Integrated MicroRNA Network Analyses Identify a Poor ... · Integrated MicroRNA Network Analyses Identify a Poor-Prognosis € Updated version € ... We used integrated analysis

2014;20:878-889. Published OnlineFirst December 18, 2013.Clin Cancer Res Fengju Song, Da Yang, Ben Liu, et al. Subtype of Gastric Cancer Characterized by the miR-200 FamilyIntegrated MicroRNA Network Analyses Identify a Poor-Prognosis

Updated version

10.1158/1078-0432.CCR-13-1844doi:

Access the most recent version of this article at:

Material

Supplementary

http://clincancerres.aacrjournals.org/content/suppl/2013/12/18/1078-0432.CCR-13-1844.DC1.html

Access the most recent supplemental material at:

Cited Articles

http://clincancerres.aacrjournals.org/content/20/4/878.full.html#ref-list-1

This article cites by 43 articles, 15 of which you can access for free at:

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

Subscriptions

Reprints and

[email protected]

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

Permissions

[email protected]

To request permission to re-use all or part of this article, contact the AACR Publications Department at

on February 19, 2014. © 2014 American Association for Cancer Research. clincancerres.aacrjournals.org Downloaded from

Published OnlineFirst December 18, 2013; DOI: 10.1158/1078-0432.CCR-13-1844



http://clincancerres.aacrjournals.org/lookup/doi/10.1158/1078-0432.CCR-13-1844

http://clincancerres.aacrjournals.org/lookup/doi/10.1158/1078-0432.CCR-13-1844





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

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

mailto:[email protected]




http://clincancerres.aacrjournals.org/




Human Cancer Biology

Integrated MicroRNA Network Analyses Identify aPoor-Prognosis Subtype of Gastric Cancer Characterizedby the miR-200 Family

Fengju Song1, Da Yang6, Ben Liu1, Yan Guo1, Hong Zheng1, Lian Li1, Tao Wang5, Jinpu Yu2,4, Yanrui Zhao1,Ruifang Niu1, Han Liang3, Hans Winkler7, Wei Zhang6, Xishan Hao3, and Kexin Chen1

AbstractPurpose: Our aim was to investigate whether microRNAs can predict the clinical outcome of patients

with gastric cancer. We used integrated analysis of microRNA and mRNA expression profiles to identify

gastric cancer microRNA subtypes and their underlying regulatory scenarios.

Experimental Design:MicroRNA-based gastric cancer subtypes were identified by consensus clustering

analysis ofmicroRNAprofiles of 90 gastric cancer tissues. Activated pathways in the subtypes were identified

by gene expression profiles. Further integrated analysis was conducted to model a microRNA regulatory

network for each subtype. RNA and protein expression were analyzed by RT-PCR and tissue microarray,

respectively, in a cohort of 385 gastric cancer cases (including the 90 cases for profiling) to validate the key

microRNAs and targets in the network. Both in vitro and in vivo experiments were carried out to further

validate the findings.

Results: MicroRNA profiles of 90 gastric cancer cases identified two microRNA subtypes significantly

associated with survival. The poor-prognosis gastric cancer microRNA subtype was characterized by

overexpression of epithelial-to-mesenchymal transition (EMT) markers. This gastric cancer "mesenchymal

subtype" was further validated in a patient cohort comprising 385 cases. Integrated analysis identified a key

microRNA regulatory network likely driving the gastric cancer mesenchymal subtype. Three of the

microRNAs (miR-200c, miR-200b, and miR-125b) targeting the most genes in the network were signif-

icantly associated with survival. Functional experiments demonstrated that miR-200b suppressed ZEB1,

augmented E-cadherin, inhibited cell migration, and suppressed tumor growth in a mouse model.

Conclusions:We have uncovered a key microRNA regulatory network that defines the mesenchymal

gastric cancer subtype significantly associated with poor overall survival in gastric cancer. Clin Cancer

Res; 20(4); 878–89. �2013 AACR.

IntroductionGastric cancer is a highly aggressive and life-threatening

malignancy. It is the second leading cause of cancer-relateddeaths worldwide, accounting for nearly 10% of all cancerdeaths. More than half of the gastric cancer–related deathsoccur in East Asia, mainly in China (1). The prognosis forpatients with gastric cancer is heterogeneous, and the 5-yearoverall survival rate is only approximately 20% (2). Surgeryis the mainstay of treatment, but the results are oftendisappointing. The lack of successful treatment strategieshas led researchers to comprehensively measure genomicand epigenomic abnormalities of gastric tumors to identifygastric cancer microRNA subtypes and their underlyingregulatory scenarios (3).

Accumulated evidence shows thatmicroRNAsplay impor-tant roles in gastric cancer development and progression (4).MicroRNA expression patterns can be especially rich inbiologic information, as variations in expression of hun-dreds of protein-coding genesmay, to an extent, be capturedin the expression patterns of one or a few microRNAs that

Authors' Affiliations: Departments of 1Epidemiology and Biostatistics,2Immunology, and 3Gastric Cancer, 4TMUCIH-J&J Joint Laboratory,Key Laboratory of Cancer Prevention and Therapy, Tianjin, NationalClinical Research Center of Cancer, Tianjin Medical University CancerInstitute and Hospital; 5Department of Gastroenterology, Tianjin MedicalUniversity General Hospital, Tianjin, PR China; 6Department of Pathol-ogy, The University of Texas MD Anderson Cancer Center, Houston,Texas; and 7Janssen Research and Development, a Division of JanssenPharmaceutica, Beerse, Belgium

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

F. Song, D. Yang, and B. Liu contributed equally to this work.

Current address: Hans Winkler, GNS Healthcare, 58 Charles Street, Cam-bridge, MA 02141.

Corresponding Authors: Kexin Chen, Department of Epidemiology andBiostatistics, Tianjin Medical University Cancer Hospital and Institute,Tianjin, PR China 300060. Phone: 86-0-2223372231; Fax: 86-0-2223372231; E-mail: [email protected]; andWei Zhang,Depart-ment of Pathology, Unit 85, The University of Texas MD Anderson CancerCenter, 1515 Holcombe Blvd., Houston, TX 77030. Phone: 713-745-1103;Fax: 713-792-5549; E-mail: [email protected]

doi: 10.1158/1078-0432.CCR-13-1844

�2013 American Association for Cancer Research.

ClinicalCancer

Research

Clin Cancer Res; 20(4) February 15, 2014878





regulate them (5, 6). Although findings from microRNAprofiling studies are promising, they have limitations inelucidating microRNA function and identifying interactionsbetween microRNAs and targeted mRNAs. Simultaneousprofiling of the expression patterns of mRNAs and micro-RNAs in the same panel of patients with cancer has beenshown to be a highly integrative and reproducible way ofdissecting the molecular basis of human cancer (7, 8). Suchstudies have the potential not only to identify microRNAsthat are independent prognostic factors but also to improveour understanding of gene regulation and systems-levelmodeling. However, integrated analysis of microRNA andmRNA global expression profiles has yet to be explored inprognostic studies of gastric cancer.In this study, we analyzed microRNA and mRNA expres-

sionprofiles in aChinese gastric cancer cohort.Our purposewas to investigate whether microRNAs could predict theclinical outcome of patients with gastric cancer and thushave potential as prognostic markers. To identify candidatemicroRNA-regulated networks of gene expression that maybe involved in gastric cancer survival, we integrated thesemicroRNA expression profiles with mRNA gene expressiondata we obtained from the same samples. Our findingssuggest that certain microRNA regulatory pathways mayhave potential as both clinical biomarkers and therapeutictargets for gastric cancer.

Materials and MethodsStudy design and patient samplesThe study was conducted in 2 phases. In the first phase,

global microRNA and mRNA expression profiling for 90gastric cancer tissues and 10 adjacent normal tissues wereobtained throughmicroarray analysis. In the second phase,the candidate microRNAs and targets identified in the firstphase were validated and evaluated for their potential asbiomarkers of gastric cancer survival in 385 gastric cancercases (including the 90 cases in the first phase) from TianjinMedical University Cancer Institute and Hospital (Tianjin,PR China). All the patients were randomly selected and hadhistologically confirmed gastric cancer diagnosed between2001 and 2009 at the Tianjin Medical University CancerInstitute andHospital. Patients from this cohort were askedto complete a follow-up questionnaire annually withupdated information on their disease progression. Morethan 90% of the study participants had completed andreturned every questionnaire they received during the study

period. The study was approved by the Institutional ReviewBoard of Tianjin Medical University; informed consent wasobtained from all patients.

MicroRNA expression profiling. GeneChip microRNAarrays (Affymetrix) containing 2,202 probe sets unique topre-microRNA were analyzed according to Affymetrix pro-tocols.Microarray processing procedureswere conducted asdescribed in the Affymetrix Gene-Chip Expression AnalysisManual.

mRNA expression profiling. Genechip HT HG-U133þPM 96-array plates from Affymetrix, containing probe setsfor more than 47,000 transcripts, were analyzed accordingto Affymetrix protocols. Sample labeling and processing,GeneChip hybridization, and scanning were conductedusing the GeneTitan Instrument (Affymetrix) as the proto-col described. Total RNAwas isolated from liquid nitrogen–frozen gastric cancer tissues (n ¼ 90) and normal adjacenttissues (n ¼ 10). The total RNA was extracted and purifiedwith TRIzol reagent (Invitrogen) and ethanol precipitationaccording to the instructions of the manufacturer. RNAquality and concentration were determined by Nano-Drop-8000.

Statistical analysis. In the profiling phase, cluster anal-yses were conducted to look for natural groupings in themicroRNA and mRNA expression profiles. Consensus clus-tering was conducted as in previous studies (9, 10). Increas-ing values ofK (2 through 6, inclusive) were used to identifyoptimal segregation. For each K, 1,000 random iterationswere conducted to characterize the clusters. The Benjamini–Hochberg correction was used to estimate the false discov-ery rate when multiple testing was applied. Consensusk-mean clustering (11) of the 90 tumor samples identified2 robust clusters with clustering stability decreasing for k ¼2–6 (Supplementary Fig. S1). Cluster significance was eval-uated using SigClust (12) with 1,000 times simulation. Theclass boundary was statistically significant (P < 10�16).

To validate the association between gastric cancer survivaland expression of the candidate microRNAs and epithelial-to-mesenchymal transition (EMT) markers, the correlationof the expressionof candidatemicroRNAsbymicroarray andby quantitative RT-PCR (qRT-PCR) analysis was determinedby the Spearman rank test and was statistically significant.RepresentativeqRT-PCR results are shown inSupplementaryFig. S2. For survival analysis, we used univariate and mul-tivariate Cox proportional hazards models to estimate theHR between patients with high expression and those withlow expression of candidate microRNAs and EMT markers.Variables included in the multivariate model were patients’sex, age, smoking status, and alcohol consumption anddisease characteristics, including pathologic type, differen-tiation, location, stage, and treatment. The Kaplan–Meiermethod was used to estimate the survival curves.

The following approach was used for separation of thepatients into 2 groups according to relative expression levelsof candidatemicroRNAs and EMTmarkers. FormicroRNAs,the lowest quintile values of the expression data were usedas the cutoffs. For the EMT markers, values around themedian expression were used as the cutoffs. Survival was

Translational RelevanceOur observations on the role of themiR-200 family in

regulating epithelial-to-mesenchymal transition (EMT)enhance our understanding of the microRNA regulatorypathways influencing the clinical progression and prog-nosis of gastric cancer, potentially opening up a newavenue for therapeutic intervention in patients withlocalized primary gastric cancer.

MicroRNA Network Predicts Gastric Cancer Survival

www.aacrjournals.org Clin Cancer Res; 20(4) February 15, 2014 879





defined as the interval from the date of diagnosis until dateof death from gastric cancer, date of death from other cause,or the end of follow-up (May 31, 2012), whichever camefirst. Patients lost to follow-up were censored at the date oflast follow-up contact. Statistical analyses were conductedusing R 2.10.0 (R Foundation). All P values were 2-tailedand are reported as significant when P < 0.05.

ResultsClinical characteristics of gastric cancer patients

A total of 385 patients with pathologically confirmedgastric cancerwere included in this study. Their demograph-ic and clinical characteristics are summarized in Supple-mentary Table S1. The male:female ratio was 2.7:1. Themean age of the participants at diagnosis was 60.5 � 9.3years. Themedian follow-up intervalwas 35months (range,1–112 months), and 180 patients died of gastric cancerduring this period.

Identification of two gastric cancer subtypes withdistinct prognoses

To identify microRNA subtypes of gastric cancer, consen-sus clustering was applied to the microRNA expression

profile of 90 gastric tumors, on the basis of themost variable50% of microRNAs across all samples. The analysis iden-tified 2 clusters with distinct microRNA expression patterns(Fig. 1A). Cluster 1 comprised 31 gastric cancer cases thatoverexpressed 43 microRNAs. Cluster 2 comprised 59 gas-tric cancer cases that overexpressed 54microRNAs. Survivalanalysis revealed that patients in cluster 1 had significantlyshorter overall survival and progression-free survival thanthose in cluster 2 (P¼0.050 andP¼0.022, respectively; Fig.1B and C). These microRNA subtypes remained strongpredictors of survival in amultivariateCox regressionmodelthat included sex, age, disease grade, and metastasis status(yes or no; P¼ 0.015 and P¼ 0.006 for overall survival andprogression-free survival, respectively).

Functional characterization of the two gastric cancersubtypes

To determine whether the 2 gastric cancer subtypes werefunctionally distinct, we identified signature genes andpathways that were specifically altered in each subtype.Using the genome-wide protein-coding gene expressiondata on the 90 tumors, we identified 1,245 and 965 signa-ture genes for clusters 1 and 2, respectively (Fig. 2A).

1.0Cluster 2 (n = 59)Cluster 1 (n = 31)

A B

C

0.8

0.6

0.4

0.2

0.0

1.0

0.8

0.6

0.4

0.2

0.0

0 12 24 36 48 60

Survival time (months)

0 12 24 36 48 60

Progression-free time (months)O

vera

ll surv

ival

Pro

gre

ssio

n-f

ree s

urv

ival

+ Censored

Logrank P = 0.050

+ Censored

Logrank P = 0.022

Cluster

Cluster 1 Cluster 2

Cluster

Cluster 1 Cluster 2

Figure 1. Consensus clustering identifies 2 gastric cancer subtypes with distinct microRNA profiles. A, analysis of microRNA profile in 90 gastric cancer casesidentified 2 clusters (31 cases in cluster 1 and 59 cases in cluster 2) with distinct microRNA expression patterns. Red color represents high expressionand green color represents lowexpression. Not all themicroRNAs in the figurewere labeled. B, Kaplan–Meier curves for overall survival of patientswith gastriccancer in clusters 1 and 2; the solid line represents cluster 1, and the dashed line represents cluster 2. C, Kaplan–Meier curves for progression-free survival ofpatients with gastric cancer in clusters 1 and 2; the solid line represents cluster 1, and the dashed line represents cluster 2.

Song et al.

Clin Cancer Res; 20(4) February 15, 2014 Clinical Cancer Research880





Pathway analysis of the signature genes showed that mes-enchymal phenotype–related pathways, including EMT,regulation of EMT, and regulation of mesenchymal cellproliferation,were activated in cluster 1, the poor-prognosissubtype (Fig. 2B). The biosynthetic- and metabolic-relatedpathways were upregulated in cluster 2, the favorable-prog-nosis subtype (Fig. 2B). Specific investigation of the mes-enchymal and epithelial markers in the 2 subtypes showedthat mesenchymal markers, such as N-cadherin, vimentin,ZEB1, ZEB2, and Slug, were significantly upregulated incluster 1 compared with cluster 2 (P < 0.001, Fig. 2C).Epithelial markers, such as E-cadherin and cytokeratin, weresignificantly downregulated in cluster 1 (P < 0.001, Fig. 2C).Our observations in protein-coding gene and microRNA

profiles suggested that clusters 1 and 2 were 2 gastric cancersubtypeswith distinctmolecular and clinical characteristics.We thus named cluster 1 the mesenchymal subtype andcluster 2 the epithelial subtype.Identification of key microRNAs regulating the mesenchy-

mal and epithelial subtypes. To predict candidate keymicroRNAs that play driving roles in the mesenchymal andepithelial subtypes, the MIRACLE algorithm (13) was usedto identify microRNAs whose expression was significantly

upregulated in one subtype compared with the other sub-type and normal tissue (Supplementary Methods). Thisanalysis revealed 24 key microRNAs for the mesenchymalsubtype and 15 key microRNAs for the epithelial subtype.We next integrated the microRNA and protein-coding geneexpression data to predict the potential targets for eachmicroRNA. These analyses revealed 19microRNAs targeting269 genes for themesenchymal subtype and 10microRNAstargeting 288 genes for the epithelial subtype. Among the 39keymicroRNAs identified inour analyses, 10were predictedto regulate 79.2% (411 of 557) of all targets. Besides havingbinding sites on the 30-untranslated regions (UTR) of theirpredicted targets, expression levels of these 10 microRNAswere inversely correlated with the expression levels of theirpredicted targets.

Three key microRNAs associated with gastric cancer sur-vival. Among the 10 keymicroRNAswith themost targets,6 showed significant upregulation in the mesenchymalsubtype compared with both the epithelial subtype andnormal tissues (Fig. 3A). Specifically, miR-125b was upre-gulated by more than 4-fold in the mesenchymal subtype,and its overexpression was significantly associated withpoor prognosis (P ¼ 0.01). Among the 4 microRNAs

Figure 2. Cluster and pathwayanalyses of mRNA profile dataidentify distinct functionalcharacteristics of the 2 clusters.A, clusters 1 and 2 showed distinctmRNA expression patterns. Thered color represents highexpression and the green colorrepresents low expression.B, functional pathways wereconstructed for each cluster, anddifferences between the 2 clustersfor each pathway were analyzed.C, the mesenchymal markers N-cadherin, vimentin (VIM), ZEB1,ZEB2, and Slug were significantlyupregulated in cluster 1 comparedwith cluster 2. The epithelialmarkers E-cadherin andcytokeratin were significantlydownregulated in cluster 1compared with cluster 2.







downregulated in the mesenchymal subtype, 3 (miR-200a,miR-200b, and miR-200c) belong to the miR-200 family(Fig. 3A). In our analysis, miR-200a (P ¼ 0.05) and miR-200b (P ¼ 0.02) were both associated with good gastriccancer prognosis and were predicted to target ZEB1/2 andother targets (Fig. 3B and C). Detailed information aboutthe keymicroRNA identification can be seen in Supplemen-tary Table S2.

Validation of themesenchymal and epithelial subtypes in anindependent population. We identified an independentdataset, a genome-wide gene expression profile compris-ing 200 gastric cancer cases from Singapore, with whichto evaluate the validity of the mesenchymal and epithe-lial subtypes and the microRNA regulatory network.Consensus clustering using the 164 genes in our micro-RNA regulatory network segregated the 200 gastric can-cers into 67 mesenchymal cases and 133 epithelial cases.Consistently with our previous observation, the mesen-chymal cases had significantly shorter progression-freesurvival (P ¼ 0.02) than the epithelial cases (Supple-mentary Fig. S3).

Validation of the association between keymicroRNA expres-sion and gastric cancer survival. We further validated theassociation between the expression of key microRNAs andgastric cancer prognosis among the 385 gastric cancer casesfrom the Tianjin Medical University Cancer Institute andHospital. On the basis of their association with survival inthe first phase of the analysis, 3 microRNAs (miR-200a,miR-200b, and miR-125b) were selected for validation.miR-200c was also selected because it is a member of themiR-200 family. Among these 4 microRNAs, 3 were signif-icantly associated with gastric cancer survival. Interestingly,the associations of miR-200a and miR-200b with survivalwere significant only inwomen: womenwith higher expres-sion of either miR-200a or miR-200b had a more favorableprognosis (P ¼ 0.027 and P ¼ 0.048, respectively; Supple-mentary Fig. S4A and S4B). The association of miR-125bwith gastric cancer survival was significant overall: patientswith higher miR-125b expression had poor prognosis (P ¼0.005). Again, however, the association was significant inwomen (P ¼ 0.002) but not in men (P ¼ 0.1348; Supple-mentary Fig. S4C). The associations between the expression

miR-200aA B

C

Exp

ressio

nE

xp

ressio

nE

xp

ressio

nE

xp

ressio

n

Exp

ressio

nE

xp

ressio

nE

xp

ressio

nE

xp

ressio

n

Exp

ressio

nE

xp

ressio

n

miR-200b

miR-200c

Cluster1 Cluster2 Normal Cluster1 Cluster2 Normal

Cluster1 Cluster2 NormalCluster1 Cluster2 Normal Cluster1 Cluster2 Normal

miR-194

miR-27b miR-27a miR-125b

miR-125a-5p miR-199a-5p miR-130a

Number of targets

150 100 50 0OS

miR-200c

miR-200b

miR-125b

miR-27a

miR-27b

miR-125a-5p

miR-200a

miR-194

miR-199a-5p

miR-130a

P < 0.01

P < 0.05

Poor prognosis

Good prognosis

Figure 3. Integrated analysis identifies key microRNAs that play driving roles in the mesenchymal and epithelial subtypes. A, expression of the 10 keymicroRNAs was compared among the mesenchymal subtype, the epithelial subtype, and the adjacent normal tissue. Four microRNAs showedsignificantly lower expression in the mesenchymal subtype, whereas 6 microRNAs showed significantly higher expression in the mesenchymal subtype.B, integrated analysis revealed the functional targets of the 10 key microRNAs. C, the 10 key microRNAs were ranked by number of targets; 3 microRNAsshowed significant association with gastric cancer survival.

Song et al.






ofmiR-200c and gastric cancer survival were not statisticallysignificant. Detailed results on the associations between theexpression of the 4 key microRNAs and overall survival andprogression-free survival of gastric cancer are shown inSupplementary Table S3.Validation of the association between expression of EMT

markers and gastric cancer survival. To further evaluate the

relationship between expression of EMTmarkers and gastriccancer survival, we conducted immunohistochemical anal-ysis for 11 EMT markers in 364 gastric tumor tissues assem-bled on a tissue microarray. Representative cases are shownin Fig. 4A and B. Among the 11 EMT markers, 5 wereassociated with gastric cancer survival. Expression of E-cad-herin, cytokeratin, or b-catenin was significantly associated

+ Censored

+ Censored

log rank P < 0.0001

log rank P < 0.0148

+ Censored

log rank P < 0.0467

+ Censored

log rank P = 0.0627

+ Censored

log rank P = 0.0466

+ Censored

log rank P = 0.0403

E_cadherin High Low

cytokeratin High Low

b_catenin High Low

N_cadherin High Low

Hightwist2 Low

HighZEB1 Low

0 12 24 36 48 60

0 12 24 36 48 60

0 12 24 36 48 60

0 12 24 36 48 60

0 12 24 36 48 60

0 12 24 36 48 60

1.0

0.8

0.6

0.4

0.2

0.0

1.0

0.8

0.6

0.4

0.2

0.0

1.0

0.8

0.6

0.4

0.2

0.0

1.0

0.8

0.6

0.4

0.2

0.0

1.0

0.8

0.6

0.4

0.2

0.0

1.0

0.8

0.6

0.4

0.2

0.0

Survival time (months)

Surv

ival pro

babili

tyS

urv

ival pro

babili

tyS

urv

ival pro

babili

tyS

urv

ival pro

babili

tyS

urv

ival pro

babili

tyS

urv

ival pro

babili

ty

Case 2Case 1A CB

200 mm 50 mm 200 mm 50 mm

H&

EE

-cadheri

nC

yto

kera

tins

b-cate

nin

N-c

adheri

nTw

ist2

ZE

B1

Figure 4. Representative cases and Kaplan–Meier curves of patients with gastric cancer with low expression versus high expression of EMTmarkers. A and B,immunohistochemical analysis on consecutive tissuemicroarray slides of gastric cancer tissues showed different expression of EMTmarkers in patients withgastric cancer. Representative case 1 had low expression of E-cadherin, cytokeratin, and b-catenin and high expression of N-cadherin, Twist2, andZEB1 (A, scale bars, 200 and 50mm, respectively). Representative case 2 had high expression of E-cadherin, cytokeratin, and b-catenin and low expression ofN-cadherin, Twist2, and ZEB1 (B, scale bars, 200 and 50 mm, respectively). C, expression of E-cadherin, cytokeratin, or b-catenin was associated withlonger survival (log-rank test). Expression of ZEB1 or Twist2 was associated with shorter survival (log-rank test). Expression of N-cadherin was borderlineassociated with gastric cancer survival.







Figure 5. Overexpression of miR-200b in gastric cancer cells induces epithelial phenotype. A, changes in microRNA and mRNA levels in MGC-803and SGC-7901 cells transfected with miR-200b or control miRNA (miR-Ctrl) as measured by real-time RT-PCR (TaqMan). Two independent timecourse experiments were carried out; the average � SE (indicated by the error bars) of the 2 experiments are shown. B, MTT assay in MGC-803 andSGC-7901 cells transfected with miR-200b or miR-Ctrl. (Continued on the following page.)

Song et al.






with longer survival (P < 0.0001, P ¼ 0.0148, and P ¼0.0467, respectively, log-rank test; Fig. 4C). Expression ofZEB1 or Twist2 was associated with poor survival (P ¼0.0405 and P ¼ 0.0466, respectively, log-rank test; Fig.4C). Expression of N-cadherin was borderline associatedwith gastric cancer survival (P¼ 0.0627, log-rank test; Fig.4C). Expression of vimentin was associated with poorprogression-free survival of gastric cancer (SupplementaryTable S4). The associations between expression of Twist1,Sip1, Slug, or Snail and gastric cancer survival were notstatistically significant. Details of the associations betweenthe 11 EMT markers and overall survival and progression-free survival of gastric cancer are shown in SupplementaryTable S4. Tumors with low E-cadherin expression exhib-ited a more mesenchymal phenotype, with elongatedtumor cells and looser connections between tumor cells,whereas those with high E-cadherin expression exhibitedmore of an epithelial phenotype, such as a papillarystructure that was covered by the typical cobblestonemorphologic characteristics of epithelial cells (Fig. 4Aand B).miR-200b promoted the epithelial phenotype in vitro. To

determine whether forced expression of miR-200b canpromote the epithelial phenotype, we transfected gastriccancer cells MGC-803 and SGC-7901 with either miR-200bmimic (miR-200b) or a scrambled negative microRNAcontrol (miR-Ctrl). miR-200b overexpression significantlyincreased the expression of the epithelialmarker E-cadherinin both cell lines (Fig. 5A). In addition, the growth-inhib-itory effect of miR-200b was detected by MTT assay (Fig.5B). These results suggested that cells overexpressing miR-200b gained an epithelial signature characterized by induc-tion of E-cadherin expression and suppression of mesen-chymal markers.To further confirm these results, we conducted immu-

nofluorescence staining to directly visualize the effect ofmiR-200b on E-cadherin expression, localization, and cellmorphology. As shown in Fig. 5C (left), miR-200b–trans-fected MGC-803 and SGC-7901 cells showed epithelialcell features, characterized by aggregated cells with typicalcobblestone structure; immunofluorescence stainingrevealed that E-cadherin protein was localized on themembrane at cell–cell junctions, indicative of epithelialcells (Fig. 5C, left). In addition, F-actin distribution wasrearranged to a cortical pattern, another hallmark of theepithelial phenotype (Fig. 5C, left). In contrast, the cellstransfected with miR-Ctrl showed a mesenchymal phe-notype, indicated by an absence of E-cadherin on the cellmembrane and rearrangement of F-actin from a corticalto a stress-fiber pattern (Fig. 5C, left). Consistently, forcedmiR-200b expression decreased ZEB1 expression and

markedly decreased expression of mesenchymal markersvimentin and N-cadherin (Fig. 5C, right; SupplementaryFig. S5). In a Transwell invasion assay, miR-200b expres-sion significantly decreased invaded cell numbers com-pared with miR-Ctrl (Fig. 5D, left). In addition, ectopicmiR-200b expression decreased cell migration comparedwith miR-Ctrl in a wound-healing assay (Fig. 5D, right).

Systematic delivery of miR-200b suppressed tumor growth,inhibited ZEB1, and induced E-cadherin expression in vivo.We established a gastric cancer transplantation mousemodel in BALB/C nude mice by administering a subcu-taneous injection of MGC-803 cells (see SupplementaryMethods for details). For this model, we used in vivoJetPEI (Polyplus Transfection) as a carrier for delivery ofmiR-200b, and this resulted in significant reduction intumor volumes (P ¼ 0.013; Fig. 6A and B) compared withmiR-Ctrl. We conducted immunohistochemical stainingof E-cadherin, N-cadherin, vimentin, and ZEB1 in thetumors to determine whether systemic delivery of miR-200b affected the expression of these EMT markers. Rep-resentative sections stained for these markers are shownin Fig. 6C. Compared with miR-Ctrl, miR-200b treatmentsignificantly suppressed the expression of N-cadherin(P < 0.05; Fig. 6D), vimentin (P < 0.05; Fig. 6D), andZEB1 (P < 0.05; Fig. 6D) and significantly induced E-cadherin (P < 0.05; Fig. 6D).

DiscussionUsing integrated approaches, we have uncovered a key

microRNA regulatory network that reproducibly defines themesenchymal gastric cancer subtype significantly associatedwith poor overall survival. Tissue microarray validation in385 gastric cancer cases solidified our discovery at theprotein level that patients with tumors showing the mes-enchymal phenotype had a poor prognosis in comparisonwith patients whose tumors were of the epithelial pheno-type. This study is a major step forward from currentapproaches for predicting gastric cancer outcome in that itreveals regulatory mechanisms associated with the sub-types. In particular, our integrated analysis highlights theimportant role of a microRNA regulatory network consist-ingof 10keymicroRNAs for themesenchymal gastric cancersubtype. Notably, three of the top key microRNAs (miR-200c, miR-200b, and miR-125b) were associated with sur-vival in both microarray discovery patients and PCR vali-dation patients, suggesting their essential role in gastriccancer progression. Our extensive functional studies con-sistently validated miR-200b as a potent EMT inhibitor thatmay have therapeutic potential in gastric cancer, one of themost aggressive cancer types among women. To the best of

(Continued.) ��, P < 0.01. C, left, inverse phase microscopy and E-cadherin/F-actin staining of MGC-803 and SGC-7901 cells transfected with miR-200b ormiR-Ctrl for 72 hours. Cell nuclei were stained with DAPI. Scale bars, 20 mm. C, right, Western blotting of epithelial and mesenchymal markers inMGC-803 and SGC-7901 cells transfected with miR-200b or miR-Ctrl from the same transfection as in A. D, left, in vitro Transwell invasion assay. Cells fromthe same transfection as in A were seeded into triplicate Matrigel-coated invasion chambers at 24 hours posttransfection and allowed to invade towardserum for 22 hours. The invading cell numbers on each filter were counted. ��,P < 0.01. D, right, wound-healing assay. Cells from the same transfection as in Awere seeded into 6-well dishes, and a scratch wound was applied at 24 hours posttransfection.







our knowledge, this is the first integrated analysis of micro-RNA, mRNA, and protein expression data in a study ongastric cancer survival.

The integrated profiling method has been used success-fully in studies on cancer outcome. However, previousstudies using microRNA profiling have identified few con-sistent and repeatable prognostic markers for gastric cancer(14–17). This may be due partly to population heteroge-neity. Selection of markers based solely on statistical asso-ciation and neglecting functional context may have madethe results less reliable. The miR-200a/b identified in ourstudy, although never reported as gastric cancer prognostic

markers in previous microRNA profiling studies, is func-tionally related to gastric cancer.

The miR-200 family consists of 5 members organized in 2clusters: miR-200a, miR-200b, and miR-429 on chromo-some 1 and miR-200c and miR-141 on chromosome 12. Sofar, no population study has demonstrated an associationbetween the miR-200 family and gastric cancer survival,whereas an in vitro study found that miR-200b has thepotential to regulate metastasis in gastric cancer (18). In fact,membersof themiR-200 familyhavebeenusedasprognosticmarkers for several cancer types (19–24). The predominantfunction of the miR-200 family in cancer progression is

miR-NC

A B

C D

miR-200b

miR-NC

50 mM 50 mM

HE

E-c

ad

he

rin

N-c

ad

he

rin

Vim

en

tin

ZE

B 1

miR-200b

Nude mouse (n = 9)

miR-NCmiR-200b

miR-NCmiR-200b

E-cadherin

N-cadherin

Vimentin

ZEB1

P = 0.023

P = 0.019

8

6

4

2

0

8

6

4

2

0

2,000

1,500

1,000

500

00 7 14 21 28 35

Paired t test

Time (d)

Tu

mo

r vo

lum

e (

mm

3)

P = 0.013

P = 0.046

P = 0.022

3

2

1

0

Pro

tein

exp

ressio

n

(IH

C s

co

re)

Pro

tein

exp

ressio

n

(IH

C s

co

re)

Pro

tein

exp

ressio

n

(IH

C s

co

re)

Pro

tein

exp

ressio

n

(IH

C s

co

re)

10

8

6

4

2

0

Figure 6. miR-200b inhibits tumorgrowth in an orthotopic mousemodel of gastric cancer. A,representative images of tumornodules in control miRNA- andmiR-200b–treated mice. Scalebar, 1 cm. B, quantification oftumor volume in control- andmiR-200b–treated mice. Error bars,�SEM. C, tumor samples fromcontrol- and miR-200b–treatedmice were sectioned and stainedfor E-cadherin, N-cadherin,vimentin, and ZEB1 byimmunohistochemistry (IHC).Scale bars, 50 mm. D,quantification of E-cadherin,N-cadherin, vimentin, and ZEB1protein expression. Error bars,�SD.

Song et al.






suppression of EMT, the initiating step of metastasis. ThemiR-200 family has been recognized as a master regulator ofthe epithelial phenotype by targeting transcriptional repres-sors of the cell adherence gene (25, 26). Each of the 5 familymembers has been shown to inhibit EMT and cell migration.EMT plays a key role in invasion and metastasis during

carcinogenesis. One of the molecular hallmarks drivingEMT is functional loss of E-cadherin, a cell adhesion proteinand amajor constituent of adherens junctions that acts as asuppressor of migration and invasion during carcinomaprogression (27). The mechanism for E-cadherin transcrip-tional silencing during EMT has been proposed to be directinhibition by transcriptional repressors such as ZEB, Twist,and Snail. During EMT, gastric cancer cells with fibroblasticmorphologic changes show increased migration and inva-siveness as a result of decreased cell–cell adhesion, and thecells then acquire a spindle-shaped, highlymotile fibroblastphenotype (28). Several studies have reported associationsbetween EMT-related proteins and tumor metastasis andprognosis in gastric cancer (29–31). Generally, loss ofepithelial proteins (such as E-cadherin and cytokeratin)and/or acquisition of mesenchymal proteins [such asb-catenin (nuclear) and N-cadherin] are associated withpoor tumor differentiation, advanced stage, and poor out-come in gastric cancer (30), consistent with our findings.Target screening and luciferase assays have linked the

miR-200 family with ZEB1 and ZEB2 (32). Several studiesdemonstrated direct binding sites for themiR-200 family onthe 30-UTR of ZEB (33, 34). It has been reported thatupregulation of miR-200 reduced the expression of ZEBand increased the expression of E-cadherin in the plasmamembrane. Increased expression of miR-200 in gastriccancer cells was associated with a change in their morphol-ogy to more epithelial-like and with inhibition of cellularinvasion, migration, and proliferation (18). Our data sug-gest that miR-200a/bmay negatively regulate EMT and thusresult in better prognosis in gastric cancer. There was asignificant inverse correlation between miR-200a/b andZEB expression. Our study, for the first time, shows theassociation from a population view between the miR-200/EMT regulatory network and gastric cancer prognosis. Ourstudy extends previous studies in identifyingmiR-200a/b asa prognostic marker of gastric cancer, thus capturing thebiologic information of the complex EMT regulatory net-work in a single microRNA.In our validation, significant association between miR-

200a/b expression and gastric cancer survival was observedmainly in women, not in men. We cannot absolutely ruleout chance findings in this study, although there are severalstudies indicating that themiR-200 familymay be related tohormones (35, 36). Gastric cancer is a hormone-relatedcancer. Treating male mice with estrogen dramatically low-ers their rates of gastric cancer (37). A population-basedSwedish cohort study, designed to detect possible effects ofestrogen in the etiology of gastric cancer, revealed a reducedrisk of gastric cancer among a cohort of patients withprostate cancer, most of whom had received estrogen treat-ment (38). Male and female gastric cancers differ in their

etiology, and it is possible that the miR-200 family isfunctionally dependent on estrogen and affects gastric can-cers more in women than in men.

The expression of miR-200c was not associated withgastric cancer prognosis in our study. The members of themiR-200 family largely target a common subset of genesthat includes ZEB, and members from each cluster are co-expressed. However, the expression of miR-200 familymembers in the 2 gastric cancer clusters does not appearto be highly correlated (32, 39). Expression of miR-200aand miR-200b was highly correlated, but their expressionwas not significantly correlated with miR-200c. More oftenthan not, no synergy is shown between the two clusters ofthe miR-200 family. Hur and colleagues investigated therole of miR-200 members in the pathogenesis of metastaticcolorectal cancer and found that miR-200c, but not miR-200a/b, plays an important role in mediating EMT andmetastatic behavior in the colon (40). In a similar study onovarian cancer, researchers found that low-level expressionof the miR-200a/b cluster predicts poor survival (19).

This study has several limitations. First, miR-125b, as awell-known oncomiR, has been associated with poor sur-vival in many cancer types, including gastric cancer. Uedaand colleagues identified miR-125b as the most importantprogression-related signature of gastric cancer among 237microRNAs analyzed (41). miR-125b may act as an onco-gene in gastric cancer by dysregulating gastric cell prolifer-ation and apoptosis (42). A recent study found miR-125bexpression correlates inversely with HER2 status, and dysre-gulation of miR-125b and HER2 is an early event in thegastric (intestinal-type) oncogenesis (43). In our integrateddata analysis, no cancer-related regulatory network wasconstructed specifically formiR-125b. The underlyingmech-anismfor its associationwithgastric cancer survivalhas yet tobe explored. Second,we focusedonEMT, soother functionalpathways such as nucleotide metabolism and transcriptionregulation were not explored. Although EMT is closelyrelated to both the miR-200 family and gastric cancer pro-gression, other pathwaysmay also hold great insight into thedifferential survival of the 2 subtypes of gastric cancer.

Our observation on the role of the miR-200 family (miR-200a/b) in regulating EMT through ZEB1 and E-cadherinenhances our understanding of the microRNA regulatorypathways influencing the clinical progression and progno-sis of gastric cancer, especially in women. The miR-200family may serve as a good prognostic marker for gastriccancer, potentially opening up a new avenue for therapeuticintervention in patients with localized primary gastric can-cer. Further studies arewarranted to replicate our findings indifferent populations.

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

Authors' ContributionsConception anddesign: F. Song,D. Yang, B. Liu,W. Zhang, X.Hao, K. ChenDevelopment of methodology: D. Yang, H. Zheng, W. ZhangAcquisitionofdata (provided animals, acquired andmanagedpatients,provided facilities, etc.): F. Song, D. Yang, B. Liu, Y. Guo, H. Zheng,T. Wang, J. Yu, Y. Zhao, R. Niu, H. Liang







Analysis and interpretation of data (e.g., statistical analysis, biosta-tistics, computational analysis): F. Song, D. Yang, B. Liu, L. Li, J. Yu,H. Winkler, W. ZhangWriting, review, and/or revision of the manuscript: F. Song, D. Yang,B. Liu, H. Winkler, W. Zhang, K. ChenAdministrative, technical, or material support (i.e., reporting or orga-nizing data, constructing databases): Y. Guo, Y. ZhaoStudy supervision: W. Zhang

AcknowledgmentsThe authors thank Kathryn L. Hale of the Department of Scientific

Publications at The University of Texas MD Anderson Cancer Center forediting the manuscript. They also thank Dr Yan Sun, Department of Pathol-ogy, Tianjin Medical University Cancer Institute and Hospital, for her helpand assistance in our experiments. They thank Karin Verstraeten and TinekeCasneuf from Janssen Research and Development, a Division of JanssenPharmaceutica NV, for help in gene expression profiling.

Grant SupportThis study was supported by the Program for Changjiang Scholars and

Innovative Research Team in University (PCSIRT) in China (IRT1076), theNational Key Scientific and Technological Project (2011ZX09307–001–04),and the National Natural Science Foundation of China (No.81172762,81071627). The tissue bank is jointly supported by the Tianjin CancerInstitute andHospital and the USNational Foundation for Cancer Research.D. Yang is an Odyssey Fellow, supported by the Odyssey Program and theTheodoreN. LawEndowment for Scientific Achievement at TheUniversity ofTexas MD Anderson Cancer Center.

The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely to indicatethis fact.

Received July 3, 2013; revised November 15, 2013; accepted December 8,2013; published OnlineFirst December 18, 2013.

References1. Ferlay J, Shin HR, Bray F, Forman D, Mathers C, Parkin DM. Estimates

ofworldwide burden of cancer in 2008:GLOBOCAN2008. Int JCancer2010;127:2893–917.

2. Catalano V, Labianca R, Beretta GD, Gatta G, de Braud F, Van CutsemE. Gastric cancer. Crit Rev Oncol Hematol 2009;71:127–64.

3. Wu WK, Lee CW, Cho CH, Fan D, Wu K, Yu J, et al. MicroRNAdysregulation in gastric cancer: a new player enters the game. Onco-gene 2010;29:5761–71.

4. Link A, Kupcinskas J,Wex T,Malfertheiner P. Macro-role of microRNAin gastric cancer. Dig Dis 2012;30:255–67.

5. Bartel DP. MicroRNAs: genomics, biogenesis, mechanism, and func-tion. Cell 2004;116:281–97.

6. Song FJ, Chen KX. Single-nucleotide polymorphisms among micro-RNA: big effects on cancer. Chin J Cancer 2011;30:381–91.

7. Li X, Chen J, Hu X,HuangY, Li Z, Zhou L, et al. ComparativemRNA andmicroRNA expression profiling of three genitourinary cancers revealscommon hallmarks and cancer-specific molecular events. PLoS One2011;6:e22570.

8. Zhou L, Chen J, Li Z, Li X, Hu X, Huang Y, et al. Integrated profiling ofmicroRNAs andmRNAs:microRNAs located onXq27.3 associatewithclear cell renal cell carcinoma. PLoS One 2010;5:e15224.

9. Noushmehr H, Weisenberger DJ, Diefes K, Phillips HS, Pujara K,Berman BP, et al. Identification of a CpG island methylator phenotypethat defines a distinct subgroup of glioma. Cancer Cell 2010;17:510–22.

10. VerhaakRG,HoadleyKA,PurdomE,WangV,Qi Y,WilkersonMD, et al.Integrated genomic analysis identifies clinically relevant subtypes ofglioblastoma characterized by abnormalities in PDGFRA, IDH1, EGFR,and NF1. Cancer Cell 2010;17:98–110.

11. Wilkerson MD, Hayes DN. ConsensusClusterPlus: a class discoverytool with confidence assessments and item tracking. Bioinformatics2010;26:1572–3.

12. Liu Y, Hayes DN, Nobel A. Statistical Significance of Clustering forHigh-Dimension, Low-Sample Size Data. J Am Statist Assoc 2008;103:1281–93.

13. YangD,SunY,HuL, ZhengH, Ji P, Pecot CV, et al. Integrated analysesidentify a master microRNA regulatory network for the mesenchymalsubtype in serous ovarian cancer. Cancer Cell 2013;23:186–99.

14. Li X, Zhang Y, Zhang Y, Ding J, Wu K, Fan D. Survival prediction ofgastric cancer by a seven-microRNA signature. Gut 2010;59:579–85.

15. Brenner B, Hoshen MB, Purim O, David MB, Ashkenazi K, Marshak G,et al. MicroRNAs as a potential prognostic factor in gastric cancer.World J Gastroenterol 2011;17:3976–85.

16. KogoR,Mimori K, TanakaF,KomuneS,MoriM.Clinical significanceofmiR-146a in gastric cancer cases. Clin Cancer Res 2011;17:4277–84.

17. Wu WY, Xue XY, Chen ZJ, Han SL, Huang YP, Zhang LF, et al.Potentially predictive microRNAs of gastric cancer with metastasis tolymph node. World J Gastroenterol 2011;17:3645–51.

18. Kurashige J, KamoharaH,WatanabeM,Hiyoshi Y, IwatsukiM, TanakaY, et al. MicroRNA-200b regulates cell proliferation, invasion, and

migration by directly targeting ZEB2 in gastric carcinoma. Ann SurgOncol 2012;19:S656–64.

19. Hu X, Macdonald DM, Huettner PC, Feng Z, El NI, Schwarz JK, et al. AmiR-200 microRNA cluster as prognostic marker in advanced ovariancancer. Gynecol Oncol 2009;114:457–64.

20. Baffa R, Fassan M, Volinia S, O'Hara B, Liu CG, Palazzo JP, et al.MicroRNA expression profiling of humanmetastatic cancers identifiescancer gene targets. J Pathol 2009;219:214–21.

21. Roybal JD, Zang Y, Ahn YH, Yang Y, Gibbons DL, Baird BN, et al. miR-200 Inhibits lung adenocarcinoma cell invasion and metastasis bytargeting Flt1/VEGFR1. Mol Cancer Res 2011;9:25–35.

22. Smith CM, Watson DI, Leong MP, Mayne GC, Michael MZ, WijnhovenBP, et al. miR-200 family expression is downregulated upon neoplasticprogression of Barrett's esophagus. World J Gastroenterol 2011;17:1036–44.

23. Leskela S, Leandro-Garcia LJ, Mendiola M, Barriuso J, Inglada-PerezL, Munoz I, et al. The miR-200 family controls beta-tubulin III expres-sion and is associated with paclitaxel-based treatment response andprogression-free survival in ovarian cancer patients. Endocr RelatCancer 2011;18:85–95.

24. Nam EJ, Yoon H, Kim SW, Kim H, Kim YT, Kim JH, et al. MicroRNAexpression profiles in serous ovarian carcinoma. Clin Cancer Res2008;14:2690–5.

25. DavalosV,MoutinhoC, VillanuevaA,BoqueR,SilvaP,Carneiro F, et al.Dynamic epigenetic regulation of the microRNA-200 family mediatesepithelial and mesenchymal transitions in human tumorigenesis.Oncogene 2012;31:2062–74.

26. Spaderna S, Brabletz T, Opitz OG. The miR-200 family: central playerfor gain and loss of the epithelial phenotype. Gastroenterology 2009;136:1835–7.

27. Birchmeier W, Behrens J. Cadherin expression in carcinomas: role inthe formation of cell junctions and the prevention of invasiveness.Biochim Biophys Acta 1994;1198:11–26.

28. Thiery JP. Epithelial-mesenchymal transitions in development andpathologies. Curr Opin Cell Biol 2003;15:740–6.

29. Ru GQ, Wang HJ, Xu WJ, Zhao ZS. Upregulation of Twist in gastriccarcinoma associated with tumor invasion and poor prognosis. PatholOncol Res 2011;17:341–7.

30. Kim MA, Lee HS, Lee HE, Kim JH, Yang HK, Kim WH. Prognosticimportance of epithelial-mesenchymal transition-related proteinexpression in gastric carcinoma. Histopathology 2009;54:442–51.

31. Castro AC, Rosivatz E, Schott C, Hollweck R, Becker I, Sarbia M, et al.Slug is overexpressed in gastric carcinomas and may act synergisti-cally with SIP1 andSnail in the down-regulation of E-cadherin. J Pathol2007;211:507–15.

32. Park SM, Gaur AB, Lengyel E, Peter ME. The miR-200 family deter-mines the epithelial phenotype of cancer cells by targeting the E-cadherin repressors ZEB1 and ZEB2. Genes Dev 2008;22:894–907.

33. Hurteau GJ, Carlson JA, Spivack SD, Brock GJ. Overexpression of themicroRNA hsa-miR-200c leads to reduced expression of transcription

Song et al.






factor 8 and increased expression of E-cadherin. Cancer Res 2007;67:7972–6.

34. Christoffersen NR, Silahtaroglu A, Orom UA, Kauppinen S, Lund AH.miR-200b mediates post-transcriptional repression of ZFHX1B. RNA2007;13:1172–8.

35. Renthal NE, Chen CC, Williams KC, Gerard RD, Prange-Kiel J, Men-delson CR. miR-200 family and targets, ZEB1 and ZEB2, modulateuterine quiescence and contractility during pregnancy and labor. ProcNatl Acad Sci U S A 2010;107:20828–33.

36. Panda H, Pelakh L, Chuang TD, Luo X, Bukulmez O, Chegini N.Endometrial miR-200c is altered during transformation into cancerousstates and targets the expression of ZEBs, VEGFA, FLT1, IKKbeta,KLF9, and FBLN5. Reprod Sci 2012;19:786–96.

37. Sheh A, Ge Z, Parry NM,Muthupalani S, Rager JE, Raczynski AR, et al.17beta-estradiol and tamoxifen prevent gastric cancer by modulatingleukocyte recruitment and oncogenic pathways in Helicobacter pylori-infected INS-GASmalemice. Cancer PrevRes (Phila) 2011;4:1426–35.

38. Lindblad M, Ye W, Rubio C, Lagergren J. Estrogen and risk of gastriccancer: a protective effect in a nationwide cohort study of patients with

prostate cancer in Sweden. Cancer Epidemiol Biomarkers Prev 2004;13:2203–7.

39. KorpalM, Lee ES, HuG, Kang Y. ThemiR-200 family inhibits epithelial-mesenchymal transition andcancer cellmigration bydirect targetingofE-cadherin transcriptional repressors ZEB1 and ZEB2. J Biol Chem2008;283:14910–4.

40. Hur K, ToiyamaY, TakahashiM, Balaguer F, Nagasaka T, Koike J, et al.MicroRNA-200c modulates epithelial-to-mesenchymal transition(EMT) in human colorectal cancer metastasis. Gut 2013;62:1315–26.

41. Ueda T, Volinia S, Okumura H, Shimizu M, Taccioli C, Rossi S, et al.Relation between microRNA expression and progression and prog-nosis of gastric cancer: amicroRNAexpression analysis. LancetOncol2010;11:136–46.

42. Yang ZX, Lu CY, Yang YL, Dou KF, Tao KS. MicroRNA-125b expres-sion in gastric adenocarcinoma and its effect on the proliferation ofgastric cancer cells. Mol Med Rep 2013;7:229–32.

43. Fassan M, Pizzi M, Realdon S, Balistreri M, Guzzardo V, Zagonel V,et al. The HER2-miR125a5p/miR125b loop in gastric and esophagealcarcinogenesis. Hum Pathol 2013;44:1804–10.







Documents

Integrated MicroRNA Network Analyses Identify a Poor ... · Integrated MicroRNA Network Analyses Identify a Poor-Prognosis € Updated version € ... We used integrated analysis