early diagnostic biomarkers for esophageal adenocarcinoma ...early diagnostic biomarkers for...

Review

Early Diagnostic Biomarkers for EsophagealAdenocarcinoma—The Current State of Play

Alok Kishorkumar Shah1, Nicholas A. Saunders1, Andrew P. Barbour2, and Michelle M. Hill1

AbstractEsophageal adenocarcinoma (EAC) is one of the two most common types of esophageal cancer with

alarming increase in incidence and very poor prognosis. Aiming to detect EAC early, currently high-risk

patients are monitored using an endoscopic-biopsy approach. However, this approach is prone to sampling

error and interobserver variability. Diagnostic tissue biomarkers related to genomic and cell-cycle abnormal-

ities have shown promising results, although with current technology these tests are difficult to implement in

the screening of high-risk patients for early neoplastic changes. Differential miRNA profiles and aberrant

protein glycosylation in tissue samples have been reported to improve performance of existing tissue-based

diagnostic biomarkers. In contrast to tissue biomarkers, circulating biomarkers are more amenable to

population-screening strategies, due to the ease and low cost of testing. Studies have already shown altered

circulating glycans and DNA methylation in BE/EAC, whereas disease-associated changes in circulating

miRNA remain to be determined. Future research should focus on identification and validation of these

circulating biomarkers in large-scale trials to develop in vitro diagnostic tools to screen population at risk for

EAC development. Cancer Epidemiol Biomarkers Prev; 22(7); 1185–209. �2013 AACR.

IntroductionAfterheart disease, cancer is the second leading cause of

death globally. Four major cancer sites account for half ofthe cancer-related mortalities: lung, colorectal, prostatein men, and breast in women. In past 2 decades, a steadydecrease in deaths of these 4 major site malignancies ledto an overall decrease in cancer-related death rates inmenand women (1). In contrast, the incidence of esophagealadenocarcinoma (EAC) is increasing faster than any othercancer type. EAC togetherwith esophageal squamous cellcarcinoma (ESCC) is the eighth most-common cancer byprevalence and sixth most-common cause of cancer-relat-ed death globally (2). In 1970s, the incidence of EACrepresented less than 5% of total esophageal cancer, anda majority of esophageal cancer cases diagnosed wereESCC. Over a period of 3 decades, EAC incidences havebeen increasing continuously, especially inwestern coun-tries among Caucasians. Now almost half of the esoph-ageal malignancy cases diagnosed are EAC (3, 4). EACand ESCC show marked differences in their geographicspread. EAC is more common in developed countriessuch as the United Kingdom (8 in 100,000 individuals;

ref. 5), Australia, and the United States. Within Europe,southern Europe has the highest EAC incidence (5). Onthe other side, ESCC is the most common type of esoph-ageal cancer amongdevelopingAsian countries (6). Racialdisparity also occurs between the 2 types of esophagealcancer. ESCC is more prevalent among Blacks, whereasEAC is at least twice as common in Whites as comparedwith other ethnic groups (7, 8). Once diagnosed, Blackpatients showed poorer overall survival than Whites(9, 10). Taken together, strong genetic and environmentalfactors relating to ethnicity and geographic distributionseem to be playing critical roles in the incidence of esoph-ageal cancer. Studies also suggest possible links betweensocioeconomic status and the prevalence of esophagealcancer phenotype (6).

Risk FactorsIn themajority of cases, EAC is diagnosed at a late stage,

leading to a poor 5-year survival of less than 15% (11).Hence, recent research for EAC has focused on under-standing risk factors and the identification of early diag-nostic biomarkers.

Esophageal cancer is unlikely to develop in individualsyounger than 40 years of age; however, after that theincidence increases significantly with each decade of life(9). Changing lifestyle and food habits are primarilyresponsible for the dramatic epidemiologic changes inEAC as described in recent reviews (11–13). Known EACrisk factors include accumulation of visceral fat in theabdomen (14), male gender, high intake of dietary fat andcholesterol with low intake of fruits and vegetables (15),tobacco smoking (16), reduction in Helicobacter pylori

Authors' Affiliations: 1The University of Queensland Diamantina Institute;and 2School of Medicine, The University of Queensland, Woolloongabba,Queensland, Australia

Corresponding Author: Michelle M. Hill, The University of QueenslandDiamantina Institute, Level 5, Translational Research Institute, 37 KentStreet, Woolloongabba, QLD 4102. Phone: 61-7-3443-7049; Fax: 61-7-3443-6966; E-mail: [email protected]

doi: 10.1158/1055-9965.EPI-12-1415

�2013 American Association for Cancer Research.

CancerEpidemiology,

Biomarkers& Prevention

www.aacrjournals.org 1185

on June 7, 2021. © 2013 American Association for Cancer Research. cebp.aacrjournals.org Downloaded from

Published OnlineFirst April 10, 2013; DOI: 10.1158/1055-9965.EPI-12-1415

http://cebp.aacrjournals.org/

infections (17), and Barrett’s esophagus (BE), a metaplas-tic change to the esophageal lining. Individuals withBarrett’s esophagus carry almost 30 to 125 times morerisk for EAC development, and 0.5% to 1% of patientswith Barrett’s esophagus are estimated to develop EACeach year (18). Barrett’s esophagus is characterized byreplacement of normal stratified squamous epitheliumwith metaplastic columnar epithelium and is consideredto be a successful adaptation of the distal esophagus inresponse to chronic gastroesophageal reflux disorder(GERD; ref. 19).

GERD is a very common condition in the western pop-ulation with around 20% reporting weekly symptoms ofheartburn and acid regurgitation (20). Refluxate-contain-ing bile acid, along with gastric acid, is considered to bemore harmful, leading to inflammation, ulceration, Bar-rett’s esophagus, and ultimately EAC. Development ofBarrett’s esophagus is a slow process and distinctivemucus-secreting goblet cell formation can take 5 to 10years (21, 22). Typically, EAC develops through metapla-sia–dysplasia–carcinoma sequence involving genetic andepigenetic modifications, leading to uncontrolled cellproliferation, and is characterized by the presence ofintestinal metaplasiawith low-grade (LGD) to high-gradedysplasia (HGD), which eventuallymay progress to inva-sive carcinoma (20).

Current Diagnosis ScenarioTo detect pathologic changes leading to EAC develop-

ment before onset of disease, current clinical practiceinvolves endoscopic screening of patients with high-riskGERD and to characterize the degree of dysplasia inbiopsy samples collected during endoscopy (23, 24).Enrollment of patients into an endoscopic screening pro-grammay be facilitated by a patient questionnaire of self-evaluated symptoms/complications (25, 26). Onceenrolled into the screening program, a patient undergoesendoscopy-biopsy every 3 months to 2 years dependingon the degree of dysplasia, during which 4 quadrantbiopsy samples are taken every 1 to 2 cm and evaluatedfor histologic changes by expert pathologists (23, 24). As asignificant number of patients histologically diagnosedwith HGD develop EAC, endoscopic mucosal ablation oresophageal resection (esophagectomy) are options to stopfurther disease progression in those high-risk patients(27, 28). Significantly improved survival is observed inpatients diagnosed at an early stage during surveillanceendoscopy program as compared with symptomaticallydiagnosed EAC (29–32).

Although current screeningmethodology shows prom-ise, outcome of endoscopy-biopsy in many cases is non-reproducible due to interobserver variability and sam-pling error (28, 33). Furthermore, histologic dysplasticchanges may be patchy and present heterogeneously inthe tissue sample. This makes the diagnosis challenging,especially in the early stages of transition to LGD (28, 34).In up to 40%of patients, invasive cancer has been found inresected tissue despite negative endoscopic examination

for the malignancy (35). Moreover, false-positive resultsalso occur, meaning despite intramucosal carcinoma in abiopsy, the subsequently resected tissue has no signs ofcarcinoma (28). These evidence suggest dysplasia gradingis an imperfect measure of cancer risk.

Despite extensive screening with currently availabletechniques, more than 80% of EACs are diagnosed with-out any prior diagnosis of Barrett’s esophagus or GERD(36, 37). According to an estimate, more than 80% ofBarrett’s esophagus cases are undiagnosed and thereforeare not getting the benefit of the screening program (38).On the other hand, a large proportion of patients under-going routine biopsy screening do not progress to EAC(13). These suggest inability of current methodologies inscreening population to detect high-risk patients and todistinguish between disease progressors from nonpro-gressors. In addition, the screening procedure is not verycost-effective (39). To overcome these challenges, adjunctuse of biomarker has been proposed to stratify the riskassociated with EAC development.

Biomarkers in EACAccording to United States’ NIH, a biomarker is "a

characteristic that is objectively measured and evaluatedas an indicator of normal biologic processes, pathogenicprocesses, or pharmacologic responses to a therapeuticintervention (40)."

In transit from intestinal metaplasia to LGD to HGD toEAC, cells acquire abilities to become self-sufficient forgrowth, evade apoptosis, proliferate uncontrollably, pro-mote angiogenesis, invade underlined epithelium, andstart to metastasize. These changes are accompanied withhistologic changes in tissue architecture, genomic insta-bility, development of tumor microenvironment, modu-lation of immune response, and are therefore reflected inbody fluids (serum/plasma/mucus/urine) or tissue sam-ples and differentiate in terms of their genome/prote-ome/metabolome profile (41). Thus, a biomarker can befrom any of these sources and reflect underlying patho-logic or homeostatic changes. Table 1 summarizes differ-ent classes of biomarkers proposed for BE/EAC.

National Cancer Institute Early Detection ResearchNetwork (EDRN) guidelines outline biomarker discoveryanddevelopment to a 5-phase process summarized below(42) and depicted in Fig. 1.

Phase I—Preclinical exploratory study: it comparesnormal versus cancer samples (body fluids/tissue)using technologies such as genomics, microarrayexpression, proteomics, immunohistochemistry, orimmunoblotting to detect significant changes inproteins/genes/metabolites between the groups.

Phase II—Clinical assay development and validation: it isaimed at developing a clinical assay using a minimallyinvasive sample collection method. The assay is meantto be robust, reproducible, and suitable for storedclinical samples to be used in later phases ofdevelopment. At the end of this phase, one should

Shah et al.

Cancer Epidemiol Biomarkers Prev; 22(7) July 2013 Cancer Epidemiology, Biomarkers & Prevention1186




expect high specificity and sensitivity for the assay.However, it remains to be determined how early thebiomarker can predict the disease.

Phase III—Retrospective longitudinal repository studies:the assay is applied on prospectively collected storedsamples to determine the ability of the biomarker todetect the disease before clinical presentation. If so, thencriteria for positive screening is determined for futureuse.

Phase IV—Prospective screening: the test is prospectivelyapplied to real population to detect the extent andcharacteristic of disease detected by the biomarker. Thisphase gives positive predictive value for the test andgives an idea about feasibility for last phase of controltrials.

Phase V—Cancer control studies: it comprises large-scaleclinical trial to determine the impact of new screeningprocess on the disease burden in the community.

With respect to EAC, none of the biomarkers, includinghigh-grade dysplasia, have been evaluated in phase V,whereas very feware evaluated in phase III and IV. Figure1 summarizes proposed EAC biomarkers and how wellthey are characterized in the process of biomarker dis-covery. The following sections will discuss some of theclasses of BE/EAC biomarkers.

Genomic InstabilityMany groups have studied genomic instability induced

by aneuploidy, tetraploidy, DNAmethylation, allelic lossand shown some predictive power for these changes. Arole for hypermethylation in the promoter regions oftumor-suppressor genes during the development of EAChas also been well established. Table 2 summarizes DNAmethylation changes associated with metaplasia–dyspla-sia–carcinoma development. In the majority of patients,methylation changes are acquired very early during EACdevelopment, hence these alterations could be used as anearly diagnostic biomarker. Apart from discriminatingpatients at different stages of EAC development, DNAmethylation signatures may be useful as predictors forprogression from Barrett’s esophagus to EAC (43, 44) andfor response to chemotherapy and survival in patientswith EAC (45, 46).

Although the individual genomic abnormality has thepotential to diagnose disease at different stages, bestresults are obtained when they are used in combination(47–49). LOH at chromosome 9p and 17p locus are con-sidered to be early events during Barrett’s esophaguspathogenesis (50). If present with other chromosomalalterations such as aneuploidy and tetraploidy, itincreases the 10-year risk for development of EAC from12% to approximately 80% (51). However, with the cur-rent flow cytometry technology, it is technically verychallenging for clinical laboratories to assess these geno-mic biomarkers in the samples, which limits widespreaduse of these biomarkers in the clinic.

Alternatively, genomic alterations canbedetected at theprotein level using immunohistochemistry. One of themost common and earliest genomic abnormality occurs atchromosome 17p, which codes for tumor-suppressor p53protein. Loss of p53 protein expression in tissue samplescorrelates with disease progression (52). However, as p53expression only reflects alterations at one particular gene,it has lower predictive value as comparedwith techniquesmonitoring multiple genomic abnormalities. Further-more, sensitivity drops as mutations or deletions at geno-mic level may not necessarily be detected at the proteinlevel (53).

In line with the genomic abnormalities described ear-lier, single-nucleotide polymorphism (SNP)–based geno-typing can also stratify cancer risk in patients with Bar-rett’s esophagus. As summarized in Table 3, in the past

None

Phase V: Cancer control studies

Bio

mar

ker d

isco

very

and

dev

elop

men

t

Phase IV: Prospective screening

Phase III: Retrospective longitudinal repository studies

Phase I and II: Preclinical exploration, clinical assay development and validation

High-gradedysplasis

DNA methylation, LOH,ploidy, p53 loss, cyclin D1

PCNA, Ki-67, EGFR, COX-2,miRNA, cMYC, HER2, NF-κB, Bcl-2,VEGF, E-cadherin, p16 abnormalities,

β-catenin, glycoproteins, etc.

Figure 1. Summary of current BE/EAC Biomarkers with respect to EDRNclinical phase of development.

Table 1. Comprehensive summary of differentclasses of BE/EAC biomarkers

Biomarker class Ref.

Tissue biomarkersGenomic abnormalities(ploidy and LOH)

(47–51)

DNA methylation Refer to Table 2SNPs/expression array studies Refer to Table 3

Inflammatory markersCOX-2 (69, 72–77)NF-kB (78–81)Cytokines (67, 79, 81–86)MMPs (87–93)Cell-cycle abnormalities (94, 95, 101)miRNA Refer to Table 4Glycosylation changes (121, 123–125)

Circulatory biomarkersDNA methylation changes (130–132)Glycan alterations (135–138)Metabolic profiling (142–145)

Biomarkers for Esophageal Adenocarcinoma

www.aacrjournals.org Cancer Epidemiol Biomarkers Prev; 22(7) July 2013 1187




Tab

le2.

Sum

maryof

hypermethy

latedge

nesduringBE/EAC

dev

elop

men

t Num

ber

(%)o

fsa

mplessh

owinghy

permethy

lationorstud

yfind

ings

Gen

eLo

cation

Func

tion

Metho

dNorm

alBE

LGD

HGD

EAC

Ref.

p16

(orCDKN2A

orINK4A

)9p

21Cyc

lin-dep

ende

ntkina

seinhibitor

Methy

latio

n-sp

ecificPCR

5/9(56%

)14

/18(77%

)—

—18

/21(85%

)(146

)

Methy

latio

n-se

nsitive

sing

le-stran

dco

nformationan

alysis

0/10

(0%)

4/12

(33%

)3/11

(27%

)3/10

(30%

)18

/22(82%

)(147

)

Methy

latio

n-sp

ecificPCR

0/17

(0%)

14/47(30%

)9/27

(32%

)10

/18(56%

)22

/41(54%

)(148

)Methy

latio

n-sp

ecificPCR

2/64

(3%)

14/93(15%

)—

—34

/76(45%

)(149

)Methy

latio

n-sp

ecificPCR

—3/10

(30%

)—

—5/11

(45%

)(150

)Methy

latio

n-sp

ecificPCR

—27

/41(66

%)

21/45(47%

)17

/21(81%

)65

/107

(61%

)(151

)Methy

latio

n-sp

ecificPCR

0%1/15

(7%)

4/20

(20%

)12

/20(60%

)8/15

(53%

)(152

)Methy

latio

n-sp

ecificPCR

Sep

aratelyde

term

ined

exon

1an

d2methy

latio

n.Five

of16

(31%

)exo

n-1,

8/16

(50%

)exo

n2in

EAC

patient

samples

show

edhy

permethy

latio

n.Exo

n2

methy

latio

nco

rrelates

with

stag

eof

thetumor

(P¼

0.01

)

(153

)

O6-M

ethy

lgua

nine

-10

q26

DNArepa

irMethy

light

tech

nique

2/10

(20%

)8/13

(62%

)—

—84

/132

(64%

)(154

)DNA Methy

ltran

sferase

(orMGMT)

Methy

latio

n-sp

ecificPCR

6/29

(21%

)24

/27(89%

)13

/13(100

%)

37/47(79%

)(155

)

APC

5q21

-q22

Wnt/b-caten

insign

aling

Methy

latio

n-sp

ecificPCR

0/17

(0%)

24/48(50%

)14

/28(50%

)14

/18(78%

)20

/32(63%

)(148

)

Methy

latio

n-se

nsitive

sing

le-stran

dco

nformation

analysis

and

methy

latio

n-se

nsitive

dot

blotas

say

0/16

(0%)

11/11(100

%)

——

20/21(95%

)(156

)Eight

of14

histolog

ically

norm

alga

stric

muc

osaad

jace

ntto

EAC

show

edsign

ifica

ntly

differen

tmethy

latio

nof

APC

promoter.

(157

)

GSTM

21p

13.3

Antioxidan

tsan

dprotec

tionag

ains

tDNAdam

age

Bisulfite

pyrose

quen

cing

(sam

ple

size

:EAC-

100,

BE-11,

dys

plasia-

11,n

ormal

esop

hage

al/

gastric

muc

osa-37

)

Tab

le2.

Sum

maryof

hypermethy

latedge

nesduringBE/EAC

dev

elop

men

t(Con

t'd)

Num

ber

(%)ofsa

mplessh

owinghy

permethy

lationorstud

yfind

ings

Gen

eLo

cation

Func

tion

Metho

dNorm

alBE

LGD

HGD

EAC

Ref.

Tach

ykinin-1

(TAC1)

7q21

-22

Smoo

thmus

cle

contractility,e

pith

elial

iontran

sport,

vasc

ular

permea

bility

andim

mun

efunc

tion

Methy

latio

n-sp

ecificPCR

5/67

(7.5%)

38/60(63.3%

)12

/19(63.2%

)11

/21(52.4%

)41

/67(61.2%

)(162

)

Rep

rimo

2q23

Reg

ulates

p53

-med

iatedce

ll-cy

cle

arrest

inG2-pha

se

Methy

latio

n-sp

ecificPCR

0/19

(0%

)9/25

(36%

)—

7/11

(64%

)47

/75(63%

)(163

)

E-C

adhe

rin16

q22

.1Caþ

2-dep

enden

tintercellularad

hesion

andmaintains

norm

altis

suearch

itecture

Methy

latio

n-sp

ecificPCR

0/4(0%)

——

—26

/31(84%

)(164

)

SOCS-3

17q25

.3Inhibits

cytokine

sign

aling

Methy

latio

n-sp

ecificPCR

0%4/30

(13%

)6/27

(22%

)20

/29(69%

)14

/19(74%

)(165

)

SOCS-1

16p13

.13

0%0/30

(0%

)1/27

(4%

)6/29

(21%

)8/19

(42%

)

Sec

retedfrizzled

-related

proteins(SFR

P)

SFR

P1

8p11

.21

Wnt

antago

nist

Methy

latio

n-sp

ecificPCR

7/28

(25%

)30

/37(81%

)—

—37

/40(93%

)(166

)SFR

P2

4q31

.318

/28(64%

)33

/37(89%

)—

—33

/40(83%

)SFR

P1

8p11

.21

Methy

latio

n-se

nsitive

sing

le-stran

dco

nformationan

alysis

andmethy

latio

n-se

nsitive

dot

blotas

say

1/12

(8%

)6/6(100

%)

——

23/24(96%

)(156

)

SFR

P2

4q31

.311

/15(73%

)6/6(100

%)

——

19/25(76%

)SFR

P4

7p14

.1Methy

latio

n-sp

ecificPCR

9/28

(32%

)29

/37(78%

)—

—29

/40(73%

)(166

)SFR

P5

10q24

.16/28

(21%

)27

/37(73%

)—

—34

/40(85%

)Plako

philin-1

(PKP1)

1q32

Cella

dhe

sion

and

intrac

ellularsign

aling

Methy

latio

n-sp

ecificPCR

5/55

(9.1%)

5/39

(12.8%

)—

1/4(25%

)20

/60(33.3)

(167

)

GATA

-48p

23.1-p22

Tran

scrip

tionfactor

andregu

late

cell

differen

tiatio

n

Methy

latio

n-sp

ecificPCR

0/17

(0%

)—

——

31/44(71%

)(168

)

GATA

-520

q13

.33

0/17

(0%

)—

——

24/44(55%

)CDH13

(orH-

cadhe

rinor

T-ca

dhe

rin)

16q24

Cella

dhe

sion

Methy

latio

n-sp

ecificPCR

0/66

(0%

)42

/60(70%

)15

/19(78.9%

)16

/21(76.2)

51/67(76.1%

)(169

)

(Con

tinue

don

thefollo

wingpag

e)






Tab

le2.

Sum

maryof

hypermethy

latedge

nesduringBE/EAC

dev

elop

men

t(Con

t'd)

Num

ber

(%)o

fsa

mplessh

owinghy

permethy

lationorstud

yfind

ings

Gen

eLo

cation

Func

tion

Metho

dNorm

alBE

LGD

HGD

EAC

Ref.

NELL

-1(nel-like1)

11p15

Tumor

suppress

orMethy

latio

n-sp

ecificPCR

0/66

(0%)

28/60(46.7%

)8/19

(42.1%

)13

/21(61.9%

)32

/67(47.8%

)(170

)Eye

sAbse

nt4

6q23

Apop

tosismod

ulator

Methy

latio

n-sp

ecificPCR

2/58

(3%)

27/35(77%

)—

—33

/40(83%

)(171

)A-kinasean

choring

protein

12(or

Gravinor

AKAP12

)

6q24

-25.2

Cell-sign

aling,

adhe

sion

,mito

gene

sis,

and

differen

tiatio

n

Methy

latio

n-sp

ecificPCR

0/66

(0%)

29/60(48.3%

)10

/19(52.6%

)11

/21(52.4%

)35

/67(52.2)

(172

)

Vim

entin

10p13

Cytos

keletonprotein

Methy

latio

n-sp

ecificPCR

0/9(0%

)10

/11(91%

)—

5/5(100

%)

21/26(81%

)(173

)RUNX3

1p36

Tran

scrip

tionfactor

Methy

latio

n-sp

ecificPCR

1/63

(2%)

23/93(25%

)—

—37

/77(48%

)(149

)HPP1

19pter-p13

.1Tu

mor

suppress

or2/64

(3%)

41/93(44%

)—

—55

/77(71%

)3-OST-2

16p12

Sulfotran

sferas

een

zyme

1/57

(2%)

47/60(78%

)—

—28

/73(38%

)

Wnt

inhibitory

factor-1

(WIF-1)

12q14

.3Wnt

antago

nist

Methy

latio

n-sp

ecificPCR

81%

ofpa

tientswith

Barrett's

esop

hagu

ssu

fferingfrom

EAC

show

edhy

permethy

latedWIF-1

asco

mpared

with

20%

ofpa

tientswith

Barrett's

esop

hagu

swith

outEAC

(174

)

CHFR

(che

ckpoint

with

forkhe

adasso

ciated

and

ringfing

er)

12q24

Mito

sisch

eckpoint

protein

Bisulfite

pyros

eque

ncing

EAC

samples31

%(18/58

)sho

wed

sign

ifica

ntly

high

erCHFR

promoter

methy

latio

nas

compared

with

norm

alsa

mples(P

¼0.01

).(175

)

Metallothione

in3

(orMT3

)16

q13

Metal

homeo

stas

isan

dprotec

tion

agains

tDNAda

mag

e

Bisulfite

pyros

eque

ncing

(sam

ple

size

:normal-33,

BE-5,E

AC-78)

Iden

tified

2region

s(R2an

dR3)

ofCpG

nucleo

tides

,which

show

edsign

ifica

ntly

high

ermethy

latio

nin

EACas

compared

with

norm

alep

ithelium

(FDR<0.00

1).

Increa

sedDNAmethy

latio

nof

MT3

promoter

R2co

rrelates

with

adva

nced

tumor

stag

e(P

¼0.00

5)an

dlymphno

demetas

tasis(P

¼0.03

).DNAmethy

latio

nof

MT3

promoter

R3co

rrelates

with

tumor

stag

ing(P

¼0.03

)but

notwith

lymph

nodestatus

(P¼

0.4).

(176

)

Methy

lationmarke

rpan

el

Sam

ple

size

Metho

dFind

ings

Ref.

EAC-35un

dergo

ing

chem

orad

iotherap

yMethy

latio

n-sp

ecificPCR

Com

bine

dmea

nof

promoter

methy

latio

nof

p16

,Rep

rimo,

p57

,p73

,RUNX-3,

CHFR

,MGMT,

TIMP-3,a

ndHPP1was

lower

inpatientswho

resp

onded

toch

emorad

iotherap

y(13/35

)asco

mpared

with

patientswho

did

notresp

ond

(22/35

;P¼

0.00

3).

(46)

BE-62(28patientswith

Barrett's

esop

hagu

sprogres

sedto

EAC

andremaining

34pa

tientswith

Barrett's

esop

hagu

swere

nonp

rogres

sors)

Methy

latio

n-sp

ecificPCR

Three-tie

redstratifi

catio

nmod

elwas

dev

elop

edus

ingmethy

latio

nindex

(p16

,HPP1,

andRUNX3),B

arrett's

esop

hagu

sleng

than

dpatho

logy

.Com

bine

dmod

elbas

edon

2-(ROC:0

.838

6)an

d4-ye

ar(ROC:0

.791

0)predictionwas

able

toca

tego

rizepatientswith

Barrett'ses

opha

gusinto

low-risk,

interm

ediate-risk,

andhigh

-riskgrou

psforEAC

dev

elop

men

t.

(44)

(Con

tinue

don

thefollo

wingpag

e)

Shah et al.





Tab

le2.

Sum

maryof

hypermethy

latedge

nesduringBE/EAC

dev

elop

men

t(Con

t'd)

Num

ber

(%)ofsa

mplessh

owinghy

permethy

lationorstud

yfind

ings

Gen

eLo

cation

Func

tion

Metho

dNorm

alBE

LGD

HGD

EAC

Ref.

BE-195

(145

patientswith

Barrett's

esop

hagu

sprog

ressed

toEAC

andremaining

50patientswith

Barrett's

esop

hagu

swere

nonp

rogres

sors)

Methy

latio

n-sp

ecificPCR

HPP1(P

¼0.00

25),p16

(P¼

0.00

66),an

dRUNX3(P

¼0.00

02)w

eresign

ifica

ntly

hypermethy

latedin

progres

sors

asco

mpared

with

nonp

rogres

sors.In

combination,

pan

elof

8methy

latio

nmarke

rs(p16

,HPP1,

RUNX3,

CDH13

,TA

C1,

NELL

1,AKAP12

,and

SST)

show

edse

nsitivitie

sof

0.44

3an

d0.62

9at

spec

ificity

of0.9an

d0.8forEAC

progres

sion

inpatientswith

Barrett's

esop

hagu

sus

ingco

mbined

mod

eldes

igne

don

thebas

isof

2an

d4ye

arsof

follo

w-up.

(43)

EAC-41(adjace

ntno

rmal

samples

asco

ntrol)

Methy

latio

n-sp

ecificPCR

Patientsha

ving

morethan

50%

oftheirge

nesmethy

lated(APC,E

-cad

herin

,MGMT,

ER,p

16,D

AP-kinase,

andTIMP3)

show

edsign

ifica

ntly

poo

r2-ye

arsu

rvival(P

¼0.04

)and

2-ye

arrelapse

-freesu

rvival(P

¼0.03

)asco

mpared

with

thepa

tientsha

ving

less

than

50%

methy

latio

n.

(45)

BE-18,

EAC-38(m

ultip

lebiopsies

weretake

nan

dclas

sified

into

norm

al,B

arrett's

esop

hagu

s,HGD,a

ndEAC)

Bisulfite-m

odified

DNAwith

PCR

Themethy

latio

nfreq

uenc

iesof

9ge

nes(APC,C

DKN2A

,ID4,

MGMT ,

RBP1,

RUNX3,

SFR

P1,

TIMP3,

andTM

EFF

2)foun

dto

be95

%,5

9%,7

6%,5

7%,7

0%,

73%

,95%

,74%

,and

83%

,res

pec

tively,inEACsa

mples,whe

reas

95%

,28%

,78

%,4

8%,5

8%,4

8%,9

3%,8

8%,a

nd75

%,res

pec

tively,

inBarrett's

esop

hagu

ssa

mples,

which

was

sign

ifica

ntly

high

eras

compared

with

norm

alsq

uamou

sep

ithelium.T

hemethy

latio

nfreq

uenc

yforC

DKN2A

andRUNX3was

sign

ifica

ntly

high

erforEAC

asco

mpared

with

Barrett's

esop

hagu

sbiopsy

samples

.

(177

)

Normal-30,

BE-29,

HGD-8,E

AC-29

Illum

inaGolden

Gatemethy

latio

nbea

darray

Ove

rallmed

ianmethy

latio

nat

thetotal706

numbersof

mos

tinformativeCpG

sites

grad

ually

increa

sedfrom

norm

al-B

E-H

GD/EAC

(P<0.00

1).T

heau

thors

differen

tiatedbetwee

nEACvs.normal,H

GDvs.normal,B

arrett'ses

opha

gusvs.

norm

al,E

ACvs

.Barrett's

esop

hagu

s,an

dHGDvs

.Barrett's

esop

hagu

sbas

edon

422,

225,

195,

17,a

nd3nu

mbersof

CpG

sites,which

issh

owingdifferen

tial

methy

latio

nbetwee

nresp

ectiv

egrou

ps.

(178

)

Iden

tifica

tionpha

se(BE-22,

EAC-

24);retros

pec

tiveva

lidation

pha

se(BE-60,

LGD/H

GD-36,

EAC-90);p

rosp

ectiv

eva

lidation

pha

se(98pa

tientsun

der

surveillanc

e).

Iden

tifica

tionpha

se:IlluminaInfinium

assa

y;retros

pec

tive/prosp

ectiv

eva

lidationpha

se:

pyros

eque

ncing

Onthebas

isof

initial

iden

tifica

tionpha

se,7

gene

s(SLC

22A18

,ATP

2B4,

PIGR,

GJA

12,R

IN2,

RGN,a

ndTC

EAL7

)sho

wingmos

tprominen

tmethy

latio

nch

ange

swerese

lected

forva

lidation.

Com

binationof

4ge

nes(ROC

0.98

8)SLC

22A18

,PIG

R,G

JA12

,and

RIN2sh

owed

sens

itivityof

94%

andsp

ecificityof

97%.T

hispa

nelo

f4ge

nessh

owingdifferen

tialm

ethy

latio

n,stratifi

edpatients

into

low-,interm

ediate-,an

dhigh

-riskgrou

psforEAC

dev

elop

men

tin

prosp

ectiv

eva

lidation.

(179

)

Non

dys

plasticBarrett'ses

opha

gus

(not

progres

sedto

EAC)-16

,Barrett's

esop

hagu

smuc

osa

from

patientsprogres

sedto

EAC-12

Methy

latio

n-se

nsitive

sing

le-stran

dco

nformation

analysis

andmethy

latio

n-se

nsitive

dot

blot

assa

yBarrett'ses

opha

gussa

mplesco

llected

from

patientswho

prog

ressed

toEACin12

mon

thstim

epe

riodsh

owed

100%

,91%

,and

92%

hypermethy

latio

nof

APC,

TIMP-3,a

ndTE

RT,

resp

ectiv

ely,

asco

mpared

with

36%

,23%

,and

17%

inBarrett's

esop

hagu

smuc

osaco

llected

from

patientswho

did

notprogres

sto

EAC.

(180

)

Methy

lationmarke

rpan

el

Sam

ple

size

Metho

dFind

ings

Ref.






Tab

le3.

Sum

maryof

gene

expressionprofilingstud

iesforBE/EAC

Sam

ple

size

Array

des

criptio

nOutco

me

Find

ings

Externa

lvalidation

Ref.

BE-21(pairedno

rmal

esop

hage

alan

dga

stric

samplesas

control)

Seriala

nalysisof

gene

express

ion,

PCRan

dim

mun

oblotting

Disea

seprog

ression

Ofno

te,5

34tags

weresign

ifica

ntly

differen

tially

expressed

betwee

nno

rmales

opha

gealsq

uamou

sep

ithelium

andBarrett's

esop

hagu

s.Th

emos

tup

regu

latedge

nesin

Barrett's

esop

hagu

sas

compared

with

norm

alep

ithelium

wereiden

tified

tobetrefoilfac

tors,a

nnex

inA10

andga

lectin-4

with

each

differen

ttypeof

tissu

esh

owed

anun

ique

cytoke

ratin

expres

sion

.

No

(181

)

Barrett's

esop

hagu

san

dHGD-11

(match

edbiopsy

samples)

cDNAmicroarray

Disea

seprog

ression

Using

2.5-fold

cutoff,iden

tified

131up

regu

latedan

d16

dow

nreg

ulated

gene

sinHGD.Twen

ty-fou

rof28

mos

tsign

ifica

ntly

differen

tge

nessh

owed

similar

chan

gesduringva

lidation.

Rea

l-tim

ePCR

(182

)

EAC-91

Oligo-microarray

Disea

seprog

ression

A4-ge

nepan

elco

nsists

ofdeo

xycy

tidinekina

se,3

0 -pho

spho

aden

osine50-pho

spho

sulfa

tesy

ntha

se2,

sirtuin-2,

andtripartitemotif-co

ntaining

44predicted

5-ye

arsu

rvival.

Immun

ohistoch

emistry

(183

)

Twen

ty-three

pairedBarrett's

esop

hagu

san

dno

rmal

epith

elium

samples

Tran

scrip

tiona

lprofiling

andproteo

mics

Disea

seprog

ression

Iden

tified

2,82

2ge

nesto

bedifferen

tially

expressed

betwee

nBarrett's

esop

hagu

san

dno

rmal

epith

elium.S

ignifica

ntly

overex

press

edge

nes

duringBarrett's

esop

hagu

sbe

long

edto

cytokine

san

dgrow

thfactors,

cons

titue

ntsof

extrac

ellular

matrix

,bas

emen

tmem

brane

andtig

htjunc

tions

,proteinsinvo

lved

inprostag

land

inan

dpho

spho

inos

itolm

etab

olism,n

itric

oxide

produc

tionan

dbioen

erge

tics.

While

gene

sen

codingHSPan

dva

rious

kina

seswere

dow

nreg

ulated

.

No

(184

)

Lymphno

demetas

tatic

(n¼

55)

andno

nmetas

tatic

(n¼

22)E

AC

samples

Oligo-microarray

Disea

seprog

ression

Lymphno

de–

pos

itive

samplessh

owed

sign

ifica

ntdow

nreg

ulationof

arginino

succ

inatesynthe

tase

asco

mpared

with

lymphno

deno

nmetas

tatic

samples(P

¼0.04

8).

No

(185

)

EAC-6

andga

stric

cardiaca

ncer-8

aCGH

Disea

seprog

ression

Iden

tified

HGF(45%

)and

BCAS1(27%

)tobemos

tfreq

uently

overex

pressed

gene

sresp

ectiv

elyat

7q21

and20

q13

locu

s.

No

(186

)

Eleve

nmatch

edsa

mplese

ts(hea

lthy-BE-EAC

match

ed-6,

norm

al-B

Ematch

ed-4

and

norm

al-EAC

match

ed-1)

SNPmicroarray

Disea

seprog

ression

60%

ofBarrett's

esop

hagu

san

d57

%of

EAC

samplesco

ntaine

dat

leas

ton

eof

thege

nomic

alteratio

nsin

theform

ofdeletions

,dup

lications

,am

plifica

tions

,cop

ynu

mber

chan

ges,

andne

utral

LOH.

No

(187

)

(Con

tinue

don

thefollo

wingpag

e)

Shah et al.





Tab

le3.

Sum

maryof

gene


iesforBE/EAC

(Con

t'd)

Sam

ple

size

Array

des

cription

Outco

me

Find

ings

Externa

lvalidation

Ref.

Normal-39,

BE-25,

EAC-38,

and

ESCC-26

cDNAmicroarray

Disea

seprogres

sion

Clusteringsh

owed

these

parationof

samples

into

4distinct

grou

ps.

Ofno

te,2

,158

clon

eswere

differen

tially

expressed

betwee

nno

rmal

and

Barrett's

esop

hagu

ssa

mples,

whe

reas

1,30

6be

twee

nBarrett's

esop

hagu

san

dEAC.B

E/EAC

samplessh

owed

differen

tiale

xpressionof

hydrolase

s,lyso

zyme,

fuco

sidas

e,tran

scrip

tion

factors,

muc

ins,

andthetrefoilfac

tors.

No

(188

)

BE-20,

LGD-19,

HGD-20an

dEAC-42

SNPmicroarray

Disea

seprogres

sion

Increa

sing

numbersof

SNPsan

dloss

ofch

romos

omes

with

disea

seprogres

sion

.Chrom

osom

aldisruptio

nwas

iden

tified

intheFH

IT,

WWOX,R

UNX1,

KIF26

B,M

GC48

628,

PDE4D

,C20

orf133

,GMDS,D

MD,a

ndPARK2ge

nesin

EAC.

No

(189

)

EAC-75sp

ecim

ensfrom

64pa

tients,ad

jace

ntpairedno

rmal

tissu

efrom

patientswith

EAC-

28

DNAmicroarray

Disea

seprogres

sion

Iden

tified

AKR1B

10,C

D93

,CSPG2,

DKK3,

LUM,

MMP1,

SOX21

,SPP1,

SPARC,a

ndTW

IST1

gene

sas

biomarke

rbas

edon

tran

scrip

tomicsdata.

Qua

ntita

tivereal-tim

ePCRiden

tified

SPARC

and

SPP1ge

nesto

beas

sociated

with

EAC

patient

survival

(P<0.02

4).

Rea

l-tim

ePCR

(190

)

EAC-8,g

astric

cardia

canc

er-3

aCGH

andcD

NA

microarray

Disea

seprogres

sion

Tran

scrip

tomicsdataiden

tified

11ge

nesto

be

differen

tially

expressed

(ELF

3,SLC

45A3,CLD

N12

,CDK6,

SMURF1

,ARPC1B

,ZKSCAN1,

MCM7,

COPS6,

FDFT

1 ,an

dCTS

B).IHCan

alys

isreve

aled

sign

ifica

ntov

erex

pressionof

CDK6ace

ll-cy

cle

regu

latorin

tumor

samples.

No

(191

)

BE-20

aCGH

arrays

andhigh

den

sity

SNP

geno

typing

Disea

seprogres

sion

Cop

ynu

mber

loss

esweredetec

tedat

FRA3B

(81%

),FR

A9A

/C(71.4%

),FR

A5E

(52.4%

),an

dFR

A4D

(52.4%

)site

sin

early

Barrett's

esop

hagu

s.Validationstud

yco

nfirm

edloss

ofFR

A3B

and

FRA16

Din

early

Barrett's

esop

hagu

ssa

mples.

Rea

l-tim

ePCRan

dpyros

eque

ncing

(192

)

BE-11,

gastroes

opha

geal

junc

tion

(GEJ)

aden

ocarcino

ma-11

aCGH

with

awho

lech

romos

ome8q

contig

array

Disea

seprogres

sion

Ove

rexp

ress

ionof

MYC

andEXT1

,while

downreg

ulationof

MTS

S1,

FAM84

B,a

ndC8o

rf17

issign

ifica

ntly

asso

ciated

with

GEJ

aden

ocarcino

ma.

(193

)

BE-14,

EAC-5,E

SCC-3

cDNAmicroarray

Disea

seprogres

sion

Iden

tified

160ge

nesthat

candifferen

tiate

betwee

nBarrett's

esop

hagu

san

des

opha

geal

canc

er.

No

(194

)

Twen

ty-fou

rpa

iredsa

mplesof

norm

al,B

arrett's

esop

hagu

s,an

dEAC

phen

otyp

e

cDNAmicroarray

Disea

seprogres

sion

Ofno

te,2

14differen

tially

regu

latedge

nesco

uld

differen

tiate

betwee

nno

rmal,B

arrett'ses

opha

gus,

andEACphe

notype.

Gen

esinvo

lved

inep

idermal

No

(195

)

(Con

tinue

don

thefollo

wingpag

e)






Tab

le3.

Sum

maryof

gene


iesforBE/EAC

(Con

t'd)

Sam

ple

size

Array

des

criptio

nOutco

me

Find

ings

Externa

lvalidation

Ref.

differen

tiatio

nareun

derex

pressed

inEAC

asco

mpared

with

Barrett's

esop

hagu

s.Exp

ress

ion

ratio

ofGATA

6to

SPRR3ca

ndifferen

tiate

betwee

n3ph

enotyp

esstud

ied.

Poo

ledbiosp

ysa

mplesfrom

Barrett's

esop

hagu

s,es

opha

geal

squa

mou

s,ga

stric

,an

dduo

den

um

Oligo-microarray

Disea

seprog

ression

Differen

tiate

differen

ttis

sueclus

ters

bas

edon

gene

expressionprofile.Iden

tified

38ge

nesthat

are

upregu

latedin

Barrett's

esop

hagu

stis

sueclus

ter,

which

belong

toce

llcy

cle(P1c

dc4

7,PCM-1),ce

llmigratio

n(urokina

se-typ

eplasm

inog

enrece

ptor,

LUCA-1/H

YAL1

),grow

thregu

latio

n(TGF-b

superfamily

protein,a

mphiregu

lin,C

yr61

),stress

resp

onse

s(calcy

clin,A

TF3,

TR3orpha

nrece

ptor),

epith

elialc

ells

urface

antig

ens(epsilon-BP,E

SA,

integrin

b4,m

esothe

linCAK-1

antig

enprecu

rsor),

and4muc

ins.

No

(196

)

Normal-24,

BE-18,

EAC-9

cDNAmicroarray

Disea

seprog

ression

Iden

tified

457,

295,

and36

differen

tially

expres

sed

gene

s,resp

ectiv

ely,betwee

nno

rmal-EAC,normal-

Barrett's

esop

hagu

s,an

dBE–EAC

grou

ps.

No

(197

)

89-EAC

cDNA-m

ediated

anne

aling,

selection,

extens

ion,

andlig

ation

assa

ywith

502kn

own

canc

er-related

gene

s

Disea

seprog

ression

Iden

tified

differen

tialg

eneex

pressionbe

twee

nea

rlystag

esof

EAC(T1an

dT2

)vs.late

(T3an

dT4

).Gen

eex

pressionprofile

reve

aled

ERBB4,

ETV

1,TN

FSF6

,MPLge

nesto

beco

mmon

betwee

nad

vanc

edtumor

stag

ean

dlymphno

demetas

tasis.

No

(198

)

Normal

esop

hage

almuc

osa-9,

esop

hagitis

-6,B

E-10,

EAC-5,

GEJad

enoc

arcino

ma-9,

stom

achsa

mples-32

(normal

muc

osa-11

,IM-9,intes

tinal-

typead

enoc

arcino

ma-7,

and

diffus

eca

rcinom

a-5)

cDNAmicroarray

Disea

seprog

ression

Onthebas

isof

theex

pressionprofile,g

enes

asso

ciated

with

thelip

idmetab

olism

andcy

tokine

nodulearefoun

dto

besign

ifica

ntly

altered

betwee

nEAC

andothe

rgrou

ps.

No

(199

)

Sev

enteen

pairedsa

mples

ofno

rmal,B

E/EAC

cDNAmicroarray

Disea

seprog

ression

Eac

htis

suetype

expresses

distin

ctse

tof

gene

s,which

candifferen

tiate

betwee

ntheirph

enotyp

es.

Barrett'ses

opha

gusan

dEACex

presses

similarset

ofstromal

gene

sthat

aredifferen

tfrom

norm

alep

ithelium.

No

(200

)

BE-19,

EAC-20(98tis

sue

spec

imen

swereco

llected

and

catego

rized

into

differen

tgrou

ps)

Onthebas

isof

previou

smicroarraystud

ies23

gene

swereva

lidated

usingreal-tim

ePCR

Disea

seprog

ression

Out

of23

gene

s,pa

nelo

f3ge

nes(BFT

,TSPAN,a

ndTP

)was

able

todiscrim

inatebetwee

nBarrett's

esop

hagu

san

dEACin

internal

valid

ationwith

0%clas

sifica

tionerror.

N.A.

(201

)

(Con

tinue

don

thefollo

wingpag

e)

Shah et al.





Tab

le3.

Sum

maryof

gene


iesforBE/EAC

(Con

t'd)

Sam

ple

size

Array

des

cription

Outco

me

Find

ings

Externa

lvalidation

Ref.

Normal-30,

BE-31,

gastric

muc

osa-34

,duo

den

um-18

Biomarke

rsforBarrett's

esop

hagu

swere

iden

tified

using3

pub

licly

available

microarraydatas

ets

andva

lidated

using

real-tim

ePCRan

dim

mun

ohistoch

emistry.

Disea

seprogres

sion

Out

of14

gene

siden

tified

,dop

ade

carbox

ylas

e(DDC)

andTrefoilfac

tor3(TFF

3)wereva

lidated

tobe

upregu

latedin

Barrett's

esop

hagu

s.

N.A.

(202

)

EAC-56

Olig

onuc

leotide

microarrayan

daC

GH

Disea

seprogres

sion

Iden

tified

4ne

wge

nes(EGFR

,WT1

,NEIL2,

and

MTM

R9)to

beov

erex

pressed

in10

%to

25%

EAC.

Exp

ress

ionleve

lsof

thes

e4ge

nesdifferen

tiated

patie

ntswith

EAC

into

3grou

psna

melygo

od,

averag

e,an

dpoo

rdep

endingup

ontheirp

rogn

osis

(P<0.00

8)

Immun

ohistoch

emistry

(203

)

BE/LGD-72,

HGD-11,

EAC-15

Bac

teria

lartificial

chromos

omeaC

GH

Disea

seprogres

sion

Cop

ynu

mber

chan

gesweremoreco

mmon

and

larger

asdisea

seprogres

sto

laters

tage

s.Patients

having

copynu

mber

alteratio

nsinvo

lvingmore

than

70Mbpwereat

increa

sedris

kof

progres

sion

toEAC

(P¼

0.00

47)

No

(60)

EAC-30,

BE-6,L

GD-9,H

GD-10

Gen

ome-wideCGH

Disea

seprogres

sion

Loss

of7q

33-q35

was

foun

din

HGDas

compared

with

precu

rsor

LGD(P

¼0.01

).Lo

ssof

16q21

-q22

andga

inof

20q1

1.2-q1

3.1was

sign

ifica

ntly

differen

tbetwee

nHGDan

dEAC

(P¼

0.02

and

0.03

,res

pec

tively).

No

(56)

EAC-30,

lymphno

demetas

tasis-

8,HGD-11,

LGD-8,a

ndBE-6

from

30EAC

patient

biopsy

samples

CGH

Disea

seprogres

sion

Iden

tified

region

sun

dergo

ingco

pynu

mber

loss

and

amplifica

tionduringea

chstag

eof

tran

sitio

n.Ave

rage

number

ofch

romos

omal

imba

lanc

ese

que

ntially

increa

sedfrom

BE–LG

D–HGD–EAC–

lymphno

demetas

tasis.

No

(54)

Forty-tw

opatientsreprese

ntdifferen

tstag

esof

disea

seSNParray

Disea

seprogres

sion

SNPab

norm

alities

increa

sesfrom

2%to

morethan

30%

asthedise

aseprog

ress

from

Barrett's

esop

hagu

sto

EAC.T

otalnu

mbe

rofS

NPalteratio

nsin

tissu

esa

mples

istig

htlyco

rrelated

with

DNA

abno

rmalities

such

asan

euploidy

andLO

H.

No

(57)

EAC-27an

dmatch

edno

rmal-14

SNParray

Disea

seprogres

sion

Con

firm

edprev

ious

lydes

cribed

geno

mic

alteratio

nssu

chas

amplifica

tionon

8qan

d20

q13

ordeletion/

LOH

on3p

and9p

.Alsoiden

tified

alteratio

nsin

seve

raln

ovel

gene

san

dDNAregion

sin

EAC

samples

.

No

(58)

(Con

tinue

don

thefollo

wingpag

e)






Tab

le3.

Sum

maryof

gene


iesforBE/EAC

(Con

t'd)

Sam

ple

size

Array

des

cription

Outco

me

Find

ings

Externa

lvalidation

Ref.

EAC-26

SNParray

Disea

seprogres

sion

Con

firm

edpreviou

slyreportedfreq

uent

chan

gesto

FHIT,C

DKN2A

,TP53

,and

MYC

gene

sin

EAC.

Iden

tified

PDE4D

andMGC48

628as

tumor-

suppress

orge

nes.

No

(59)

EAC-35

cDNAmicroarray

Res

pon

seto

chem

othe

rapy

Iden

tified

165differen

tially

expres

sedge

nesbetwee

npo

or(n

¼17

)and

good

outcom

e(n

¼18

)patient

grou

ps.

Topfunc

tiona

lpathw

aybas

edon

differen

tialg

eneex

pressionwas

iden

tified

tobe

Toll-rece

ptorsign

aling.

No

(204

)

EAC-47(lo

cally

adva

nced

tumor)

cDNAmicroarray

Res

pon

seto

chem

othe

rapy

Iden

tified

86ge

nessh

owingat

leas

t2-folddifferen

cebe

twee

nch

emothe

rapy

resp

onders(n

¼28

)and

nonres

pon

ders(n

¼19

).EphrinB3rece

ptor,w

hich

show

edhigh

estdifferen

cebetwee

nthegrou

ps,

show

edstrong

mem

brane

staining

inch

emothe

rapyresp

ondingtumorsus

ing

immun

ohistoch

emistry.

No

(205

)

Patientswith

EAC-19un

dergo

ing

chem

orad

iotherap

yOlig

o-microarray

Res

pon

seto

chem

orad

iotherap

yRed

uced

expressionof

IVL,

CRNN,N

ICE-1,S

100A

2,an

dSPPR3ge

nesco

rrelated

with

poo

rsurviva

land

nonres

pon

seto

chem

othe

rapy.

No

(206

)

19patients(EAC-16,

ESCC-2

and

aden

osqua

mou

sca

rcinom

a-1)

undergo

ingch

emorad

iotherap

y

Olig

o-microarray

Res

pon

seto

chem

orad

iotherap

yLo

wer

expressionforp

anelof

gene

sPERP,S

100A

2,an

dSPRR3was

asso

ciated

with

nonres

pon

seto

therap

y.Pathw

ayan

alysis

iden

tified

downreg

ulationof

apop

tosisin

nonres

pon

ders.

No

(207

)

EAC-174

,ESCC-36

SNPsas

sociated

with

the

chem

othe

rapy

drug

actio

npa

thway

Res

pon

seto

chem

orad

iotherap

yIden

tified

asso

ciationbetwee

nge

netic

polymorphism

san

dresp

onse

topreop

erative

chem

othe

rapy(fluo

rourac

ilan

dplatinum

compou

nds)

andradiotherap

y.

No

(208

)

NOTE

:Majority

ofstud

iesdes

cribed

inthis

table

includ

eva

lidationof

resu

ltsin

thesa

mepatient

coho

rtas

used

indisco

very

pha

se.Onlystud

iesthat

includ

edva

lidationus

ing

indep

enden

tpatient

coho

rtarede

scrib

edas

external

valid

ation.

Shah et al.





decade, several studies conducted using advanced geno-mic techniques such as array-comparative genomichybridization (aCGH) and SNP arrays confirmed previ-ously reported copy number alterations and identifiednovel genomic loci undergoing changes during process ofmetaplasia–dysplasia–carcinoma development (54–60). Ithas been shown that as the disease progresses from earlyto late stages, SNP abnormalities increase from approxi-mately 2% to 30% (54, 57). The total number of SNPalterations in tissue samples is tightly correlated withpreviously reported DNA abnormalities such as aneu-ploidy, copy number alterations, and LOH highlightingthe application of SNP-based genotyping to assess geno-mic abnormalities (54–60). Thus, SNP-based genotypingprovides an alternative way to assess genomic abnormal-ities during EAC pathogenesis.Studies on gene expression changes in EAC have been

propelled by recent progress in genomic technologies,each identifying unique sets of gene expression profile,which can be used as a biomarker panel for diseasediagnosis, prognosis, or to predict response to therapy(Table 3).Moreover, determination of the gene expressionchanges has been extremely helpful to understanddetailed pathogenesis and will form basis for developingfuture therapies. However, future validation using inde-pendent sample cohorts will be necessary for themajorityof these potential biomarkers.Apart from genomic abnormalities associated with the

disease progression, inheriting genetic factors are alsoimplicated for EAC development. Risk for BE/EAC andGERD is increased by 2- to 4-fold when a first-degreerelative is already affected by any of these conditions (61).Recently, a study conducted by The Esophageal Adeno-carcinoma Genetics Consortium and TheWellcome TrustCase Control Consortium identified link between SNPs atthe MHC locus and chromosome 16q24.1 with risk forBarrett’s esophagus (62). They also identified SNPs asso-ciated with body weight measures that were present withmore than expected frequency in Barrett’s esophagussamples supporting epidemiologic findings about obesityas a risk factor for Barrett’s esophagus and EAC (62). Wuand colleagues examined the relationship between pres-ence of risk genotypes and the onset of EAC. They iden-tified 10 SNPs associatedwith the age of EAConset. Genesassociated with 5 of 10 SNPs identified were known to beinvolved in apoptosis (63).Recently, published cancer genome–sequencing stud-

ies have given deeper insights into the genomic abnor-malities associated with the EAC pathogenesis. The com-parative genomic analysis between EAC and ESCCreported by Agrawal and colleagues (64) confirmed pre-viously verywell-described association of p53 genemuta-tions with esophageal cancer development. The authorsalso conducted comparative genome-wide analysisbetween matched Barrett’s esophagus and EAC patienttissue samples and concluded that the majority of geno-mic changes occur early during EAC development, at thestageofBarrett’s esophagus (64). Similar conclusionswere

made by next-generation sequencing of biopsy samplesobtained from the same patient at the stage of Barrett’sesophagus and EAC (65). The authors also identifiedARID1A as novel tumor-suppressor gene and around15% of patientswith EAC showed loss of ARID1Aproteinin tissue samples. In vitro studies suggested it to beassociated with cell growth, proliferation, and invasion(65). Very recently published high-resolution methylomeanalysis has provided first evidence for methylationchanges at genomic regions that encodenoncodingRNAs.The authors identified longnoncodingRNA,AFAP1-AS1,to be severely hypomethylated in Barrett’s esophagus andEAC tissue samples, silencing of which significantlyreduced aggressiveness of EAC cell lines OE33 andSKGT4 (66).

Taken together, genomic abnormalities play key rolesduring each stage of transformation from normal squa-mous epithelium to EAC.

Cancer-Related InflammationGastric and bile acid exposure in the esophageal epi-

thelium leads to the development of chronic inflamma-tory conditions mainly driven by elevated levels of proin-flammatory cytokines. Chronic inflammatory responsesinduce cell survival and increase cell proliferation, henceplay key roles in the development of EAC (67, 68). Expres-sions of various inflammatory molecules such as COX-2,NF-kB, interleukin (IL)-6, IL-8, and matrix metalloprotei-nases (MMP) have been evaluated as prognostic biomar-kers for BE/EAC development.

Exposure to gastric/bile acid and cytokines leads toincreased COX-2 expression (69). COX-2 is a rate-limitingenzyme that regulates synthesis of prostaglandins fromarachidonic acid. COX-2 directly increases cell prolifera-tion and promotes tumor invasion (69), andCOX-2–medi-ated increase in prostaglandin synthesis could result intumor growth and angiogenesis (70). COX-2 expressionhas been detected in disease-free esophageal tissue homo-genates using immunoblotting (69). In comparison withGERD, patients suffering from erosive reflux show slight-ly higher gene expressions of this enzyme in tissue sam-ples (71). Several studies have shown significantlyincreased COX-2 expression correlating with the diseaseprogression from Barrett’s esophagus to dysplasia andEAC (69, 72–75). Furthermore, expression levels of COX-2have been shown to have a prognostic value in EAC withhigher levels associated with poor survival and increasedchances of tumor relapse (76, 77).

Another well-studied inflammatory biomarker NF-kBis activated in response to exposure with bile acid andelevated NF-kB expression levels are found during Bar-rett’s esophagus, dysplasia, and adenocarcinoma (78–80).Activated NF-kB translocates from cytoplasm to nucleusand upregulates transcription of the genes involvedin inflammatory processes. Moreover, nuclear NF-kBexpression has been shown to be correlated with thepatient response to chemoradiotherapy.All of thepatientswho showed complete response to chemoradiotherapy






had elevated NF-kB levels pretreatment and showed lackof active NF-kB posttreatment (81).

In line with NF-kB and COX-2, expression of indi-vidual or combinations of proinflammatory cytokinesIL-1b, IL-6, IL-8, and TNF-a is significantly increased inBarrett’s esophagus and EAC as compared with squa-mous epithelium (82–84). IL-1b and IL-8 expressionlevels also correlate with the stage of EAC (79). Patientswho responded to neoadjuvant chemotherapy treat-ment showed significantly reduced expressions of IL-8 and IL-1b in postchemotherapy esophageal tissuesections (81). IL-6 is activated in response to reflux andthe IL-6/STAT3 antiapoptotic pathway may underliethe development of dysplasia and tumor (85). Serum IL-6 levels were reported to provide 87% sensitivity and92% specificity for EAC diagnosis in a recent retrospec-tive study (86). However, the study only comparedbetween healthy and EAC groups. It would be interest-ing to see how early it can diagnose EAC during theprocess of metaplasia–dysplasia. Combination of cyto-kines IFN-g , IL-1a, IL-8, IL-21, and IL-23 along withplatelet proteoglycan and miRNA-375 expression pro-filing has been shown to build an inflammatory riskmodel, which has clinical use to determine prognosis forpatients with EAC (67).

MMPs are a family of proteolytic enzymes involved inthe degradation of extracellular matrix components.MMPs play a role in both inflammation and tumormetas-tasis. Immunohistochemical staining forMMP-1, MMP-2,MMP-7, andMMP-9 has been reported to be significantlyhigher in EAC as compared with healthy individuals (87,88). Higher level of MMP-1 expression has been associ-ated with the lymph node metastases and possibly poorpatient survival (89). Expression ofMMP-9 is shown to bean early event during the EAC transformation and itsexpression levels are correlated with the progression ofthe disease (90–92). Activity of MMPs is inhibited by afamily of proteins called tissue inhibitors of metallopro-teinases (TIMP). Specifically, TIMP-3 gene is methylatedin EAC development and its reduced expression is asso-ciated with stage of the tumor and patient survival (93).On contrary, Salmela and colleagues described elevatedTIMP-1 and TIMP-3 expression in EAC tumor samples(88).

Although the underlying tissue inflammation is veryclosely associated with EAC development and severalinflammation-related biomarkers have been identified,these remain to be validated in large-scale biomarkerstudies.

Cell Cycle–Related AbnormalitiesTo compensate for the tissue damage induced by gas-

tric/bile acid, the underlying epithelium starts to prolif-erate rapidly and become uncontrolled resulting in neo-plasia. To meet the proliferation requirements, the cellshave to overcome cell-cycle checkpoints. Cyclin D1 over-expression is one such means by which cells overcomeG1–S checkpoint, and cyclin D1 immunohistochemical

staining has been proposed to identify patients with Bar-rett’s esophagus with an increased risk for EAC (94). Incontrast to cyclin D1, expression of p16 protein results incell-cycle arrest in G1 phase as it has been shown to inhibitcyclin-dependent kinase–induced phosphorylation ofretinoblastoma protein. Early genomic abnormalities dur-ing EAC development significantly affect p16 proteinexpression,which can bedeterminedusing immunostain-ing and implemented as a potential biomarker (95). Fur-ther large-scale trials are required to confirm cell-cycleabnormalities during EAC development to implementthem as a biomarker.

Bottom of the pyramid in Fig. 1 represents list ofbiomarkers in the initial stages of development. Tumorsharboring overexpression of growth factor receptors [EGFreceptor (EGFR) and HER-2] are associated with poorpatient survival (96, 97), whereas those overexpressingapoptosis regulator Bcl-2 protein showed prolonged sur-vival (98). Incipient angiogenesis is a marked feature ofBarrett’s esophagus and underlining tissue expressesangiogenesis markers VEGF and its receptors (99). Neo-vascularization continues as the disease progresses fromBarrett’s esophagus to EAC. Measuring the degree ofneovascularization correlated with histopathologic gradeof the tumor and associated with the patient survival(100). Expression of 2 prominent cell proliferation mar-kers, PCNA and Ki-67, has been described to be alteredduring BE–EAC development (101).

miRNAmiRNA was first discovered in Caenorhabditis elegans

(102) and since then it has beenwidely studied in a varietyof biologic phenomena. These short stretches of approx-imately 21 nucleotides do not code for protein but playimportant roles in gene regulation by either suppressingprotein synthesis or causing mRNA cleavage. UnlikesiRNA, miRNA can target multiple genes on remote lociand therefore control diverse group of proteins. Severalkey properties of carcinogenesis have been shown to beregulated via miRNA, for example, angiogenesis andmetastasis (103).

With increased biologic understanding ofmiRNAs andtheir role in cancer, they have been proposed in severaldifferent clinical applications including cancer diagnosisand tumor prognosis, tumor classification, and also as atherapeutic target for disease intervention. Differentialtissue miRNA expression has been observed in severaldifferent malignancies and these changes can be used fordiagnosis and classification of the tumors (103). miRNAbioarrays were first used to show differential miRNAexpression in healthy, Barrett’s esophagus, and EACtissue samples (104). Since then, a number of differentstudies have identified miRNA changes associated withthe development of the BE–EAC. Table 4 summarizesprimary findings of miRNA expression profiling studiesalong with statistical significance and fold-change values.Biologic significance for some of the miRNA-relatedchanges is discussed later.

Shah et al.





Smith and colleagues identified reduced expression ofmiR-200 and miR-141 in Barrett’s esophagus and EACtissue samples. They conducted bioinformatics analysisand correlated these miRNA expression changes withcellular processes such as cell cycle, cell proliferation,apoptosis, and cell migration (105). miR-196a, which isdescribedas amarker of progression fromBarrett’s esoph-agus to EAC, can increase cell proliferation and anchor-age-independent growth and inhibit apoptosis in EACcell lines in vitro (106). The downstream targets for miR-196a are verified to be Annexin A1, S100 calcium-bindingprotein A9, small proline-rich protein 2C, and Keratin 5,which showed reduced expression in EAC patient tissuesamples as compared with normal epithelium (106, 107).Several studies described in Table 4 report overexpressionof miR-192 during EAC carcinogenesis. miR-192 has beenreported tobe a target of p53 andhas been able to suppresscancer progression in osteosarcoma and colon cancer celllines throughp21 accumulation and cell-cycle arrest (108).As shown in Table 4,miR-21 is overexpressed during BE/EAC and it can function as an oncogene as shown intumors of breast, brain, lung, prostate, pancreas, colon,liver, and chronic lymphocytic leukemia. It negativelyregulates tumor- and metastasis-suppressor genes PTEN,TPM1, PDCD4, and Sprouty2 (109–112). miR-194 expres-sion is regulated by hepatocyte nuclear factor (HNF)-1atranscription factor, which is induced during BE/EACand may lead to upregulation of miR-194 (109). Higherexpression of miR-194 is also observed in metastaticpancreatic cell lines (113). Among miRNAs found to bedownregulated during EAC development, let-7 family ofmiRNAs is tumor-suppressive and negatively regulatesRas oncogene. Fassan and colleagues confirmed upregu-lation of HMGA2, which is one of the target of let-7miRNA, using immunohistochemistry in tissue samples(110, 112, 114). Further studies in the regards of miRNAandmiRNA target geneswill improve the biologic under-standing of EAC pathogenesis and may also providenovel molecular targets for disease intervention.Notably, miRNAs are found to be stable in serum

encapsulated in microvesicles and can be accessed easily(115). In fact, circulating miRNA profiling has showndistinct expression patterns in a number of cancers, otherthan EAC (116). This opens up new avenues for circulat-ing miRNA changes as a potential biomarker for EAC.

GlycoproteinsProtein glycosylation is a common posttranslational

modification with almost half of the proteins synthesizedundergoing 1 of the 2 major types either N-linked or O-linked glycan modifications. The biosynthetic process ofglycosylation is regulated by the expression and localiza-tion of glycosyltransferases/glycosidases and the avail-ability of substrate glycans (117).Aberrant glycosylation changes have previously been

reported in several different cancers namelybreast cancer,prostate cancer, melanoma, pancreatic cancer, ovariancancer, etc. (118, 119). These changes include truncated

forms of O-glycans, increased degree of branching in N-glycans, and elevated sialylation, sulfation, and fucosyla-tion with a range of other possible variations (119). Thedifferential glycosylation can alter protein interactions,stability, trafficking, immunogenicity, and function (118).Tumor-specific glycosylation changes are activelyinvolved in neoplastic progression, namely metastasis,as glycoproteins are found abundantly on cell surfacesand extracellularmatrices and therefore play a vital role incellular interactions.

Lectins are a family of glycan-binding proteins exten-sively used in glycobiology due to preferential binding ofeach lectin to recognize specific glycan structures (119,120). The first effort to identify differential glycosylationin the progression to Barrett’s esophagus and EAC wasmade in 1987 by Shimamoto and colleagues using differ-ential binding pattern to 5 lectins in tissue specimens(121). The glycoconjugate expression profile in Barrett’sesophagus was found to be significantly different fromnormal esophageal epithelium. Interestingly, glycoconju-gate expression between Barrett’s esophagus and normalduodenum was quite similar. There were minimal glyco-conjugate expression changes between Barrett’s esopha-gus and LGD. However, EAC tissue samples showedsignificantly different lectin-binding pattern than BE/LGD (121). Using rabbit esophageal epithelium, Poor-khalkali and colleagues showeddifferential lectin bindingin response to acid/pepsin exposure suggesting acidexposure can induce cell surface glycosylation changes(122). In 2008, Neumann and colleagues used 4 differentlectins to identify pathologic mucosal changes (123). Theyobserved 2 distinct lectin-binding patterns. Onewas asso-ciated with the GERD, whereas the other pattern wascharacteristic for Barrett’s esophagus mucosa. Specifical-ly,UEA (Ulex europaeus) lectin bindingwasupregulated inBarrett’s esophagus tissue sections, which suggests pos-sible increase in fucosylation during the disease progress(123). A recently published study has concluded thatdysplasia can alter glycan expression and lectin bindingto the tissue samples. Fluorescently labeled WGA (wheatgerm agglutinin) lectin-binding intensity was found to beinversely related to the degree of dysplasia (124). Further-more, the authors used fluorescent-capable endoscope exvivo in the study and followed all the protocols in amanner that exactly mimics a clinical study in vivo. Fol-lowed by topical fluorescein-labeled WGA spray, theauthors measured fluorescence in the tissue samples.Measurement of lectin fluorescence was a more sensitiveapproach to identify dysplastic lesions as compared withwhite light endoscopic technique. Their data show clinicaluse of such a lectin-based endoscopic technique if devel-oped further (124). In a phase III biomarker clinical trialstudy, Bird-Liberman and colleagues combined 3 differ-ent abnormalities to predict EAC progression in patientswith Barrett’s esophagus. Along with using conventionalLGD and DNA content abnormalities they used AOL(Aspergillus oryzae) lectin binding to the tissue samples,which detects presence of a1-6 fucose on the cell surface






Table 4. Summary of literature describing miRNA expression changes in BE/EAC

Sample size Upregulated in BE/EAC Downregulated in BE/EAC Ref.

71 (BE-12, Barrett'sesophagus withoutdysplasia-20, LGD-27,EAC/HGD-12)

miR-192 (P < 0.00001), miR-196a (P < 0.05):upregulated in Barrett's esophagus ascompared with healthy tissue.miR-196aexpression is correlated with progressionfrom IM-LGD-HGD-EAC (P < 0.005).

miR203 (P < 0.00001): downregulation inBarrett's esophagus as compared withhealthy tissue.

(209)

22 (Barrett's esophaguswithout dysplasia-11,Barrett's esophagus withdysplasia-11)

miR-15b (3.3-fold; P < 0.05), miR-203 (5.7-fold; P < 0.05): upregulated in dysplasia ascompared with nondysplastic Barrett'sesophagus.

miR-486-5p (4.8-fold; P < 0.05), miR-let-7a(3.3-fold; P < 0.05): downregulated indysplasia as compared with nondysplasticBarrett's esophagus.

(110)

100 (EAC-100, adjacentnormal tissue as control)

miR-21 (�3-fold; P < 0.05), miR-223 (�2-fold;P < 0.05), miR-192 (�3.5-fold; P < 0.05),and miR-194 (�3.5-fold; P < 0.05):upregulated in EAC as compared withadjacent normal tissue.

miR-203 (�3-fold; P < 0.05): downregulatedin EAC as compared with adjacent normaltissue.

(111)

25 (Healthy-9, BE-5,HGD-1, EAC-10)

miR-192 (1.7-fold; FDR < 1 e�07), miR-194(2-fold; FDR < 1e�07), miR-21 (3.7-fold;FDR ¼ 0.0003), miR-200c (1.9-fold; FDR ¼0.0015), miR-93 (1.3-fold; FDR ¼ 0.0108):upregulated in EAC as compared withBarrett's esophagus.

miR-27b (1.43-fold; FDR ¼ 0.0003), miR-342(1.25-fold; FDR ¼ 0.0015), miR-125b (2-fold; FDR ¼ 0.0108), miR-100 (1.25-fold;FDR ¼ 0.011): downregulated in EAC ascompared with Barrett's esophagus.

(104)

75 (Healthy-15, BE-15,LGD-15, HGD-15,EAC-15)

miR-215 (62.8-fold; P < 1e�07), miR-192(6.34-fold; P < 1e�07): upregulated inBarrett's esophagus in comparison withnormal tissue and remained at similar levelswith disease progress.

miR-205 (10-fold; P ¼ 1.39e�0.5), let-7c(2.04-fold; P ¼ 3.11e�05), miR-203 (6.67-fold; P ¼ 3.2e�0.5): downregulated inBarrett's esophagus in comparison withnormal tissue and remained at similar levelsas disease progresses.

(114)

91 (LGD-31, HGD-29, EAC-31, In all cases adjacentnormal tissue used as acontrol)

miR-200a (13.5-fold; P ¼ 0.02), miR-513(1.58-fold; P ¼ 0.03), miR-125b (9.2-fold; P¼ 0.04), miR-101 (1.83-fold; P¼ 0.04), miR-197 (1.61-fold; P ¼ 0.04): upregulated inLGD to HGD transition.

miR-23b (1.45-fold; P ¼ 0.007), miR-20b(1.56-fold; P¼ 0.01), miR-181b (2.22-fold;P¼ 0.03), miR-203 (1.49-fold; P¼ 0.03), miR-193b (2.70-fold; P ¼ 0.04), miR-636 (4.17-fold; P ¼ 0.04): downregulated in LGD toHGD transition. let-7a (1.75-fold; P ¼ 0.01),let-7b (1.59-fold; P ¼ 0.009), let-7c (1.69-fold; P ¼ 0.03), let-7f (1.69-fold; P ¼ 0.03),miR-345 (2-fold; P ¼ 0.02), miR-494 (1.72-fold; P ¼ 0.03), miR-193a (2.27-fold; P ¼0.05): downregulated in HGD-EACdevelopment process.

(112)

48 (BE-19, EAC-29) miR-21 (�2.8-fold; P < 0.05), miR-143(�11.3-fold;P

(125). Thus, monitoring tissue glycan changes can becombined with existing biomarkers to improve the pre-dictive power of the currently used biomarkers.A potential mechanism responsible for these changes is

considered to be bile acid exposure-induced gene expres-sion and secretory pathway changes in esophageal epi-thelium. Using carbohydrate-specific lectins that detectN- and O-linked glycosylation and core fucosylation,Byrne and colleagues have shown differential lectin bind-ing to the cell surface and differential intracellular local-ization when normal squamous and Barrett’s metaplasticcell lines were treated with deoxycholic acid (126). Nan-carrowand colleagues profiledwhole-genome expressionin normal squamous esophageal epithelium, Barrett’sesophagus, and EAC and concluded that Barrett’s esoph-

agus is a tissue with enhanced glycoprotein synthesismachinery to provide strong mucosal defense againstacid exposure (127).

Outlook—Circulating BiomarkersLast 3 decades showed continuously increased EAC

incidences and similar trend is expected in future becauseof rising incidences of obesity and GERD in the popula-tion.Current endoscopic screeningprogrammightbenefitthe highest risk population to monitor disease progres-sion. Monitoring dysplasia in the tissue samples has notprovided fruitful outcome for early diagnosis; however,inclusion of the genomic and cell-cycle biomarkers hasshown definite improvement in the predictive powerover currently used histologic technique. Any biomarker

Table 4. Summary of literature describing miRNA expression changes in BE/EAC (Cont'd)

Sample size Upregulated in BE/EAC Downregulated in BE/EAC Ref.

11 (EAC-11, differentlesions were collectedfrom these patients andclassified into Barrett'sesophagus, LGD, HGD,and EAC)

miR-196a is overexpressed in early EAC(151-fold) > HGD (62.2-fold; P ¼ 0.00002) >LGD (31.1-fold; P ¼ 0.0005) > Barrett'sesophagus (28.9-fold; P ¼ 0.00001). Foldchanges are calculated as compared withnormal epithelium.

— (107)

45 (patients with EACundergoing surgery)

miR-143 (P ¼ 0.0148), miR-199a_3p (P ¼0.0009), miR-199a_5p (P ¼ 0.0129), miR-100 (P ¼ 0.0022) and miR-145 (P ¼ 0.1176)expression

early diagnostic biomarkers for esophageal adenocarcinoma ...early diagnostic biomarkers for...

Documents