Lack of microRNA-101 causes E-cadherin functional deregulation through EZH2 up-regulation in intestinal gastric cancer

ORIGINAL PAPERJournal of PathologyJ Pathol (2012)Published online in Wiley Online Library(wileyonlinelibrary.com) DOI: 10.1002/path.4032

Lack of microRNA-101 causes E-cadherin functional deregulationthrough EZH2 up-regulation in intestinal gastric cancerJoana Carvalho,1,2 Nicole C van Grieken,3 Patricia M Pereira,4 Sonia Sousa,1 Marianne Tijssen,3 Tineke E Buffart,3Begona Diosdado,3 Heike Grabsch,5 Manuel AS Santos,4 Gerrit Meijer,3 Raquel Seruca,1,2 Beatriz Carvalho3

and Carla Oliveira1,2*

1 Institute of Molecular Pathology and Immunology, University of Porto (IPATIMUP), Portugal2 Faculty of Medicine, University of Porto, Portugal3 Department of Pathology, VU University Medical Centre, Amsterdam, The Netherlands4 RNA Biology Laboratory, Department of Biology and CESAM, University of Aveiro, Portugal5 Pathology and Tumour Biology, Leeds Institute of Molecular Medicine, University of Leeds, UK

*Correspondence to: Carla Oliveira, IPATIMUP, Cancer Genetics, Rua Dr Roberto Frias, s/n 4200-465 Porto, Portugal. e-mail: [email protected]

AbstractE-cadherin expression disruption is commonly observed in metastatic epithelial cancers and is a crucial step ingastric cancer (GC) initiation and progression. As aberrant expression of microRNAs often perturb the normalexpression/function of pivotal cancer-related genes, we characterized and dissected a pathway that causes E-cadherin dysfunction via loss of microRNA-101 and up-regulation of EZH2 expression in GC. MicroRNA microarrayexpression profiling and array-CGH were used to reinforce miR-101 involvement in GC. By using quantitativereal-time PCR and quantitative SNaPshot genomic PCR, we confirmed that miR-101 was significantly down-regulated in GC (p < 0.0089) in comparison with normal gastric mucosas and, at least in 65% of the GC casesanalysed, this down-regulation was caused by deletions and/or microdeletions at miR-101 genomic loci. Moreover,around 40% of cases showing miR-101 down-regulation displayed concomitant EZH2 over-expression (at theRNA and protein levels), which, in turn, was associated with loss/aberrant expression of E-cadherin. Interestingly,this occurred preferentially in intestinal-type GCs, retaining allele(s) untargeted by classical CDH1-inactivatingmechanisms. We also demonstrated that miR-101 gain of function or direct inhibition of EZH2 in Kato III GC cellsled to a strong depletion of endogenous EZH2 and consequent rescue of E-cadherin membranous localization,mimicking results obtained in clinical GC samples. In conclusion, we show that deletions and/or microdeletionsat both miR-101 genomic loci cause mature miR-101 down-regulation, subsequent EZH2 over-expression andE-cadherin dysfunction, specifically in intestinal-type GC.Copyright © 2012 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

Keywords: gastric cancer; E-cadherin (CDH1); enhancer of Zeste homologue 2 (EZH2); microRNA-101

Received 13 January 2012; Revised 10 March 2012; Accepted 20 March 2012

No conflicts of interest were declared.

Introduction

Gastric cancer (GC) causes 12% of all cancer-relateddeaths each year [1,2]. The molecular mechanismsunderlying this disease are largely unknown and inac-tivation of E-cadherin (CDH1 gene; ENSG00000039068, OMIM#192 090) is the most well-establisheddefect in GC initiation and progression [3]. Intragenicmutations, large CDH1 locus deletions (loss of het-erozygosity, LOH) and CDH1 promoter methylationrepresent mechanisms inactivating E-cadherin in hered-itary [4–8] and sporadic GC with diffuse histology[9–13]. Although these classical CDH1 inactivationmechanisms have been associated with almost 50%of hereditary and sporadic GCs with a diffuse compo-nent, aberrant or absent E-cadherin protein expressionoccurs in most GCs of both diffuse and intestinal his-tology [14]. Therefore, alternative defects may exist to

explain the overall E-cadherin expression abnormali-ties in both histotypes. Appealing alternative mecha-nisms are chromatin changes, such as histone modifi-cations [15], and altered expression of specific miR-NAs [16]. Combination of both mechanisms has alsobeen previously described inactivating other genes[17].

The enhancer of Zeste homologue 2 (EZH2 ) geneencodes a histone methyltransferase which is part ofthe polycomb repressor complex 2 (PRC2) [18,19].This enzyme catalyses the addition of methyl groupsto histone H3 at lysine 27 [20], imposing an abnormalrepressive chromatin mark associated with the silenc-ing of tumour suppressor genes, such as CDH1 [21,22].EZH2 over-expression has been found in many cancers,including GC, associated with poor clinical outcomeof patients [23–26]. Moreover, high EZH2 expressionlevels stimulate cell proliferation, block apoptosis andpromote invasion and metastasis [27–30].

Copyright © 2012 Pathological Society of Great Britain and Ireland. J Pathol (2012)Published by John Wiley & Sons, Ltd. www.pathsoc.org.uk www.thejournalofpathology.com

J Carvalho et al

Profiling studies revealed the contribution of aber-rant miRNA expression to GC initiation and progres-sion [31,32] by perturbing the function of target genes.MiR-101, for instance, has been extensively studiedin a variety of human cancers and also shown tobe down-regulated in GC [33–36]. Aberrant silencingof miR-101 has been proposed to cause EZH2 over-expression [37–39]. Although miR-101-based regula-tion over EZH2, as well as regulation of E-cadherinfunction by EZH2, has been addressed independently,the interaction between these three players remains tobe systematically investigated. Therefore, in the presentstudy we sought to explore a molecular pathway lead-ing to E-cadherin dysfunction, involving miR-101 andEZH2. We profiled the expression of miRNAs in aseries of primary GC cases and confirmed wide miR-101 down-regulation. We identified chromosomal dele-tions as the genetic cause for this down-regulationin most cases and put together a cascade of events,starting with miR-101 loss, followed by EZH2 up-regulation and culminating in E-cadherin function dis-ruption, using an in vitro model. Moreover, we wereable to correlate all these steps in a series of well-characterized primary GC cases.

Materials and methods

Tissue specimens and cell linesTissue specimens (n = 76 GC and n = 10 normalgastric mucosa) were collected from the PathologyDepartments, Leeds University Hospitals and VUmcin Amsterdam. Informed consent was obtained fromall patients and the study was approved by the appro-priate hospitals’ ethics committees. GC cell lines wereobtained from ATCC (Manassas, VA, USA) (n =7) and established in IPATIMUP (n = 2) [40] (seeSupporting information, Supplementary materials andmethods).

DNA and RNA protocolsAll tumour samples used were reviewed by an expertgastrointestinal pathologist (NVG) and histologicallyverified to contain a minimum of 70% neoplastic cells.DNA from GC was isolated as described [41]. RNAwas isolated from tissues and GC cell lines, usingRecoverAll™ Total Nucleic Acid Isolation Kit for FFPE(Ambion, Austin, TX, USA) and mirVana™ miRNAIsolation Kit’s protocol (Ambion).

MicroRNA microarrayWe used microarrays (miRNAChip_human_V2) prin-ted by the National DNA-Microarray Facility, Univer-sity of Aveiro, Portugal, to profile 37 GC and 10 normalgastric mucosas (pooled to obtain the normal gas-tric reference sample). RNA labelling and hybridiza-tion on miRNAChip_human_V2 were performed asdescribed [42]. Microarray images were analysed using

Quantarray v 3.0 software (Perkin-Elmer). A globalmedian normalization of human GC microarray datawas applied using BRB-ArrayTools v 3.4.0 software.Log-transformed (base 2) cy3 intensities were obtained.Differentially expressed miRNAs from each samplewere identified using the Multiple Experiment ViewerMev software v 4.6.2. A non-parametric Wilcoxonrank sum (Mann–Whitney) test at a critical value ofp = 0.05 and FDR = 0.05 was performed to iden-tify miRNAs that demonstrated statistically significantdifference in expression between normal and tumour.Raw data was submitted in Gene Expression Omnibusdatabase, Accession Nos GPL14895, GSE33743 andGSM833952-GSM833992 (see Supporting informa-tion, Supplementary materials and methods).

Quantification of miR-101 expression by real-timePCRTaqMan microRNA assays (Applied Biosystems, Fos-ter City, CA, USA) were used to quantify the expres-sion levels of mature miR-101 and endogenous controlRNU6B or RNU48 on an ABI Prism 7000 SequenceDetection System (Applied Biosystems). Data wereanalysed by the comparative 2−��Ct method [43] (seeSupporting information, Supplementary materials andmethods).

Array comparative genomic hybridization (aCGH)dataWe analysed copy number changes at miR-101-1 andmiR-101-2 genomic loci, using aCGH data from a pre-vious study (see Supporting information, Figure S1)and validated these changes in 25 ng DNA extractedfrom 12 GC cases and peripheral blood lympho-cytes (PBL) from three control samples submitted togenomic PCR followed by quantitative SNaPshot (Q-SNaPshot). Two independent multiplex PCR reactions(Multiplex PCR Kit, Qiagen, Hilden, Germany) wereperformed, encompassing amplification of the precur-sor mir-101-1 and GAPDH, and the precursor mir-101-2 and GAPDH. SNaPshot reactions using puri-fied PCR products, SNaPshot Multiplex Kit (AppliedBiosystems) and single base extension primers formiR-101-1, miR-101-2 and GAPDH were analysed ina 310 Genetic Analyzer (Applied Biosystems). Thepeak area corresponding to each miR-101 locus inde-pendently and GAPDH was quantified and a ratio cal-culated. MiR-101:GAPDH ratios for GC and controlwere compared (see Supporting information, Supple-mentary materials and methods).

Quantification of EZH2 and CDH1 mRNA expressionby real-time PCRApproximately 200 ng RNA was reverse-transcribedusing Superscript II Reverse Transcriptase and ran-dom hexamer primers (Invitrogen). Quantitative real-time PCR (RT–PCR) was performed in triplicate forEZH2 and CDH1 and 18s RNA or GAPDH (endoge-nous controls), using TaqMan gene expression assays


MiR-101 modulates E-cadherin in gastric cancer

(Applied Biosystems) and ABI Prism 7000 SequenceDetection System (see Supporting information, Supple-mentary materials and methods).

Gain/loss of function experimentsKatoIII cells were stably transfected with the miR-101expressing vector (Kato-miR-101) or the miR-negativecontrol plasmid (Kato-miR-NC) and transiently trans-fected with siRNA targeting EZH2 and a non-silencingsiRNA (150 nM On-Target plus Smart pool and Non-targeting siRNA, Thermo Scientific Dharmacon, UK),using Lipofectamine 2000 (Invitrogen). Three biolog-ical replicas were obtained for each experiment (seeSupporting information, Supplementary materials andmethods).

Western blotTotal protein lysates were immunoblotted with antibod-ies to E-cadherin (1 : 1000; 24E10, Cell Signaling, MA,USA), EZH2 (1 : 1000; BD Transduction Laboratories,NJ, USA) and α-tubulin (1 : 10000, Sigma-Aldrich,MO, USA). Horseradish peroxidase-conjugated sec-ondary antibodies (GE Healthcare, Buckinghamshire,UK) were used and the signal detected with ECL (GEHealthcare) (see Supporting information, Supplemen-tary materials and methods).

Flow cytometry analysis (FCM)FCM was performed as described [44]. The experi-ments were repeated three times and analysed withWinMDI (Joe Trotter, TSRI, San Diego, CA, USA)(see Supporting information, Supplementary materialsand methods).

Co-immunofluorescence for EZH2 and E-cadherinproteinsWhole sections of FFEP GC and GC cell lines wereco-immunostained for EZH2 (1 : 150) and E-cadherin(1 : 150), using the above-mentioned antibodies. Anti-rabbit/mouse Alexa 488 and anti-rabbit/mouse Alexa594 (1 : 500; Invitrogen) were used as secondary anti-bodies (see Supporting information, Supplementarymaterials and methods).

E-cadherin immunohistochemistryHeat-induced antigen retrieval was performed usingcitrate buffer solution, pH 6.0. Slides were incubatedwith 24E10 E-cadherin antibody (1 : 50) and labelledwith Envision Detection System Anti-Rabbit/Mouse(Dako, CA, USA). E-cadherin immunoreactivity wasscored according to the percentage of positive tumourcells and cellular localization (see Supporting informa-tion, Supplementary materials and methods).

CDH1 promoter methylation analysisThis was performed as described [7] (see Supportinginformation, Supplementary materials and methods).

CDH1 mutation screening and LOH analysisAll CDH1 exons were screened for somatic muta-tions in four independent multiplex PCR reactionsand sequenced with exon-flanking-specific primers.LOH analysis was performed as described [7] (seeSupporting information, Supplementary materials andmethods).

Statistical analysisDifferences in expression levels between controlsand tumour samples were evaluated applying theMann–Whitney non-parametric U-test. For clinico-pathological correlations, Fisher’s test was used tocalculate the two-tailed p value. Student’s t-test wasused to establish significant differences observed within vitro experiments. All data are presented as mean± SEM and were analysed using Prism 5.0 software(GraphPad). Differences were taken to be significantwhen p < 0.05.

Results

Deregulation of microRNA expression in GCMiRNA microarray expression profiling picked up 70miRNAs differentially expressed in GC and normalmucosas: 63 up-regulated and seven down-regulated(p < 0.05, FDR ≤ 0.05; Figure 1A). MiR-501-5p,miR-149, miR-519e∗, miR-218-2∗ and miR-429 werethe five most up-regulated, while miR-451, miR-101, miR-502-3p, miR-33a and miR-516a-3p/516b∗were the most down-regulated miRNAs (Figure 1B).Amongst those down-regulated, miR-101 ranked sec-ond (fold change 0.22, range 0.16–0.35), reinforcingits role in GC. MiR-101 expression levels were val-idated in 37 primary GCs and a control pool of fivenormal gastric mucosas by RT–PCR and were signifi-cantly decreased in tumours in comparison with normalcounterparts (p < 0.0089; Figure 1C). Following thesame trend, 8/9 GC cell lines expressed lower levelsof miR-101 relative to commercial RNA from normalstomach (Figure 1D).

DNA loss at miR-101 loci underlies miR-101down-regulationTo elucidate the mechanism underlying mature miR-101 down-regulation in GC, we analysed an aCGHdata set (76 GCs; Buffart, et al., submitted), contain-ing the previous analysed GCs, to infer DNA copynumber changes at miR-101 loci (miR-101-1 locus atchromosome 1p31.3 and miR-101-2 locus at chromo-some 9p24.1). Changes in the interval encompassingthe miR-101-1 locus were rarely found (1.3%). How-ever, loss of the region containing the miR-101-2 locuswas detected in 51.3% of all GCs, suggesting that DNAloss specifically at this region is a frequent molecularevent in GC (see Supporting information, Figure S1).


J Carvalho et al

p < 0.00893

2

miR

-101

rel

ativ

e ex

pres

sion

(S

ampl

e/R

NA

poo

l)

1

1

2

Log 2

miR

-101

rel

ativ

e ex

pres

sion

-1

0

-2

-3

Normal S

tomach

NCI-N87

AGS

SNU638

MKN28GP202

KATO III

IPA220

MKN45SNU1

0N GC

Log2 Transformed Intensity

14.0

miR-451

miR-502-3pmiR-101miR-582-5pmiR-33amiR-151-5pmiR-516a-3p/miR-516bmiR-296-5pmiR-30b*miR-513-5pmiR-335*miR-576-5pmiR-219-2-3pmiR-331-5p

miR-889miR-152miR-992miR-93

miR-519cmiR-599miR-520a-5pmiR-631miR-550miR-136miR-22miR-515-5pmiR-127-3pmiR-374a*miR-181a*miR-192*miR-532-3p

miR-30d*miR-640miR-425*miR-92b*miR-501-5pmiR-514miR-576-3pmiR-519e*

miR-149miR-219-1-3p

miR-424*miR-220miR-96

miR-218-2*miR-649miR-215miR-182miR-122miR-524-3p

miR-187miR-526bmiR-770-5pmiR-545*miR-200bmiR-9

miR-141*miR-579miR-493miR-137miR-216a

miR-503miR-126*miR-23bmiR-99bmiR-101*miR-323-3p

miR-25miR-92a-1*miR-4291.0

0.84971863

0.69943726

T8

T38

T37

T23

T35

T26

T43

T16

T32

T34

T33

T6

T42

T39

T36

T2

T28

T20

T13

T12

T31

T18

T15

T1

T5

T4

T3

T29

T19

T11

T30

T14

T41

T24

T22

T27

T21

N4

N1

N2

N3

1.0

0.33

5368

7

-0.3

2926

26

9.06.0

B

A

C DFold Change†

(Tumour/Normal) p value*

7.51

6.11

4.18

4.17

4.05

0.16

0.22

0.26

0.27

0.35Mic

roR

NA

s do

wnr

egul

ated

Mic

roR

NA

s up

regu

late

d miR-501-5p

miR-149

miR-519e*

miR-218-2*

miR-429

miR-451

miR-101

miR-502-3p

miR-33a

miR-516a-3p/miR-516b* 5.6x10-3

4.3x10-3

3.7x10-3

4.9x10-3

6.5x10-3

4.9x10-3

3.7x10-3

3.7x10-3

2.1x10-3

1.3x10-3

Figure 1. Deregulation of microRNA expression in GC. (A) Unsupervised hierarchical clustering of deregulated miRNAs. Microarray datawas clustered (Pearson correlation, average linkage) for four normal pooled gastric mucosa and 37 gastric tumour samples (columns) and70 statistical differentially expressed miRNAs (rows). Colours in the heat map represent the log2-transformed normalized intensities: green,lower expression; grey, absent data points; red, higher expression. (B) Summary of the most differentially expressed miRNAs; ∗absolute pvalue based on the Wilcoxon rank sum test; † based on the median expression values relative to reference. (C) Scatter plot with miR-101expression levels quantified by QRT–PCR in five normal gastric mucosas and 37 GCs. Each value was normalized to the expression levelsobserved in a pool of normal mucosas. Statistical significance p value was obtained by using the Mann–Whitney test. (D) Expression levelsof miR-101 in the indicated GC cell lines were normalized to the expression levels observed in commercial RNA from normal stomach.Error bars refer to measurements in two biological replicates.



0.4

0.3

0.2

0.1

1.5

1.0

0.5

0.0

Gen

omic

qua

ntifi

catio

n (m

iR-1

01-1

locu

s/G

AP

DH

)

Gen

omic

qua

ntifi

catio

n (m

iR-1

01-2

locu

s/G

AP

DH

)

0.0

Control GCNo loss

GCloss 9p24.1

Control GCNo loss

GCloss 9p24.1

miR-101-2 locus(chr9p24.1)

A BmiR-101-1 locus

(chr1p31.3)

p < 0.01 p > 0.05p < 0.02

Figure 2. Genomic quantification of miR-101-2 (A) and miR-101-1 (B) on chromosome 9p24.1 and 1p31.3, respectively. Scatter plotsrepresent the DNA level of miR-101 in three different groups ofsamples (output of aCGH data): n = 3 blood DNA ‘control’; n = 6GCs with normal copy number at both loci, ‘GC No loss’; n = 6GCs with DNA copy number loss only at chromosome 9p24.1‘GC loss 9p24.1’. Black and grey curves represent the genomicquantification of miR-101 and GAPDH, respectively. Statisticallysignificant p value was obtained by using the Mann–Whitney test.

Given the low resolution of the BAC aCGH platformused in the previous study, we validated the aCGHresults, adopting a restrictive and quantitative strategybased on genomic PCR directed to each miR-101 locus,followed by Q-SNaPshot. We selected six GC samplesdisplaying genomic loss by aCGH only at the miR-101-2 locus (‘GC loss 9p24.1’) and six GC sampleswithout genomic loss by aCGH at either miR-101 locus(‘GC No loss’). PBLs’ DNA from three cancer-freeindividuals was used as ‘control’ for normal copy num-ber changes. The genomic quantification of miR-101-2locus revealed that 6/6 ‘GC loss 9p24. 1’ samples,according to the aCGH data, also showed loss by Q-SNaPshot, when compared with the control group (p <

0.01). Moreover, 5/6 ‘GC No loss’ samples at 9p24.1by aCGH also exhibited decreased levels of DNAwhen compared with the control group (p < 0.02;Figure 2A; see also Supporting information, TableS1). The genomic quantification of miR-101-1 locusrevealed that, overall, no copy number loss was foundin GC cases when compared with the control group(p > 0.05; Figure 2B). Nevertheless, a detailed anal-ysis of each tumour demonstrated that 7/12 tumourswithout aCGH copy number changes at the miR-101-1locus displayed loss of DNA at this region by Q-SNaPshot (see Supporting information, Table S1).

Together, these results pinpointed deletions/micro-deletions at miR-101-2 (9p24.1) and miR-101-1(1p31.3), which remained undetected by aCGH, as oneof the mechanisms responsible for mature miR-101down-regulation in at least 65% (n = 19/29) of the GCcases tested (Table 1; see also Supporting information,Figure S2).

4

2

0

Log 1

0 E

ZH

2 ex

pres

sion

-2

-4

4 p < 0.0095A B

p > 0.05p < 0.012p < 0.0006

2

0

Log 1

0 E

ZH

2 ex

pres

sion

-2

-4N GC

IntestinalN GC GC

Diffuse/Mixed

Figure 3. Over-expression of EZH2 is associated with intestinalGC. (A) Scatter plots with EZH2 mRNA expression levels in normalgastric mucosas (n = 9) and GCs (n = 29). (B) Scatter plots withEZH2 mRNA expression levels in normal gastric mucosas (n = 9),intestinal GCs (n = 14) and diffuse/mixed GCs (n = 15). EZH2expression levels were relative to 18s. Statistically significantp value was obtained by using the Mann–Whitney test.

EZH2 over-expression is associated with loss ofmiR-101 loci and decreased mature miR-101expression in intestinal GCEZH2 has been reported as one of the target genes ofmiR-101 [37]. Nevertheless, the extent of its expres-sion regulation by miR-101 has not been systemati-cally addressed in GC. To test whether miR-101 lossof expression observed in most GCs could be asso-ciated with EZH2 over-expression, we analysed byQRT–PCR the expression level of EZH2 in two bio-logical subsets: (a) 29 GC cases previously analysedfor miR-101 expression; and (b) nine normal gastricmucosa samples from cancer-free individuals.

EZH2 mRNA expression levels were significantlyincreased in tumours in comparison with normal sam-ples (p < 0.0095; Figure 3A). Moreover, EZH2 over-expression was mainly associated with intestinal his-totype GC (p < 0.0006), when compared to normalmucosas or to diffuse/mixed type tumours (p < 0.012;Figure 3B).

From 13 tumours with EZH2 over-expression (nineintestinal, four diffuse/mixed), 12 showed concomitantmature miR-101 decreased expression levels (nineintestinal, three diffuse/mixed) and of these nine hadproven DNA copy number loss at one and/or both miR-101 loci (seven intestinal, two diffuse/mixed; Table 1).

In summary, concomitant EZH2 over-expression andmiR-101 down-regulation is a de novo feature pre-sented by GC cases, mainly of the intestinal histotype.

EZH2 and CDH1/E-cadherin mRNA and proteinexpression levels are inversely correlated in GCCDH1/E-cadherin down-regulation is one of the out-comes of EZH2 over-expression in several typesof cancer including GC [22], therefore we analysedwhether in our series of primary GC this correlationwas maintained and further associated with miR-101down-regulation. To assess basal values of EZH2 andCDH1 mRNA expression in normal stomach, we


J Carvalho et al

Tabl

e1.

Mec

hani

sms

invo

lved

inE-

cadh

erin

regu

latio

nin

29ga

stric

canc

ers

E-ca

dher

inex

pres

sion

Clas

sica

lCDH

1m

iR-1

01-1

miR

-101

-1m

iR-1

01-2

miR

-101

-2m

iR-1

01EZ

H2

CDH

1EZ

H2/

CDH

1La

uren

'sTu

mou

r ID

scor

e0,

1,2

inac

tivat

ing

mec

hani

sms

aCGH

SNaP

shot

aCGH

SNaP

shot

RNA

mRN

Am

RNA

mRN

Acl

assi

ficat

ion

2aAb

erra

nt1

ND

No

loss

Loss

Loss

Loss

Inte

stin

al3

Bial

lelic

met

hyla

tion

No

loss

No

loss

Loss

Loss

Inte

stin

al10

Abse

nt0

Abse

nt0

Met

hyla

tion*

No

loss

Loss

Loss

Loss

Inte

stin

al13

Aber

rant

1N

egat

ive

No

loss

Loss

Loss

Loss

Inte

stin

al15

Aber

rant

1LO

H*

No

loss

Loss

Loss

Loss

Diff

use

19Ab

sent

0N

egat

ive

No

loss

No

loss

No

loss

Loss

Inte

stin

al21

Aber

rant

1N

egat

ive

No

loss

Loss

No

loss

Loss

Mix

ed27

bAb

sent

0N

egat

ive

No

loss

Loss

No

loss

Loss

Inte

stin

al28

Aber

rant

1N

egat

ive

No

loss

Loss

No

loss

Loss

Inte

stin

al29

Abse

nt0

Neg

ativ

eN

olo

ssN

DN

olo

ssN

DIn

test

inal

30Ab

erra

nt1

Neg

ativ

eN

olo

ssN

olo

ssN

olo

ssN

olo

ssIn

test

inal

18Ab

erra

nt1

Neg

ativ

eN

olo

ssN

DN

olo

ssN

DDi

ffus

e9d

Aber

rant

1N

egat

ive

No

loss

ND

Loss

ND

Diff

use

20Ab

erra

nt1

Mut

atio

n/m

ethy

latio

n/LO

HN

olo

ssLo

ssN

olo

ssLo

ssN

EDi

ffus

e12

Abse

nt0

Abse

nt0

Abse

nt0

Abse

nt0

Neg

ativ

eN

olo

ssN

DLo

ssN

DN

EM

ixed

5N

egat

ive

No

loss

ND

Loss

ND

NE

NE

Inte

stin

al6

Neg

ativ

eG

ain

No

loss

Loss

Loss

NE

NE

Diff

use

7LO

HN

olo

ssN

DLo

ssN

DN

EM

ixed

8cAb

erra

nt1

Neg

ativ

eN

olo

ssN

DLo

ssN

DN

EDi

ffus

e11

Aber

rant

1N

egat

ive

No

loss

ND

Loss

ND

NE

Inte

stin

al14

Aber

rant

1M

ethy

latio

nN

olo

ssN

DLo

ssN

DN

EM

ixed

16Ab

erra

nt1

Neg

ativ

eN

olo

ssN

DLo

ssN

DN

EDi

ffus

e17

Abse

nt0

Mut

atio

nN

olo

ssN

DN

olo

ssN

DN

EN

EM

ixed

22Ab

erra

nt1

Neg

ativ

eN

olo

ssN

DN

olo

ssN

DN

EIn

test

inal

24Ab

sent

0N

egat

ive

No

loss

ND

No

loss

ND

NE

Mix

ed36

dAb

erra

nt1

Neg

ativ

eN

olo

ssN

DN

olo

ssN

DN

EIn

test

inal

37c

Aber

rant

1N

egat

ive

No

loss

ND

No

loss

ND

NE

Inte

stin

al34

aN

orm

al2

ND

No

loss

ND

No

loss

ND

NE

Diff

use

35a

Nor

mal

2N

DN

olo

ssN

DN

olo

ssN

DN

EDi

ffus

e

↓+ + + + + + + + + +

+ + + + + + + + + + + + + + +

+ + + + + + + + + + + + + − − − − − − − − − − − − − − − −+ + + + + + + + + + +

+ + + + + + + + + + +

+ + +θ

↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓

E-ca

dher

inex

pres

sion

scor

ing:

0,de

fined

abse

nce

ofex

pres

sion

or10

%of

E-ca

dher

in-p

ositi

vece

lls,i

ndep

ende

ntly

ofth

ece

llula

rlo

caliz

atio

nof

the

prot

ein;

1,de

fines

10–

90%

ofpo

sitiv

e-st

aine

dce

llsw

ithab

norm

alE-

cadh

erin

subc

ellu

lar

loca

lizat

ion;

and

2,de

fines

expr

essi

onpo

sitiv

ityat

the

cell

mem

bran

ein

>90

%of

cells

.∗Ca

ses

reta

inin

gal

lele

(s)

appa

rent

lyun

targ

eted

bycl

assi

calC

DH1

inac

tivat

ing

mec

hani

sms.

miR

-101

:↓,

dow

n-re

gula

ted;

↑,up

-reg

ulat

ed;

EZH

2:+,

over

-exp

ress

ed;+

θ,E

ZH2’

slo

cus

(chr

omos

ome

7q36

.1)w

asam

plifi

edby

aCG

H;N

E,no

tex

pres

sed;

CDH

1:+,

expr

esse

d;N

E,no

tex

pres

sed;

EZH

2/CD

H1:

++,h

ighe

rEZ

H2

expr

essi

onle

vels

than

CDH

1;+,

equi

vale

ntor

low

erEZ

H2

expr

essi

onle

vels

than

CDH

1;−,

EZH

2an

d/or

CDH

1w

ere

not

expr

esse

d.a C

lass

ical

CDH

1in

activ

atin

gm

echa

nism

sno

tas

sess

ed.b M

utat

ions

not

asse

ssed

.c LO

Hno

tas

sess

ed.d M

utat

ions

and

LOH

not

asse

ssed

;ND,

not

dete

rmin

ed



analysed nine samples from cancer-free individuals.While EZH2 was residually expressed, CDH1 dis-played high expression levels (p < 0.004; Figure 4A).We then compared EZH2 and CDH1 mRNA expres-sion levels in 29 primary GCs. To infer whetherEZH2 was negatively associated with CDH1 mRNAexpression levels, we analysed separately cases dis-playing EZH2 over-expression (n = 13, fold-change>2 compared with normal gastric mucosas; Figure 4B,Table 1) and cases without EZH2 expression (n = 16,fold-change <2 compared with normal gastric mucosas(Figure 4C, Table 1). Eighty-five percent (11/13) ofthe tumours with EZH2 over-expression displayedlower levels of CDH1 expression than those of EZH2(Figure 4B, Table 1: EZH2/CDH1 mRNA++), sug-gesting that EZH2 could be regulating CDH1 RNAlevel. Only two tumours with EZH2 over-expressionshowed higher and equivalent levels of CDH1 thanEZH2 (Figure 4B, grey dots; Table 1, EZH2/CDH1mRNA+). These 13 cases displayed either absent (n =5) or aberrant (n = 8) E-cadherin protein expressionby immunohistochemistry (Figure 4B1, Table 1). Fromthe 16 tumours lacking EZH2 expression, 13 expressedCDH1 mRNA (E-cadherin protein expression: eightaberrant, three absent, two normal) and three did not(Figure 4C, 4C.1, Table 1).

To further understand the correlation between EZH2and CDH1 mRNA expression levels, we calculatedthe ratio between these two molecules and observedthat the ‘EZH2 over-expression’ group was signif-icantly different from the ‘non-GC tissues’ (p <

0.0002) and from the ‘EZH2 without expression’ group(p < 0.0008), thus reinforcing our previous strategyto separate GC cases according to EZH2 expression(Figure 4D).

To analyse this correlation from the protein expres-sion standpoint, we performed co-immunofluorescenceof both molecules in primary GC whole tissue sections(Figure 4E). This would further allow us to clarifywhether increased nuclear localization of EZH2 proteinwould be associated with decreased E-cadherin expres-sion and/or mislocalization, as previously described.[22] We verified that GC cases with high EZH2nuclear expression invariably showed low expres-sion or mislocalization of E-cadherin, reflecting theresults obtained by RNA expression analysis as wellas by immunohistochemistry (Figure 4E, GC_I andGC_II).

The pathway miR-101–EZH2 explains thedisruption of E-cadherin expression mainly in GCcases that retain an allele untargeted by classicalCDH1 inactivating mechanismsWhen analysing dysfunction of the E-cadherin tumour-suppressor gene in cancer, it is important to keep inmind the existence of inactivating classical genetic andepigenetic mechanisms, as well as the tumours’ histo-logical types. After performing the screening of muta-tions, promoter methylation and LOH, we explored

whether down-regulation of miR-101 with consequentEZH2 over-expression constitute an additional mecha-nism disrupting CDH1/E-cadherin expression/functionin GC or, on the other hand, synergizes with classicalCDH1-inactivating mechanisms.

Classical CDH1-inactivating mechanisms were iden-tified in 7/26 (26.9%) GC cases available for anal-ysis: promoter hypermethylation in three, mutationin one, LOH in two and all alterations togetherin one case (Table 1; see also Supporting informa-tion, Table S2, Figure S3). Importantly, while inmost GC cases a single CDH1 allele seemed tobe affected, in two cases biallelic CDH1 inactiva-tion was identified (Table 1): case 3, an intestinaltype GC displayed biallelic CDH1 promoter hyper-methylation; and case 20, a diffuse type GC show-ing concomitantly a splice-site CDH1 mutation, pro-moter hypermethylation and LOH, probably reflect-ing intra-tumoural heterogeneity previously describedin HDGC lesions. [7] The molecular and immuno-histochemistry results were concordant for case 3,which completely lacked E-cadherin expression, andfor case 20, which only showed 20% of positivetumour cells but with aberrant E-cadherin expression(see Supporting information, Figure S3). By crossingthe above results with those obtained for the path-way involving miR-101–EZH2, we verified that con-comitant miR-101 down-regulation and EZH2 over-expression (higher than CDH1 ) occurred preferentiallyin cases retaining allele(s) apparently untargeted byclassical CDH1-inactivating mechanisms (n = 9), sug-gesting that this pathway is a player in the disturbanceof CDH1/E-cadherin expression and function, on itsown, in approximately 82% (9/11) of these GC cases(upper grey part of Table 1).

The miR-101–EZH2–E-cadherin pathway ispreferentially associated with intestinal-type GC

We integrated all data in 29 GCs (Table 1). In 38%(11/29) of the cases, we observed concomitant miR-101down-regulation, inverse correlation between EZH2and CDH1 mRNA levels and aberrant or absentE-cadherin protein expression (upper grey part ofTable 1). In this setting, 7/10 cases had proven genomicloss of both miR-101-1 and miR-101-2 loci, while 2/10cases displayed genomic loss only at the miR-101-2locus, causing mature miR-101 down-regulation. Theremaining 62% (18/29) of the cases did not show suchassociation.

Importantly, most GC cases associated with the miR-101–EZH2–E-cadherin pathway were of the intestinalhistotype (p = 0.0259), while most of those apparentlyunrelated to this pathway were of the diffuse and mixedtypes (Table 1). This result demonstrates a novel cas-cade of genetic events leading to E-cadherin disruption,preferentially affecting the intestinal type of GC.


J Carvalho et al

Normal Gastric Mucosas4

2

0

Log 1

0 re

lativ

e m

RN

A e

xpre

ssio

n

Log 1

0 R

NA

rel

ativ

e ex

pres

sion

rat

io

-2

-4

4

2

0

Log 1

0 E

ZH

2/C

DH

1 re

lativ

e ex

pres

sion

-2

-4Non-GC tissues EZH2

EZH2

Nor

mal

Gas

tric

Muc

osa

GC

_IG

C_I

IG

C_I

II

E-cadherin MergedDapi

EZH2over-expression without expression

EZH2 CDH1 EZH2-4.0

-2.0

0.0

2.0

4.0

Log 1

0 m

RN

A r

elat

ive

expr

essi

on r

atio

-4.0

-2.0

0.0

2.0

4.0

over-expression

E-cadherin expression

Absent Aberrant

CDH1 EZH2without expression

Normal E-cadherin expression

CDH1

p > 0.12

p < 0.0008p < 0.0002

p < 0.004

A

D

B.1 C.1

E

B C

50 μm

50 μm

50 μm

50 μm50 μm

50 μm

50 μm

50 μm50 μm

50 μm

50 μm

50 μm50 μm

50 μm

50 μm

50 μm

Figure 4. EZH2 and CDH1/E-cadherin are inversely correlated in GC. (A) Box plots represent EZH2 and CDH1 mRNA expression levels relativeto 18s in nine normal gastric mucosas. Statistically significant p value was obtained by using the Mann–Whitney test. (B) Correlationbetween EZH2 over-expression and CDH1 mRNA expression in GCs; dark dots represent cases with higher EZH2 mRNA expression levelsthan CDH1; grey dots represent cases with CDH1 mRNA expression higher than or equal to EZH2. (B.1) Representative examples of absentor aberrant E-cadherin expression in GC cases displaying higher EZH2 mRNA expression levels than CDH1. (C) Correlation between EZH2without expression and CDH1 mRNA expression in GCs; dark dots represent cases with no expression of EZH2 and with higher levels ofCDH1; grey dots represent GCs without CDH1 expression. (C.1) Representative example of a GC with normal E-cadherin expression. (B,C)Values were normalized to the expression levels observed in a panel of nine normal gastric mucosas. (D) Box plots represent the ratio ofEZH2:CDH1 mRNA expression levels in the ‘non-GC-tissues’, ‘EZH2 over-expression’ and ‘EZH2 without expression’ groups. Statisticallysignificant p value was obtained by using the Mann–Whitney test. (E) Co-immunofluorescence of EZH2 (red staining) and E-cadherin(green staining) in normal stomach and GCs. Blue staining shows nuclear staining with DAPI. Upper panel shows normal stomach lackingEZH2 nuclear protein expression and displaying high expression levels of E-cadherin at the membrane. GC_I shows an intestinal typeGC with high nuclear expression levels of EZH2 and low and mislocalized E-cadherin expression. GC_II shows a diffuse GC with EZH2high nuclear expression levels and high but mislocalized E-cadherin expression levels. GC_III shows a diffuse GC lacking EZH2 nuclearexpression and displaying low and mislocalized E-cadherin expression levels. GC_I, II and III are cases retaining allele(s) untargeted byclassical CDH1-inactivating mechanisms.



6

4

2

Rel

ativ

e R

NA

exp

ress

ion

leve

ls

3

2

1

0

Kato-MiR

-NC

Kato-MiR

-101

Kato-MiR

-NC

Kato-MiR

-101

Kato-MiR

-NC

Kato-MiR

-101

Kato-MiR

-NC

Kato-MiR

-NC

Kato-MiR

-101

Kato-MiR

-101

Kato-MiR

-101

Kato-MiR

-NC

Kato-MiR

-NC

Kato-MiR

-101

MiR-N

C

MiR-101

Rel

ativ

e R

NA

exp

ress

ion

leve

ls

1.5

1.0

0.5

0.0

Kato-MiR-NC

Number of positive cellsFluorescence Intensity

FL1 LOG

104103102101100

030

0E

vent

s

Kato-MiR-101

Rel

ativ

e pr

otei

n ex

pres

sion

leve

ls

0

-2EZH2

EZH2

miR

-NC

miR

-101

E-cadherin Merged

MiR-101 CDH1

MiR-101

EZH2

a-Tubulin

a-Tubulin

E-cadherin

EZH2CDH1

EZH2CDH1

2.5

2.0

1.5

E-c

adhe

rin s

urfa

ce e

xpre

ssio

n

1.0

0.5

0.0

A

D E

B C

Figure 5. MiR-101 rescues E-cadherin membrane localization through EZH2 inhibition. RNA expression levels of miR-101, EZH2 and CDH1were analysed by QRT–PCR in KatoIII cell line (A), Kato-MiR-NC and Kato-MiR-101 cells (B); RNU48 and GAPDH were used as endogenouscontrols. (C) EZH2 and E-cadherin were detected in whole Kato-MiR-NC and Kato-MiR-101 cell lysates by WB. (D) EZH2 (red staining) andE-cadherin (green staining) cellular localization in Kato-MiR-NC and Kato-MiR-101 cells was assessed by immunofluorescence; imageswere acquired with ×400 amplification, keeping the same setting parameters. (E) E-cadherin cell surface expression was assessed byFCM. Each histogram represents 50 000 cell counts and the cell surface expression of E-cadherin in Kato-MiR-NC cells (black line) andKato-MiR-101 cells (grey line); black area represents the cell surface expression of E-cadherin in the cells stained only with secondaryantibody (negative control). All graphics show mean ± SEM of three independent experiments (∗p < 0.05, Student’s t-test).

Induced over-expression of miR-101 leads toincreased membranous E-cadherin localizationthrough EZH2 inhibition

We used KatoIII cells to test whether induced miR-101 over-expression could promote EZH2 inhibi-tion and rescue membranous E-cadherin expression.KatoIII retains one CDH1 allele untargeted by classi-cal inactivating mechanisms [45], displays low expres-sion of mature miR-101 (Figure 1D), high EZH2mRNA and protein levels, low CDH1 mRNA levels(Figure 5A; see also Supporting information, FigureS4) but measurable E-cadherin levels (Figure 5D),rarely found at cell–cell contacts, therefore turning itin a good model to test the recovery of E-cadherinexpression/function.

We stably expressed miR-101 and a negative con-trol in KatoIII (Kato-miR-101 and Kato-miR-NC).The miR-101 gain of function, but not miR-NC, ledto a reduction of endogenous EZH2 mRNA lev-els, although not statistically significant (p > 0.11;Figure 5B), but to a robust reduction (50%) of endoge-nous EZH2 protein levels (p < 0.0005; Figure 5C),suggesting that miR-101 preferentially regulates EZH2

protein levels rather than the corresponding mRNA.We analysed the E-cadherin cellular localization qual-itatively by immunofluorescence and quantitatively byFCM to infer whether miR-101 over-expression couldrescue E-cadherin protein membranous localization.For this, we used an antibody that targets E-cadherinamino acid residues that lie downstream of the pre-mature termination codon present in KatoIII CDH1mutant allele, which is expected to label E-cadherinwild-type full-length molecules only. Immunofluores-cence showed that E-cadherin became more intenselyexpressed and localized at the cell membrane andcell–cell contacts, with a measurable two-fold higherintensity of E-cadherin potentially functional mole-cules, as shown by FCM analysis, in Kato-miR-101 comparing with Kato-miR-NC cells (p < 0.025;Figure 5D, E). The number of E-cadherin-positive cellsdid not vary between Kato-miR-101 and Kato-miR-NCcells (Figure 5E).

Interestingly, CDH1 mRNA levels were onlymodestly up-regulated (Figure 5B), and protein expres-sion levels remained unaffected upon miR-101 over-expression (Figure 5C).


J Carvalho et al

1.5 EZH2

EZH2

a-Tubulin

a-Tubulin

E-cadherin

CDH1

EZH2CDH1

siControlsiEZH2

1.0

Rel

ativ

e m

RN

A e

xpre

ssio

n le

vels

0.5

0.0

1.5

1.0

Rel

ativ

e pr

otei

n ex

pres

sion

leve

ls

Eve

nts

0.5

3

2

E-c

adhe

rin s

urfa

ce e

xpre

ssio

n

1

0

siCon

trol

siCon

trol

siEZH2

siEZH2

0.0

siContro

l

siContro

l

siEZH2

siEZH2

siContro

l

siContro

l

siEZH2

siEZH2

EZH2

siC

ontr

olsi

EZ

H2

E-cadherin

siContro

l

siEZH2

Merged

170

0

FL1 LOG

Number of positive cellsFluorescence Intensity

104103102101100

20 μm20 μm20 μm

20 μm 20 μm 20 μm

A

C

B D

Figure 6. EZH2 inhibition by siRNA leads to increased membranous E-cadherin localization. (A) EZH2 and CDH1 mRNA expression levels ofsiControl and siEZH2 cells were analysed by QRT–PCR; GAPDH was used as endogenous control. (B) EZH2 and E-cadherin were detected inwhole siControl and siEZH2 cell lysates by WB. (C) EZH2 (red staining) and E-cadherin (green staining) cellular localization in siControl andsiEZH2 cells was assessed by immunofluorescence. Images were acquired with ×400 amplification, keeping the same setting parameters.(D) E-cadherin cell surface expression was assessed by FCM. Each histogram represents 50 000 cell counts and the cell surface expressionof E-cadherin in siControl cells (black line) and in siEZH2 cells (grey line). The black area represents the cell surface expression of E-cadherinin the cells stained only with secondary antibody (negative control). The graphics show mean ± SEM of three independent experiments(∗p < 0.05, Student’s t-test).

To prove that miR-101 expression rescues mem-branous localization of E-cadherin via EZH2 down-regulation, EZH2 was transiently depleted in KatoIIIcells by RNAi (Kato-siEZH2). A reduction of 70%in EZH2 transcript levels (p < 1.2 × 10−6) led toincrease of CDH1 mRNA level (p < 0.02; Figure 6A),and co-immunofluorescence and FCM analysis con-firmed that E-cadherin was recovered at the plasmamembrane in EZH2-depleted cells compared tonon-silencing siRNA cells (p < 0.028; Figure 6C,D). Similarly to what was observed with Kato-miR-101 over-expressing cells, WB analyses demon-strated that EZH2 expression was attenuated byEZH2-siRNA-induced depletion (p < 9.6 × 10−6),while E-cadherin levels remained unchanged(Figure 6B).

These results suggest that miR-101 exerts its activ-ity over EZH2 preferentially at the post-transcriptionallevel and demonstrate that miR-101- and EZH2-dependent signals similarly regulate E-cadherin cellularlocalization, likely using the same signalling pathway.

Discussion

The mechanisms leading to E-cadherin mRNA andprotein expression deficit and mislocalization in can-cer cells are far from being identified. In the presentstudy, we aimed to identify and characterize a novelcascade of events leading to E-cadherin dysfunction,using GC patients’ samples and in vitro cell models.Our data allow us to propose a pathway by which miR-101 and the histone methyltransferase EZH2 contributeto E-cadherin dysfunction in the specific subtype ofintestinal GC (Figure 7).

The finding that miR-101 was down-regulated inprimary GC and GC cell lines, together with similarreports in this and other types of cancer [35,36,38,46],establishes loss of mature miR-101 as a widespreadevent in cancer. The mature form of miR-101 isencoded by two different genes, miR-101-1 and miR-101-2, located at chromosomes 1p31.3 and 9p24.1,respectively, which were reported to be somaticallylost in prostate cancer [37]. By adopting a validationstrategy of the aCGH data, we have noticed that not



A Normal context

miR-101-1

miR-101-2

EZH2 mRNA

Translationinhibition

Gastric cancerB

3′UTR

Chr

1

Chr

9

miR-101-2 loss

3′UTR

EZH2

EZH2

EZH2EZH2 mRNA

EZH2 mRNA

Tran

slat

ion

EZH2 mRNA EZH2

miR-101-1 loss

Chr

1

Chr

9

E-cadherin

E-cadherin

Figure 7. Proposed model for the pathway involving miR-101loss, up-regulation of EZH2 histone methyltransferase and E-cadherin dysfunction in GC. (A) In a normal context, maturemiR-101 binds to the 3′-UTR of EZH2, inhibiting its translation.In consequence, no EZH2 protein is expressed to impair CDH1/E-cadherin expression and function. E-cadherin becomes expressedand correctly localized at the cell membrane. (B) In a proportionof GC cases, deletions/microdeletions at the miR-101 loci lead todown-regulation of mature miR-101 and release EZH2 from miR-101-mediated repression. As a consequent, up-regulation of EZH2protein is able to impair E-cadherin function and membranouslocalization.

only deletions at miR-101-2 but also microdeletions atthis and at miR-101-1 loci, previously undetected bythe aCGH platform, are among the underlying mecha-nisms for miR-101 down-regulation. Nevertheless, wehave to keep in mind that, aside from structural alter-ations (chromosomal micro/deletions), other geneticand/or epigenetic mechanisms are likely involved in thestrong and overall down-regulation of miR-101 in GC.Promoter hypermethylation [47–49] and regulation bytranscription factors, such as HIF1α/HIF1β [39], mightalso be implicated. Moreover, a recently recognizedrelationship between EZH2 and the histone demethy-lase NDY1, that together bind miR-101 promoters andrepress their expression, may also be involved [50].However, this process can only occur if NDY1 isover-expressed [50], and this was not addressed in thepresent study. The repression of miR-101 in multipletypes of cancers leads to the up-regulation of its targetgenes, including some oncogenes [35,46,51,52].

EZH2 has been extensively studied in prostate[37,39], bladder [38], liver [53], and head and neck[17] cancers and also in GC [36]. Moreover, an array

of EZH2-target genes have been identified, includ-ing rap1GAP [17], ADRB2 [54], RUNX3 [55], andCDH1 [21,22]. However, to the best of our knowl-edge, the circuitry miR-101–EZH2–E-cadherin, actingto maintaining E-cadherin function integrity in epithe-lial cells, has not yet been described. Our in vitro datademonstrated for the first time that over-expression ofmature miR-101 in GC cells leads to strong deple-tion of endogenous EZH2 protein without significantlyaffecting EZH2 mRNA levels, and enables EZH2-depleted cells to increase the frequency of E-cadherinmolecules at the cell membrane. By specifically inhibit-ing EZH2 without interfering directly with miR-101expression, we were able to show the exact same effectin E-cadherin membranous localization, demonstratinga cause–consequent relationship between miR-101 andE-cadherin, which is intermediated by EZH2. Apartfrom proving the connection between these three play-ers, we further pinpoint a layer of miR-101 reg-ulation that preferentially affects protein expressionlevels (EZH2) and correct subcellular localization (E-cadherin). Although somehow similar results have beenpreviously obtained [22] for the effect of EZH2 over E-cadherin, the dependence on miR-101 deletion has not.

These in vitro results nicely validated the correla-tions that we obtained in clinical GC samples. Weobserved that near 40% of GC cases, herein stud-ied, displayed miR-101 loci deletion, mature miR-101expression loss and concomitant EZH2 up-regulation(mRNA and protein), and E-cadherin (mRNA andprotein) dysfunction. More importantly, our studyallowed the stratification of GC cases in two maingroups, one of which follows the miR-101–EZH2–E-cadherin pathway, encompassing mainly intestinal-typetumours, and the other one not. This is interest-ing because, despite presenting E-cadherin dysfunc-tion preferentially at latter stages of progression [56],intestinal-type GC remained without a strong mech-anism underlying this phenotype. Although a previ-ous report has claimed that E-cadherin dysfunction inintestinal-type GC would depend greatly on a pathwayinvolving EZH2 and E-cadherin [22], this report lackedthe study of the driver miR-101 loss effect as well asthe analysis of GC cases of other histological types.

The group of GC cases herein analysed that didnot follow the miR-101–EZH2–E-cadherin pathway,despite showing some degree of miR-101-2 DNA loss,miR-101 down-regulation and E-cadherin dysfunction,it lacked EZH2 over-expression. In this group, eitherthe EZH2 gene is inactivated, as described in myeloidmalignancies where loss-of-function mutations havebeen identified [57,58], or the impact of miR-101expression loss is probably affecting other target genes,as demonstrated in other carcinomas [34,35,46], whicheventually may also affect E-cadherin expression andfunction.

We have systematically analysed GC cases for clas-sical CDH1 genetic and epigenetic alterations. Inter-estingly, the pathway miR-101–EZH2, on its own,


J Carvalho et al

occurs preferentially in cases that retain CDH1 wild-type alleles but present great disturbance of E-cadherinexpression and function. Our data reinforce the criticalrole of E-cadherin in GC and pinpoint EZH2 over-expression, as a consequence of miR-101 loci dele-tion and mature miR-101 down-regulation, as one ofthe missing mechanisms for E-cadherin inactivation inadvanced intestinal GC. This strong association sug-gests that therapeutic intervention based on restorationof miR-101 or pharmacological inhibition of EZH2may have clinical benefit in patients with intestinal-type GC.

Acknowledgment

The authors thank: Patrıcia Oliveira for technical assis-tance with statistical analysis and image production;Sara Ricardo and Joana Paredes for technical helpwith IHC; Marta Correia and Jose Carlos Machado fordiscussion; and all the members of Carla Oliveira’sand Beatriz Carvalho’s laboratories. This study wassupported by: the Portuguese Foundation for Sci-ence and Technology (FCT; Project Nos PTDC/SAU-GMG/72168/2006 and PTDC/BIA-BCM/64745/2006,and PhD Grant No. SFRH/BD/44074/2008-JC); salarysupport to CO from the European Social Fundand Portuguese Ministry of Science and Technol-ogy (MCTES; Grant No. POPH–QREN/Type 4.2);an EMBO short-term fellowship (No. ASTF 338.00-2008-JC); a Calouste Gulbenkian Foundation short-term fellowship (No. 92352-JC); and an IPATIMUPinternal project, ‘The involvement of microRNAs ingastric cancer’. IPATIMUP is an Associate Laboratoryof the Portuguese Ministry of Science, Technology andHigher Education and is partially supported by FCT.

Author contributions

JC, study design, acquisition of data, analysis andinterpretation of data and drafting of the manuscript;NCVG, acquisition of biological material and analy-sis of data; PMP, acquisition and analysis of data; SS,acquisition of data and technical support; MT, acquisi-tion of biological material and technical support; TEB,acquisition of biological material and aCGH data; BD,critical revision of the manuscript for important intel-lectual content; HG, acquisition of biological material;MASS, critical revision of the manuscript for impor-tant intellectual content; GM, critical revision of themanuscript for important intellectual content; RS, criti-cal revision of the manuscript for important intellectualcontent; BC, critical revision of the manuscript forimportant intellectual content and supervision of partof the work; CO, study design, analysis and interpre-tation of data, critical revision of the manuscript forimportant intellectual content, supervision of all workand fund raising.

Abbreviations

aCGH, array comparative genomic hybridization;ATCC, American Type Culture Collection; BAC, bac-terial artificial chromosome; CDH1, E-cadherin gene;ECL, enhanced chemiluminescence; EZH2, enhancerof Zeste homologue 2; FCM, flow cytometry; FFPE,formalin-fixed paraffin-embedded; GEO, gene expres-sion omnibus; GC, gastric cancer; HDGC, heredi-tary diffuse gastric cancer; LOH, loss of heterozygos-ity; miRNAs, microRNAs; miR-NC, microRNA neg-ative control; mRNA, messenger RNA; PBLs, periph-eral blood lymphocytes; PRC2, polycomb repressorcomplex 2; qRT–PCR, quantitative real-time poly-merase chain reaction; SEM, standard error of themean; siRNA, small interference RNA

References

1. Parkin DM, Bray F, Ferlay J, et al . Global cancer statistics, 2002.CA Cancer J Clin 2005; 55: 74–108.

2. Khushalani N. Cancer of the esophagus and stomach. Mayo ClinProc 2008; 83: 712–722.

3. Carneiro F, Oliveira C, Leite M, et al . Molecular targets andbiological modifiers in gastric cancer. Semin Diagn Pathol 2008;25: 274–287.

4. Guilford P, Hopkins J, Harraway J, et al . E-cadherin germlinemutations in familial gastric cancer. Nature 1998; 392: 402–405.

5. Grady WM, Willis J, Guilford PJ, et al . Methylation of the CDH1promoter as the second genetic hit in hereditary diffuse gastriccancer. Nat Genet 2000; 26: 16–17.

6. Oliveira C, Senz J, Kaurah P, et al . Germline CDH1 deletions inhereditary diffuse gastric cancer families. Hum Mol Genet 2009;18: 1545–1555.

7. Oliveira C, Sousa S, Pinheiro H, et al . Quantification of epige-netic and genetic 2nd hits in CDH1 during hereditary diffusegastric cancer syndrome progression. Gastroenterology 2009; 136:2137–2148.

8. Oliveira C, Bordin MC, Grehan N, et al . Screening E-cadherinin gastric cancer families reveals germline mutations only inhereditary diffuse gastric cancer kindred. Hum Mutat 2002; 19:510–517.

9. Liu YC, Shen CY, Wu HS, et al . Mechanisms inactivating thegene for E-cadherin in sporadic gastric carcinomas. World JGastroenterol 2006; 12: 2168–2173.

10. Machado JC, Oliveira C, Carvalho R, et al . E-cadherin gene(CDH1 ) promoter methylation as the second hit in sporadic diffusegastric carcinoma. Oncogene 2001; 20: 1525–1528.

11. Machado JC, Soares P, Carneiro F, et al . E-cadherin gene muta-tions provide a genetic basis for the phenotypic divergence ofmixed gastric carcinomas. Lab Invest 1999; 79: 459–465.

12. Berx G, Becker KF, Hofler H, et al . Mutations of the humanE-cadherin (CDH1 ) gene. Hum Mutat 1998; 12: 226–237.

13. Becker KF, Atkinson MJ, Reich U, et al . E-cadherin gene muta-tions provide clues to diffuse type gastric carcinomas. Cancer Res1994; 54: 3845–3852.

14. Karayiannakis AJ, Syrigos KN, Chatzigianni E, et al . E-cadherinexpression as a differentiation marker in gastric cancer. Hepato-gastroenterology 1998; 45: 2437–2442.

15. von Burstin J, Eser S, Paul MC, et al . E-cadherin regulatesmetastasis of pancreatic cancer in vivo and is suppressed bya SNAIL/HDAC1/HDAC2 repressor complex. Gastroenterology2009; 137: 361–371, e361–365.



16. Ma L, Young J, Prabhala H, et al . miR-9, a MYC/MYCN-

activated microRNA, regulates E-cadherin and cancer metastasis.

Nat Cell Biol 2010; 12: 247–256.

17. Banerjee R, Mani RS, Russo N, et al . The tumor suppressor gene

rap1GAP is silenced by miR-101-mediated EZH2 overexpres-

sion in invasive squamous cell carcinoma. Oncogene 2011; 30:

4339–4349.

18. Cao R, Wang L, Wang H, et al . Role of histone H3 lysine 27

methylation in Polycomb-group silencing. Science 2002; 298:

1039–1043.

19. Kuzmichev A, Nishioka K, Erdjument-Bromage H, et al . Histone

methyltransferase activity associated with a human multiprotein

complex containing the Enhancer of Zeste protein. Genes Dev

2002; 16: 2893–2905.

20. Cao R, Zhang Y. The functions of E(Z)/EZH2-mediated methyla-

tion of lysine 27 in histone H3. Curr Opin Genet Dev 2004; 14:

155–164.

21. Cao Q, Yu J, Dhanasekaran SM, et al . Repression of E-cadherin

by the polycomb group protein EZH2 in cancer. Oncogene 2008;

27: 7274–7284.

22. Fujii S, Ochiai A. Enhancer of Zeste homolog 2 downregulates E-

cadherin by mediating histone H3 methylation in gastric cancer

cells. Cancer Sci 2008; 99: 738–746.

23. Varambally S, Dhanasekaran SM, Zhou M, et al . The polycomb

group protein EZH2 is involved in progression of prostate cancer.

Nature 2002; 419: 624–629.

24. Bachmann IM, Halvorsen OJ, Collett K, et al . EZH2 expression

is associated with high proliferation rate and aggressive tumor

subgroups in cutaneous melanoma and cancers of the endometrium,

prostate, and breast. J Clin Oncol 2006; 24: 268–273.

25. Matsukawa Y, Semba S, Kato H, et al . Expression of the enhancer

of zeste homolog 2 is correlated with poor prognosis in human

gastric cancer. Cancer Sci 2006; 97: 484–491.

26. Kidani K, Osaki M, Tamura T, et al . High expression of EZH2 is

associated with tumor proliferation and prognosis in human oral

squamous cell carcinomas. Oral Oncol 2009; 45: 39–46.

27. Rao ZY, Cai MY, Yang GF, et al . EZH2 supports ovarian carci-

noma cell invasion and/or metastasis via regulation of TGFβ1 and

is a predictor of outcome in ovarian carcinoma patients. Carcino-

genesis 2010; 31: 1579–1583.

28. Collett K, Eide GE, Arnes J, et al . Expression of enhancer of Zeste

homologue 2 is significantly associated with increased tumor cell

proliferation and is a marker of aggressive breast cancer. Clin

Cancer Res 2006; 12: 1168–1174.

29. Wagener N, Holland D, Bulkescher J, et al . The enhancer of Zeste

homolog 2 gene contributes to cell proliferation and apoptosis

resistance in renal cell carcinoma cells. Int J Cancer 2008; 123:

1545–1550.

30. Fussbroich B, Wagener N, Macher-Goeppinger S, et al . EZH2

depletion blocks the proliferation of colon cancer cells. PLoS One

6: e21651.

31. Li X, Zhang Y, Ding J, et al . Survival prediction of gastric cancer

by a seven-microRNA signature. Gut 2009; 59: 579–585.

32. Ueda T, Volinia S, Okumura H, et al . Relation between

microRNA expression and progression and prognosis of gastric

cancer: a microRNA expression analysis. Lancet Oncol 2010; 11:

136–146.

33. Smits M, Nilsson J, Mir SE, et al . miR-101 is down-regulated in

glioblastoma resulting in EZH2-induced proliferation, migration,

and angiogenesis. Oncotarget 1: 710–720.

34. Strillacci A, Griffoni C, Sansone P, et al . MiR-101 downregula-

tion is involved in cyclooxygenase-2 overexpression in human

colon cancer cells. Exp Cell Res 2009; 315: 1439–1447.

35. Su H, Yang JR, Xu T, et al . MicroRNA-101, down-regulatedin hepatocellular carcinoma, promotes apoptosis and suppressestumorigenicity. Cancer Res 2009; 69: 1135–1142.

36. Wang HJ, Ruan HJ, He XJ, et al . MicroRNA-101 is down-regulated in gastric cancer and involved in cell migration andinvasion. Eur J Cancer 2010; 46: 2295–2303.

37. Varambally S, Cao Q, Mani RS, et al . Genomic loss ofmicroRNA-101 leads to overexpression of histone methyltrans-ferase EZH2 in cancer. Science 2008; 322: 1695–1699.

38. Friedman JM, Liang G, Liu CC, et al . The putative tumor suppres-sor microRNA-101 modulates the cancer epigenome by repress-ing the polycomb group protein EZH2. Cancer Res 2009; 69:2623–2629.

39. Cao P, Deng Z, Wan M, et al . MicroRNA-101 negatively regu-lates Ezh2 and its expression is modulated by androgen receptorand HIF-1α/HIF-1β. Mol Cancer 2010; 9: 108.

40. Gartner F, David L, Seruca R, et al . Establishment and charac-terization of two cell lines derived from human diffuse gastriccarcinomas xenografted in nude mice. Virchows Arch 1996; 428:91–98.

41. Buffart TE, Tijssen M, Krugers T, et al . DNA quality assessmentfor array CGH by isothermal whole genome amplification. Cell

Oncol 2007; 29: 351–359.42. Pereira PM, Marques JP, Soares AR, et al . MicroRNA expression

variability in human cervical tissues. PLoS One 2010; 5: e11780.43. Livak KJ, Schmittgen TD. Analysis of relative gene expression

data using real-time quantitative PCR and the 2−��Ct method.Methods 2001; 25: 402–408.

44. Joana Figueiredo JS-C, Ola Soderberg, Gianpaolo Suriano, et al .ADP-ribosylation factor 6 mediates E-cadherin recovery by chem-ical chaperones. PLoS ONE 2011; 6.

45. Karam R, Carvalho J, Bruno I, et al . The NMD mRNA surveil-lance pathway downregulates aberrant E-cadherin transcripts ingastric cancer cells and in CDH1 mutation carriers. Oncogene

2008; 27: 4255–4260.46. Li S, Fu H, Wang Y, et al . MicroRNA-101 regulates expression

of the v-fos FBJ murine osteosarcoma viral oncogene homolog(FOS ) oncogene in human hepatocellular carcinoma. Hepatology

2009; 49: 1194–1202.47. Kim K, Lee HC, Park JL, et al . Epigenetic regulation of

microRNA-10b and targeting of oncogenic MAPRE1 in gastriccancer. Epigenetics 2011; 6: 740–751.

48. Tanaka T, Arai M, Wu S, et al . Epigenetic silencing ofmicroRNA-373 plays an important role in regulating cell prolif-eration in colon cancer. Oncol Rep 2011; 26: 1329–1335.

49. Tsuruta T, Kozaki KI, Uesugi A, et al . miR-152 is a tumor sup-pressor microRNA that is silenced by DNA hypermethylation inendometrial cancer. Cancer Res 2011; 71: 6450–6462.

50. Kottakis F, Polytarchou C, Foltopoulou P, et al . FGF-2 regu-lates cell proliferation, migration, and angiogenesis through anNDY1/KDM2B-miR-101-EZH2 pathway. Mol Cell 2011; 43:285–298.

51. Hao Y, Gu X, Zhao Y, et al . Enforced expression of miR-101inhibits prostate cancer cell growth by modulating the COX-2pathway in vivo. Cancer Prev Res (Phila) 2011; 4: 1073–1083.

52. Buechner J, Tomte E, Haug BH, et al . Tumour-suppressormicroRNAs let-7 and mir-101 target the proto-oncogene MYCN

and inhibit cell proliferation in MYCN-amplified neuroblastoma.Br J Cancer 2011; 105: 296–303.

53. Chiang CW, Huang Y, Leong KW, et al . PKCα-mediated induc-tion of miR-101 in human hepatoma HepG2 cells. J Biomed Sci

2010; 17: 35.54. Yu J, Cao Q, Mehra R, et al . Integrative genomics analysis reveals

silencing of β-adrenergic signaling by polycomb in prostate cancer.Cancer Cell 2007; 12: 419–431.


J Carvalho et al

55. Fujii S, Ito K, Ito Y, et al . Enhancer of Zeste homologue 2 (EZH2)down-regulates RUNX3 by increasing histone H3 methylation.J Biol Chem 2008; 283: 17324–17332.

56. Shun CT, Wu MS, Lin JT, et al . An immunohistochemical studyof E-cadherin expression with correlations to clinicopathologi-cal features in gastric cancer. Hepatogastroenterology 1998; 45:944–949.

57. Nikoloski G, Langemeijer SM, Kuiper RP, et al . Somatic muta-tions of the histone methyltransferase gene EZH2 in myelodys-plastic syndromes. Nat Genet 2010; 42: 665–667.

SUPPORTING INFORMATION ON THE INTERNET

The following supporting information may be found in the online version of this article:

Supplementary materials and methods

Figure S1. Frequencies of gains and losses throughout the genome of gastric tumours (n = 76) determined by BAC array CGH

Figure S2. Correlation between mature expression and copy number changes of miR-101

Figure S3. CDH1 promoter methylation and mutation analysis

Figure S4. EZH2 and CDH1 mRNA expression analysis

Table S1. Validation of aCGH data by genomic PCR followed by quantitative SNaPshot

Table S2. Classical CDH1 inactivation mechanisms in GCs with E-cadherin expression disruption

58. Ernst T, Chase AJ, Score J, et al . Inactivating mutations of thehistone methyltransferase gene EZH2 in myeloid disorders. Nat

Genet 2010; 42: .2 722–726.59. ∗Buffart TE, Carvalho B, Hopmans E, et al . Gastric cancers in

young and elderly patients show different genomic profiles.J Pathol 2007; 211: 45–51.

Cited in Supporting information only.


Documents

Lack of microRNA-101 causes E-cadherin functional deregulation through EZH2 up-regulation in intestinal gastric cancer