21 subgraphs ofs in upregulated and downregulated genes, respec-. In the upregulated PPI network, there are 170es, 26 4-cliques, and 2 5-cliques (Supplementary Tableg. 2A). The count of cliques in NPC PPI network ishigher than the count of cliques in a random PPIrk (Supplementary Fig. S2). The top 30 proteins in-d in cliques are listed and ranked by the number ofiated cliques (Supplementary Table S7). BRCA1,, EGFR, TP53, and CDC2 are the top five proteinsipating in a large number of cliques.analysis of node centrality characteristics may pro-nsights into the relative roles and features of eachTo address whether clique proteins are relativelyimportant hubs in the PPI network, we prioritizeddes of the major subnetwork, which consists of 247proteins (or nodes), in the NPC upregulated PPIrk (Supplementary Table S8). The 3,725 nodes ofne major subnetwork, which consists of query pro-with neighbor nodes, were also ranked. Differentg methods, including DC, EC, and CC, were used.nodes, which are also clique proteins, are ranked

r than those that are not clique proteins (Fig. 3).ause cliques have more interactions than the restgraph, and these protein interactions may be re-ible for the formation of protein complexes or func-modules (30), we further integrated and searchedotein complexes from the Human Protein Referencease (20), the Comprehensive Resource of Mamma-rotein complexes (22), and the Protein Interacting inucleus database (21). Of upregulated cliques, thereve 3-cliques and four 4-cliques involved in fivein complexes (Table 1; Fig. 2B). The DNA syn-

1. Schematic illustration of in silico approaches to narrow downnes for target identification and potential drug discovery. We firstd and reorganized 558 upregulated and 933 downregulated generes from PubMed and various microarray studies. Then, wethe 98 upregulated clique and 51 downregulated clique genese PPI network by clique analysis. These clique genes were usedy DAVID for pathway analysis to obtain 24 upregulated andregulated bottleneck genes that curb multiple pathways. Therelated pathways were used to search for the drugs currentlylinical trials. These bottleneck genes, combined with oncogenesnes found by group functional profiling, were used to query thenk and STITCH. To increase the number of the query genes used, additional genes that appeared in complexes were added.of 38 upregulated and 10 downregulated genes were used as

me, also known as the DNA replication complex,sts of 15 subunits, including DNA polymerase,




DNAplex (pentaand rloadicompconsisC, isDNAassocassocimanytherefactoring rabe rel

and thedge,relatiogenesare reMLH

OncocliquSix

NFKB5-cliqOverERBB

Figureand comNPC geand 5-cThe qunetworkgenes inetworkand tw5-cliquecircles.tumor sTP53, Mthe toplargestmajor cNPC upRed, updownreclique gdark gr(downregulated cliques).


www.a

D


topoisomerase, and the replication factor C com-31). The replication factor C complex is a hetero-meric protein that is essential for DNA replicationepair, and is also a clamp loader required for theng of PCNA onto dsDNA (32–34). The BASClex, BRCA1-associated genome surveillance thatts of ATM, BLM, MSH2, MSH6, MLH1, and RF-involved in the recognition and repair of aberrantstructure (35). Another complex, the hNop56p-iated preribosomal ribonucleoprotein complex, isated with ribosome biogenesis (36). Interestingly,proteins are shared in these complexes. Finally,is one complex involved in the tumor necrosis-α (TNF-α)f/NF-κB pathway (37). The above find-
ises the possibility that NPC pathogenesis mightated to aberrant DNA replication, DNA repair,
esis (Ssuppr

acrjournals.org


e TNF-α/NF-κB pathway. To the best of our knowl-this finding will be the first report to provide thenship between these complexes and NPC carcino-is. Few proteins in the above five complexes thatlated to NPC include RFC1, PCNA, TOP1, ATM,1, RPL21, and RPL31.

genes and tumor suppressor genes in NPCe genesoncogenes, including EGFR, ERBB2, MYC, RELB,2, and CCND1, were found in the 4-cliques andues from the inferred upregulated NPC network.expression of these oncogenes in NPC, except2, was suggested to be related to NPC carcinogen-

2. Highly interactive cliquesplexes are associated withne signature. A, 4-cliquesliques of NPC PPI network.ery-query interactionof the NPC upregulated

s a highly connectedthat contains 26 4-cliques

o 5-cliques. The twos are grouped in redYellow, oncogenes; green,uppressor genes. BRCA1,YC, EGFR, and CDC2 arefive proteins involved in thenumber of cliques. B, fiveomplexes associated withregulated gene signatures.regulated genes; green,gulated genes. Dark red,enes (upregulated cliques);een, clique genes

upplementary Table S7; refs. 38–42). Three tumoressor genes were found in the 51 downregulated




cliqueCDKNThe dour W

It ispresscludinuclenomic∼40%breasNPCp53, wlate tasenesidly tby Mis undegraveryhigh iAccumEBV Lto expNPC:and eelimincodesp53 (transabindsmemba dealogicbeennanciovereNPC

FindiTo

ical pupregsubjec

Figurecliquescalculalevel onrankedand levNinety-relatively higher than the other nodes in the NPC PPI major subnetworkand lev

Tab

Com

DNARepBAShNo

riTNF

NOT

Lan et al.

Mol COF6

D


genes, including CDKN1A, MLH1, and ATM. Both1A and ATM are downregulated in NPC (7, 43).

el one major subnetwork.

escription of these genes can be accessed througheb site (http://140.109.23.188:8080/NPC).

cal prdown

le ciated with

p Pro

s TOP1,licC BRCA

E: The proteins involved in complexes and proteins that are in NPC u



interesting to note that there are three tumor sup-or genes found in the 98 upregulated cliques, in-ng BRCA1, TP53, and FAS. Briefly, BRCA1, aar phosphoprotein, plays a role in maintaining ge-stability. Mutations in BRCA1 are responsible forof inherited breast cancers and >80% of inherited

t and ovarian cancers; however, its expression inis still unknown. TP53 encodes the tumor proteinhich responds to diverse cellular stresses to regu-rget genes that induce cell cycle arrest, apoptosis,cence, and DNA repair. In normal cells, p53 is rap-urned over by a negative feedback loop mediatedDM2. Mutant p53, noted in 30% to 50% of cancer,able to induce MDM2 transcription and escapesdation, thereby leading to its accumulation at ahigh level in cancer (44). Although p53 levels aren NPC, the mutation of TP53 gene is relatively rare.ulated p53 in NPC is believed to be mediated byMP1 (9, 40, 45). Two reasons have been proposedlain why wild-type p53 fails to induce apoptosis inlow ARF levels due to promoter hypermethylationxcess mutated p63. Wild-type p53 function may beated by the inactivation of the ARF gene, which en-proteins that sequester MDM2 from antagonizing44). Mutated p63, which lacks the NH2-terminalctivation domain required to activate apoptosis,to normal p63 (and p53; ref. 9). FAS protein is aer of the TNF receptor superfamily and containsth domain. It plays a central role in the physio-regulation of programmed cell death and hasimplicated in the pathogenesis of various malig-es and diseases of the immune system. Fas ligandxpression is an unfavorable prognostic marker in(46, 47).

ngs by gene group functional profilingaddress how our NPC signature might turn biolog-rocess term groups (by Gene Ontology) on or off, 98ulated and 51 downregulated clique genes wereted to g:Profiler, respectively (48). A large biologi-

3. In the inferred NPC PPI network, queries with characteristics ofbelong to the top-ranked targets as determined by centralitytion. The nodes of the major subnetwork (query-query PPI) ande major subnetwork of the NPC upregulated PPI network areby DC, CC, and EC. Gray, nodes of the major subnetwork (A)el one major subnetwork (B). Red, nodes, also clique proteins.eight queries that participated in the inferred cliques are ranked

ocess term group is shared by both upregulated andregulated clique genes. The group is mainly related

c domain da
1. Five major complexes asso NPC after analysis using three publi
Clique

pregulated cliques are listed.

Molecular Cancer Th

0 American Association for Cancer Res

tabases

lex (total proteins numbers)
teins involved in complex andNPC gene signature
numbers

ynthesome (15)
TOP2A, RFC1, RFC3, RFC2, RFC5 2 ation factor C complex (5)complex (9)
RFC1, RFC3, RFC2, RFC5
2
6p-associated preribosomal
1, RFC1, RFC2, MSH6, MLH1, ATM
NCL, NPM1, TOP1
11 p5
bonucleoprotein complexes (104)-α/NF-kB pathway (12) NFKB2, NFKBIA, RELB 1

erapeutics

earch.


to thedeathfunctip53 toAmoncludinrelatethe “rThe ggulateERBBall, wdownrelate

PathwTo f

ture, wysis oP valuupregregula294 endownanalyare reP valuFromrichedpathwsignalcancepathwtary TpathwanalymoveNPCFur

for Nby usfinal(Supped toanalypret amorefinalloadearchicthe pathat mpathwamyocytokcommpressoantitu

pathwbe anTo

nect wthe Kanismpressiof TPare ov

FDA-To

approSTITC566 alatedand 2downS12).cogentargetgenestheraamondrugsplatinical trtargebeingpeuticbinati

FindiusedwhosFro

be retate cmyelocanceclinicafinedtabaschem106 uuallydrugsare unMoreogenesin clinTablepotenchemat diffprehe


www.a

D


regulation of biological processes, cell cycle, cell, and cell development. These important biologicalons are altered, thereby leading to the activation ofdeal with the disturbed physiologic circumstances.g the downregulated clique genes, three genes, in-g CDKN1A, HDAC3, and PRKCZ, are shown to bed to the “regulation of programmed cell death” andegulation of apoptosis” by using Traceable author.enes with Traceable author references in the upre-d clique genes in the phosphorylation group are2, STAT1, and TYK2 (Supplementary Fig. S3). Over-e used gene group profiling to further identify threeregulated genes and three upregulated genes thatto the growth of tumors.

ay analysis of NPC gene signatureind the enriched pathways of our NPC gene signa-e performed an overrepresentation pathway anal-

n ConsensusPathDB (23). Under the threshold of ae of <0.01, there were 484 enriched pathways forulated genes and 222 enriched pathways for down-ted genes in the original NPC signature; 409 andriched pathways were found for upregulated andregulated genes, respectively, by using the cliquesis. To avoid the complication that small pathwayslatively easier to rank higher according to theire, we used the F score to normalize the ranking.the results of the intersection of the top 100 en-pathways of upregulated gene signature, manyays are directly related to cancer, such as the p53ing pathway, cell cycle–related pathways, bladderr pathways, lung cancer pathways, prostate cancerays, and pancreatic cancer pathways (Supplemen-able S9 and S10). Moreover, most of the enrichedays and their P values did not degrade after cliquesis, suggesting that the clique analysis tends to re-genes not involved in the enriched pathways of ourgene signature.thermore, we performed another pathway analysisPC final gene signature (Supplementary Table S4)ing DAVID. The clustering result shows that thegene signature can be divided into three groupslementary Table S11). All groups are closely relat-cancers, signaling, and cell communications. Thissis provides a convenient way to biologically inter-t the “biological module” level (24). To provide ainsightful view of the relationships between thegene signature and KEGG pathways, we down-d the pathway records (P < 0.01) to perform hier-al clustering (Fig. 4) using GenePattern. Most ofthways are shown to have downregulated genesight cause disruption, whereas there are fiveays having no downregulated blocks. They aretrophic lateral sclerosis, Jak-STAT signaling, adipo-ine signaling, neurodegenerative disease, and cellunication pathways. In addition, the tumor sup-
r, ATM, is shown to be downregulated in onlymor pathways such as apoptosis, p53 signaling
(Suppof the

acrjournals.org


ay, and cell cycle. It implies that the ATM couldimportant missing piece in NPC.investigate how the final NPC gene signature con-ith each other in pathways, we manually referred

EGG pathways to draw a possible molecular mech-of NPC carcinogenesis (Fig. 5A). To confirm the ex-on of selected final upregulated genes, IHC studies53, BCL2, BAX, and MYC were done. All of themerexpressed in tumor cells (Fig. 5B).

approved drug targetsannotate the NPC upregulated genes with FDA-ved drug targets, we integrated databases fromH (25) and Drug Bank (26). We thereby derivednd 827 drug target upregulated and downregu-genes, respectively. Two hundred eighty-nine03 FDA-approved drugs target up-clique and-clique genes, respectively (Supplementary TableThe 191 drugs target up-bottleneck genes and on-es (Supplementary Table S13), whereas 100 drugsdown-bottleneck genes and tumor suppressor(Supplementary Fig. S4). Some well-known chemo-peutic agents already used in several cancers areg the top 100 drug target up-clique genes. Theseinclude paclitaxel, doxorubicin, etoposide, and cis-. Many of these drugs are being studied in NPC clin-ials (Supplementary Table S1), suggesting that ourt prioritizations, particularly those not currentlyused in clinical trials, might reveal potential thera-agents for the treatment of NPC, alone or in com-on with older chemotherapeutic agents.

ng candidate drugs for NPC from drugs beingor being studied in clinical trials in cancerse pathways are related to NPCm the results of the pathway analysis, NPC maylated to several cancer pathways, including pros-ancer, bladder cancer, pancreatic cancer, chronicid leukemia, colorectal cancer, and small cell lungr. We derived 1,692 chemical names with 3,603l trial records of the six types of cancers with re-search limited on drug from the ClinicalTrials da-e (Supplementary Table S3). By intersecting theical names with 289 up-clique drugs, we obtainedp-clique drugs under clinical trials. We then man-selected 83 drugs that are used as antitumorin those clinical trials. Of the 83 drugs, 11 drugsder NPC clinical trials (Supplementary Table S1).ver, 66 of the 83 drugs are targeting up-bottleneckand oncogenes. After excluding the drugs alreadyical trial for NPC, 57 drugs remain (SupplementaryS14). These candidate drugs might be importanttial drugs for future NPC treatment. In addition, 26otherapeutic agents suggested to treat these cancerserent stages were retrieved from the National Com-nsive Cancer Network clinical practice guidelines
lementary Table S15). Individual or combined usageabove FDA-approved drugs may improve current



NPCminim

IdenttreatmBio

genekill N

signatare geregulaseconservedture,

ureG

respatuegunre

thecks,ral salinrodcomhwaes in the figure.

Lan et al.

Mol COF8

D


treatment with enhanced therapeutic effects andized side effects.

ifying potential small molecules for NPCent by applying NPC gene signatures to cMap

active small molecules in cMap that reverse the
signature of NPC may be the potential drugs toPC cells. We used three groups of NPC gene
latedsecon



ures to query the cMap database. The first groupnes randomly selected from whole NPC 559 up-ted and 993 downregulated gene signature; thed group consists of first 70% ranked queriesas hubs; the third group are the final gene signa-

consisting of 38 upregulated and 10 downregu-

FigKEGcorsignuprdowasblolatesignneucellpatgen

genes. By querying cMapd, and the third group of g

Molecular

0 American Association for C

4. The heat map showingpathways withonding NPC final genere. In a given pathway, thelated genes and thegulated genes are denotedred blocks and the greenrespectively. Amyotrophicclerosis, Jak-STATg, adipocytokine signaling,egenerative disease, andmunication are the

ys without downregulated

with the first, theenes, there are 6, 8,

Cancer Therapeutics

ancer Research.


and 8molecbilityshowand trnificaNPCtoneantineazine,groupand letial fo

Discu

Hermicrobetwetissuee.g.,NPCwe betient pis mo

Figuremechanby NPCof selecmolecucarcinoupregudownrenamesoncogemarkedsuppreactivatitwo blothey foarrow,becausexistingselecteThe tumwere po(B, b), Bby IHCdevelopand cohematop53 watumor cwere mthe MYand cytOrigin mand D);(a, b, c


www.a

D


drugs respectively among the 10 top-ranked smallules with antitumor effect (either from cell via-tests or PubMed literatures; Table 4). Here, wecell viability tests of two drugs, trichostatin Aifluoperazine, whose gene signatures in cMap sig-ntly negatively correlated with gene signatures of(Fig. 6A and B). Trichostatin A, a member of his-deacetylase inhibitors, has been used with otheroplastic agents in several clinical trials. Trifluoper-a typical antipsychotic drug of the phenothiazine, can induce apoptosis of B16 melanoma cells (49)
ukemic cells (50). Both of them may have poten-r treating NPC in the future.
severquery

, and d).

acrjournals.org


ssion

e, we constructed NPC gene signature from thearray data of NPC tissues due to poor correlationen the microarray data of the NPC cell lines ands (4). Although we can also find potential drugs,trichostatin A and trifluoperazine, for NPC bycell line signature (Supplementary Table S16),lieve the gene signature collected from many pa-rofiles, but not cell line derived from one patient,re representative of the NPC disease. In fact,
al studies have identified promising drugs bying cMap with the gene signatures of tumor
5. Possible molecularism of NPC carcinogenesisbottleneck genes and IHCted proteins. A, possiblelar mechanism of NPCgenesis. Red blocks, geneslated in NPC; blue blocks,gulated genes. Genemarked in red arenes, and gene namesin green are tumor

ssor genes. Arrow,on; gray line, inhibition. Ifcks are close to each other,rm complexes. Bigger redthe pathway reinforcede of the lack of inhibitor andof enhancer. B, IHC of

d proteins in NPC tumor.or cells of NPC samplessitive for p53 (A, a), BCL2AX (C, c), and MYC (D, d)

. The sections wereed by diaminobenzidineunterstained withxylin. The expression ofs mainly in the nuclei ofells; the BCL2 and BAXainly in the cytoplasm; andC was presented in nucleioplasm of the target cells.agnification, ×200 (A, B, C,Origin magnification, ×400




samplect alinesTo a

act wthat chelp rHenceworkthe dothan tdownlatedgest tcompIt is

genespresshavinthe rethe nethosethis imin theresult

methohypocompintimdefectwith nFurthermore, the clique proteins were used for path-

way an

FigureTrichoswith vaCell viability was evaluated by XTT cell viability assay. Columns, meanfrom th

Table e top ins ra ifferencentr the m bnetw

Rank DC EC CC

1 GRB EGF EGFR2 MYC EPB4 GRB23 TP53 ERBB MYC4 EGFR ADA TP535 BRCA CDH BRCA6 NFKB PDCD NPM7 CDC PDCD CDC8 EPB4 SH2B PRKD9 SMAD SH2D MAPK10 ERBB SCAM KPNB11 STAT LRPPR ESR112 NFKB TOB RPS27131415

NOT

Tabldiffe centra the le majosubn k

Rank DC EC CC

1 MYC GRB MYC2 TP53 PTM EGFR3 GRB2 STAT BRCA4 CDC MYC TP535 EGFR ADA CDC26 BRCA CDC KPNA7 EPB4 NUP1 GRB28 STAT CD4 KPNB9 STAT GTF3C NPM10 HDAC CCT6 PRKD11 NFKB NPM CSE112 PRKD NFYA STAT131415

Lan et al.

Mol COF10

D


les (15–17). Nevertheless, continuous effort to col-dditional NPC signature from various NPC cellis required.sk whether genes participate individually or inter-ith each other to form a network, we hypothesizeonverting gene signature into PPI network mighteveal the potential “hubs” in the inferred network., we use POINeT to construct the NPC PPI net-. From the inferred NPC PPI network, althoughwnregulated network contained more query geneshe upregulated network, the number of PPIs of theregulated network is less than that of the upregu-network (Supplementary Table S5). The data sug-hat the upregulated PPI network in NPC is moreact than the downregulated PPI network.difficult to prioritize targets involved in the carcino-is of NPC, especially for those differentially ex-ed signatures from multiple sources. Cliquesg more interactions within themselves than withst of the graph may be an essential modularity intwork (30). As clique proteins are ranked higher thanthat are not clique proteins (Tables 2 and 3; Fig. 3),plies that clique proteinsmight play important roles

ree independent experiments; bars, SD.

NPC PPI network. Further validating these rankings is needed to find themost appropriate hub-ranking

NOT



d. From both the clique and complex analysis, wethesize that perturbation of these cliques and/orlexes, which do not act independently but actately in an intertwined network, might result in aive NPC PPI network. It is likely that targeting themew therapies may be another way to treat NPC.

E: The clique proteins are in bold.

6. cMap analysis results. Dose-dependent cytotoxicity oftatin A (A) and Trifluoperazine (B). NPC cell lines were incubatedrious concentrations of Trichostatin A and Trifluoperazine for 72 h.

E

0

2. Thality in

: The clique pr

American As

15 proteajor su

oteins are in bold

Molecular C

sociation for Ca

nked by dork

.

ancer Therapeu

ncer Research.

t

2
R 1 2
1
3 1 2 6IP 1 2 6 2 1 3 C 1 3A 1 2 P3 1 3 C IA 1 A 2 2 HDAC CDK ABL1
STAT1 TNK2 NCLBCL2 CDK4 STAT3

alysis to obtain the bottleneck genes, which

e 3. The top 15 proteins ranked by
rentetwor
lity in
vel one r
2
A 3 1
2

1
2 2 1 55 3 4 1 1 4 1 2 A C 2 1 L C 1
IA
3 EZR NFKB STATCDK4 NFKB2 HDAC2KPNB1 NCL EPB41
tics


weresupprgenesfinal gThe fsible5A). WupregTP53,ysis, aof TPprevioptosisprotei

to lescoprowhichpressneedeThe

findinare feture)the ortop-rathat t

Tab

Ran

12345678910

Ran

12345678910

Ran

12345678910

NOT


www.a

D


combined with those being also oncogenes, tumoressor genes, and genes involved in complex andfound by group functional profiling to form theene signature for NPC (Supplementary Table S4).inal gene signature is effective to depict the pos-molecular mechanism of NPC carcinogenesis (Fig.e further performed IHC studies of selected four

ulated genes from final gene signatures, includingBCL2, BAX, and MYC. Consistent with our anal-ll of them were upregulated (Fig. 5B). High level53 has been suggested to be related to EBV asusly discussed (9, 40, 45). In regard to the apo-
pathway, although BCL2 acts as an antiapoptosisn
someh

le op 10 small molecul is queried

k G

TricLYVMS

ReM

Tan

k G

TricLYS

TanTrifl

ProcPhthal

B

k G

TricIf

EtaMNoFu

Trifl

E: Drugs in boldface are those with antitumor effects shown either by

acrjournals.org


sen the effect of BCL2. High level of MYC, an on-tein, activates the CCND1 and DNK4/6 complex,results in cell proliferation. Validation the ex-

ion of additional prioritized targets by IHC isd to further elucidate NPC carcinogenesis.final gene signature is also used to query cMap forg potential drugs (Table 4). In cMap analysis, therewer genes in the third group (the final gene signa-compared with the first group (randomization fromiginal gene set), but more promising drugs on thenked drug list are obtained (Table 4). This indicatedhe narrowed down method is effective to filter
ow noisy genome-wide data. Although there are
by inhibiting BAX, upregulation of BAX seems only 38 upregulated and 10 downregulated genes in

by NPC
4. T es in cMap analys
MA

M

Antips

A

D

A sL

Antips

cell viabilit

0 Americ

gene signatures

roup 1
Description
hostatin A
HDAC inhibitors -294002 PI3-kinase inhibitor
orinostat
HDAC inhibitors edrysone Corticosteroid irolimus ammalian target of rapamycin inhibitor sveratrol ntibacterial and antifungal phytoalexin eticrane diuretic ylsulfathiazole A synthetic opioid, potent analgesic Phthal
Sulconazole Antifungal
espimycin HSP 90 inhibitor
roup 2
Description
hostatin A
HDAC inhibitors -294002 PI3-Kinase inhibitor irolimus ammalian target of rapamycin inhibitor espimycin HSP 90 inhibitor uoperazine ychotic, phenothiazine, calmodulin inhibitor hlorperazine Dopamine D2-receptor antagonist ylsulfathiazole potent analgesic, analogue of pethidine iostrepton tibiotic, also targeting forkhead box M1 Th An
Resveratrol Antibacterial and antifungal phytoalexin
ufexamac rug used for joint and muscular pain
roup 3
Description
hostatin A
HDAC inhibitors enprodil elective inhibitor of the NMDA receptor crynic acid oop diuretic, inhibit cysteine proteases iconazole Antifungal rtriptyline Antidepressant, TCA lvestrant ER downregulator uoperazine ychotic, phenothiazine, calmodulin inhibitor ossypol itor of antiapoptotic Bcl-2 family proteins G Inhib
Astemizole Antihistamine, histamine H1-receptor antagonistPerphenazine Antipsychotic, phenothiazines

y tests or literatures (PubMed).


an Association for Cancer Research.


the fintativelectionmostimentBy

upregdrugsS13).cruciabe dedatabtreatiNPCNPC,genes66 drchoosary TNPC

ConcTo

tive Wthe ex

genesworkdrugNPC thighlinetwagentproac

Discl

No p

Grant

GranCC1 (Cfrom MYang Mand CeHospitMedicaand C.Y

Theof pagein acco

Refe1. Tao

sis2. Lee

fecFu

3. Chofof9:3

4. SriprecroRe

5. ShnasJ C

6. Zenasig

7. FanpaCh

8. Chcar

9. Chmo

10. Jeoin p

11. Bapo

12. Leesub200

13. LamNa

Lan et al.

Mol COF12

D


al gene signature, these genes are highly represen-of NPC gene signature. However, which gene se-method (including the randomization method) is

appropriate for querying cMap needs further exper-al investigation.finding FDA-approved drug targets in the NPCulated PPI network, we obtain possible candidatefor treating NPC (Supplementary Table S12 andIt is supposed that when the drug disturbs morel nodes in the NPC PPI, the NPC network maystroyed. We further surveyed the ClinicalTrialsase. Those drugs being used in clinical trials forng several cancers whose pathway related topathogenesis may have a better chance to treatespecially those 66 drugs targeting up-bottleneckand oncogenes. There are 9 drugs (13.6%) of theugs already under NPC clinical trials. One cane the remaining 57 potential drugs (Supplement-able S14), which fit the above criteria, for futureclinical trials.

lusionsthe best of our knowledge, this is the first integra-
eb site (http://140.109.23.188:8080/NPC) to depictpression patterns of differentially expressed NPC
Recepublish

9;10:114.b J. The Connectivity Map: a new tool for biomedical research.

t Rev Cancer 2007;7:54–60.

14. Lagean

15. Wegeco

16. Denedid

17. EbnalunCa

18. Hsingtag

19. LinsegefershWa

20. Keere

21. Lufro

22. RupreAc

23. Kadacle

24. HuanNa



. This inventory provides a niche for NPC PPI net-construction, target prioritization, and potentialidentification. The interaction between prioritizedargets (e.g., cliques and bottleneck genes) and drugsghts a promising approach to address disease-relatedorks and to uncover potential new therapeutics. Validating potential drugs found by in silico ap-hes both in vitro and in vivo are required in the future.

osure of Potential Conflicts of Interest

otential conflicts of interest were disclosed.

Support

ts from National Service Center grant NSC98-3112-B-010-021-.Y. Huang) and grant NSC92-2320-B-038-057 (C.L. Chen), grantsinistry of Education, Aim for the Top University Plan (Nationaling University), Taipei Veterans General Hospital (V99ER2-011)nter of Excellence for Cancer Research at Taipei Veterans Generalal (DOH99-TD-C-111-007; C.Y. Huang), and a grant from Taipeil University-Center of Excellence for Cancer Research (C.L. Chen. Huang).costs of publication of this article were defrayed in part by the paymentcharges. This articlemust therefore be herebymarked advertisementrdance with 18 U.S.C. Section 1734 solely to indicate this fact.

ived 10/25/2009; revised 06/02/2010; accepted 06/28/2010;ed OnlineFirst 08/17/2010.

rencesQ, Chan AT. Nasopharyngeal carcinoma: molecular pathogene-

and therapeutic developments. Expert Rev Mol Med 2007;9:1–24.YC, Hwang YC, Chen KC, et al. Effect of Epstein-Barr virus in-

tion on global gene expression in nasopharyngeal carcinoma.nct Integr Genomics 2007;7:79–93.en X, Liang S, Zheng W, Liao Z, Shang T, Ma W. Meta-analysisnasopharyngeal carcinoma microarray data explores mechanismEBV-regulated neoplastic transformation. BMC Genomics 2008;22.uranpong V, Mutirangura A, Gillespie JW, et al. Global gene ex-ssion profile of nasopharyngeal carcinoma by laser capture mi-dissection and complementary DNA microarrays. Clin Cancers 2004;10:4944–58.i W, Bastianutto C, Li A, et al. Multiple dysregulated pathways inopharyngeal carcinoma revealed by gene expression profiling. Intancer 2006;119:2467–75.ng ZY, Zhou YH, Zhang WL, et al. Gene expression profiling ofsopharyngeal carcinoma reveals the abnormally regulated Wntnaling pathway. Hum Pathol 38:120–33.g W, Li X, Jiang Q, et al. Transcriptional patterns, biomarkers andthways characterizing nasopharyngeal carcinoma of Southernina. J Transl Med 2008;6:32.ang ET, Adami HO. The enigmatic epidemiology of nasopharyngealcinoma. Cancer Epidemiol Biomarkers Prev 2006;15:1765–77.ou J, Lin YC, Kim J, et al. Nasopharyngeal carcinoma-review of thelecular mechanisms of tumorigenesis. Head Neck 2008;30:946–63.ng H, Mason SP, Barabasi AL, Oltvai ZN. Lethality and centralityrotein networks. Nature 2001;411:41–2.tada NN, Hurst LD, Tyers M. Evolutionary and physiological im-rtance of hub proteins. PLoS Comput Biol 2006;2:e88.SA, Chan CH, Chen TC, et al. POINeT: protein interactome with-network analysis and hub prioritization. BMC Bioinformatics

mb J, Crawford ED, Peck D, et al. The Connectivity Map: usingne-expression signatures to connect small molecules, genes,d disease. Science 2006;313:1929–35.i G, Twomey D, Lamb J, et al. Gene expression-based chemicalnomics identifies rapamycin as a modulator of MCL1 and gluco-rticoid resistance. Cancer Cell 2006;10:331–42.Preter K, De Brouwer S, Van Maerken T, et al. Meta-mining of

uroblastoma and neuroblast gene expression profiles reveals can-ate therapeutic compounds. Clin Cancer Res 2009;15:3690–6.i H, Tomida S, Takeuchi T, et al. Relationship of deregulated sig-ling converging onto mTOR with prognosis and classification ofg adenocarcinoma shown by two independent in silico analyses.ncer Res 2009;69:4027–35.u CN, Chang YM, Kuo CJ, Lin YS, Huang HS, Chung IF. Integrat-high dimensional bi-directional parsing models for gene mentionging. Bioinformatics 2008;24:i286–94.KT, Liu CH, Chiou JJ, Tseng WH, Lin KL, Hsu CN. Gene name

rvice: no-nonsense alias resolution service for homo sapiensnes. Proceedings of the 2007 IEEE/WIC/ACM International Con-ence on Web Intelligence and Intelligent Agent Technology Work-ops (WI-IAT Workshops 2007); 2007 Nov 5; Silicon Valley.shington (DC): IEEE; 2007.shava Prasad TS, Goel R, Kandasamy K, et al. Human protein ref-nce database-2009 update. Nucleic Acids Res 2009;37:D767–72.c PV, Tempst P. PINdb: a database of nuclear protein complexesm human and yeast. Bioinformatics 2004;20:1413–5.epp A, Brauner B, Dunger-Kaltenbach I, et al. CORUM: the com-hensive resource of mammalian protein complexes. Nucleicids Res 2008;36:D646–50.mburov A, Wierling C, Lehrach H, Herwig R. ConsensusPathDB-atabase for integrating human functional interaction networks. Nu-ic Acids Res 2009;37:D623–8.
ang da W, Sherman BT, Lempicki RA. Systematic and integrativealysis of large gene lists using DAVID bioinformatics resources.t Protoc 2009;4:44–57.



25. Kuinte20

26. WidruD9

27. Linlish

28. Liamo(NPGe

29. Chbriwa

30. Splec

31. FroCewit

32. EllifacDN

33. Leeteinde

34. UhhumSc

35. Wacomand

36. HaPronunucJ B

37. BomaCe

38. Leo

torgo

39. Pacagroge

40. Magioanryn

41. Basiofor

42. Yaevice

43. BodoJ P

44. WeIn:Sc

45. LiviruLe

46. Ogchno

47. HoimHe

48. Rebaexp

49. Gilintum

50. Zh


www.a

D


hn M, von Mering C, Campillos M, Jensen LJ, Bork P. STITCH:raction networks of chemicals and proteins. Nucleic Acids Res

08;36:D684–8.shart DS, Knox C, Guo AC, et al. DrugBank: a knowledgebase forgs, drug actions and drug targets. Nucleic Acids Res 2008;36:01–6.CT, ChanWY,ChenW, et al. Characterization of seven newly estab-ed nasopharyngeal carcinoma cell lines. Lab Invest 1993;68:716–27.o SK, Perng YP, Shen YC, Chung PJ, Chang YS, Wang CH. Chro-somal abnormalities of a new nasopharyngeal carcinoma cell lineC-BM1) derived from a bone marrow metastatic lesion. Cancernet Cytogenet 1998;103:52–8.en TC, Lee SA, Chan CH, et al. Cliques in mitotic spindle networkng kinetochore-associated complexes to form dependence path-y. Proteomics 2009;9:4048–62.irin V, Mirny LA. Protein complexes and functional modules in mo-ular networks. Proc Natl Acad Sci U S A 2003;100:12123–8.uin I, Montecucco A, Biamonti G, Hubscher U, Spadari S, Maga G.ll cycle-dependent dynamic association of cyclin/Cdk complexesh human DNA replication proteins. EMBO J 2002;21:2485–95.son V, Stillman B. Reconstitution of recombinant human replicationtor C (RFC) and identification of an RFC subcomplex possessingA-dependent ATPase activity. J Biol Chem 1998;273:5979–87.SH, Kwong AD, Pan ZQ, Hurwitz J. Studies on the activator 1 pro-complex, an accessory factor for proliferating cell nuclear antigen-

pendent DNA polymerase δ. J Biol Chem 1991;266:594–602.lmann F, Cai J, Flores-Rozas H, et al. In vitro reconstitution ofan replication factor C from its five subunits. Proc Natl Acad

i U S A 1996;93:6521–6.ng Y, Cortez D, Yazdi P, Neff N, Elledge SJ, Qin J. BASC, a superplex of BRCA1-associated proteins involved in the recognitionrepair of aberrant DNA structures. Genes Dev 2000;14:927–39.

yano T, YanagidaM, Yamauchi Y, Shinkawa T, Isobe T, Takahashi N.teomic analysis of humanNop56p-associated pre-ribosomal ribo-cleoprotein complexes. Possible link between Nop56p and theleolar protein treacle responsible for Treacher Collins syndrome.iol Chem 2003;278:34309–19.uwmeester T, Bauch A, Ruffner H, et al. A physical and functional
p of the human TNF-α/NF-κ B signal transduction pathway. Natll Biol 2004;6:97–105.ng JL, Loh KS, Putti TC, Goh BC, Tan LK. Epidermal growth fac-
lifeinfleu

acrjournals.org


receptor in undifferentiated carcinoma of the nasopharynx. Laryn-scope 2004;114:153–7.n J, Kong L, Lin S, Chen G, Chen Q, Lu JJ. The clinical signifi-nce of coexpression of cyclooxygenases-2, vascular endothelialwth factors, and epidermal growth factor receptor in nasopharyn-al carcinoma. Laryngoscope 2008;118:1970–5.BB, Poon TC, To KF, et al. Prognostic significance of tumor an-genesis, Ki 67, p53 oncoprotein, epidermal growth factor receptord HER2 receptor protein expression in undifferentiated nasopha-geal carcinoma-a prospective study. Head Neck 2003;25:864–72.r-Sela G, Kuten A, Ben-Eliezer S, Gov-Ari E, Ben-Izhak O. Expres-n of HER2 and C-KIT in nasopharyngeal carcinoma: implicationsa new therapeutic approach. Mod Pathol 2003;16:1035–40.n J, Fang Y, Huang BJ, Liang QW, Wu QL, Zeng YX. Absence ofdence for HER2 amplification in nasopharyngeal carcinoma. Can-r Genet Cytogenet 2002;132:116–9.se S, Yap LF, Fung M, et al. The ATM tumour suppressor gene iswn-regulated in EBV-associated nasopharyngeal carcinoma.athol 2009;217:345–52.inberg RA. P53 and apoptosis: master guardian and executioner.Weinberg RA, editor. The Biology of Cancer. New York: Garlandience; 2007, p. 307–56.L, Guo L, Tao Y, et al. Latent membrane protein 1 of Epstein-Barrs regulates p53 phosphorylation through MAP kinases. Cancer

tt 2007;255:219–31.ino T, Moriai S, Ishida Y, et al. Association of immunoescape me-anisms with Epstein-Barr virus infection in nasopharyngeal carci-ma. Int J Cancer 2007;120:2401–10.SY, Guo HR, Chen HH, Hsiao JR, Jin YT, Tsai ST. Prognostic

plications of Fas-ligand expression in nasopharyngeal carcinoma.ad Neck 2004;26:977–83.imand J, Kull M, Peterson H, Hansen J, Vilo J. g: Profiler-a web-sed toolset for functional profiling of gene lists from large-scaleeriments. Nucleic Acids Res 2007;35:W193–200.-Ad I, Shtaif B, Levkovitz Y, et al. Phenothiazines induce apoptosisa B16 mouse melanoma cell line and attenuate in vivo melanomaor growth. Oncol Rep 2006;15:107–12.

elev Z, Ohba H, Bakalova R, et al. Phenothiazines suppress pro-ration and induce apoptosis in cultured leukemic cells without any
luence on the viability of normal lymphocytes. Phenothiazines andkemia. Cancer Chemother Pharmacol 2004;53:267–75.



Published OnlineFirst August 17, 2010.Mol Cancer Ther Ming-Ying Lan, Chi-Long Chen, Kuan-Ting Lin, et al. Treatment StrategyFrom NPC Therapeutic Target Identification to Potential

Updated version

10.1158/1535-7163.MCT-09-0966doi:

Access the most recent version of this article at:

Material

Supplementary

http://mct.aacrjournals.org/content/suppl/2010/08/17/1535-7163.MCT-09-0966.DC1

Access the most recent supplemental material at:

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

Subscriptions

Reprints and

[email protected] at

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

Permissions

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

.http://mct.aacrjournals.org/content/early/2010/08/13/1535-7163.MCT-09-0966To request permission to re-use all or part of this article, use this link

on March 28, 2020. © 2010 American Association for Cancer Research. mct.aacrjournals.org Downloaded from


http://mct.aacrjournals.org/lookup/doi/10.1158/1535-7163.MCT-09-0966

http://mct.aacrjournals.org/content/suppl/2010/08/17/1535-7163.MCT-09-0966.DC1

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

mailto:[email protected]

http://mct.aacrjournals.org/content/early/2010/08/13/1535-7163.MCT-09-0966


therapeutic discovery molecular cancer therapeutics from ... · connected subgraphs, in the...

Documents