ACC

XW*CBT

BtiteAiaXXwbteb37w2ftGpma2AOa16Aii

Hrg

GASTROENTEROLOGY 2006;131:420–427

denosine Diphosphate Ribosyl Transferase and X-Ray Repairross-Complementing 1 Polymorphisms in Gastric Cardiaancer

IAOPING MIAO,* XUEMEI ZHANG,*,‡ LINGQIANG ZHANG,§ YONGLI GUO,* BINGTAO HAO,§,�

EN TAN,* FUCHU HE,§,� and DONGXIN LIN*Department of Etiology & Carcinogenesis, Cancer Institute and Hospital, Chinese Academy of Medical Sciences and Peking Union Medicalollege, Beijing; ‡Department of Biology Sciences, North China Coal Medical College, Tangshan; §Department of Genomics and Proteomics,eijing Institute of Radiation Medicine, Beijing Proteome Research Center, Beijing; and �Department of Biology Sciences and Biotechnology,

singhua University, Beijing, China

rnmicrmDdssdjcsrapcDbXXot

avo

risc

ackground & Aims: Adenosine diphosphate ribosylransferase (ADPRT) and x-ray repair cross-complement-ng 1 (XRCC1) are major DNA base excision repair pro-eins acting interactively in repair processes. This studyxamined the effects of ADPRT Val762Ala and XRCC1rg399Gln polymorphisms on ADPRT-XRCC1 interaction

n vitro in cells and their contributions to gastric cardiadenocarcinoma (GCA) risk. Methods: The ADPRT-RCC1 interaction in cells transfected with ADPRT andRCC1 variant complementary DNA (cDNA) constructsere examined by immunoprecipitation and immuno-lotting analysis. Genotypes were analyzed in 500 pa-ients and 1000 controls, and odds ratios (ORs) werestimated by logistic regression. Results: Interactionsetween ADPRT-762Val and XRCC1-399Arg or XRCC1-99Gln were robust, but interactions between ADPRT-62Ala and either XRCC1-399Arg or XRCC1-399Glnere very weak. A case-control analysis showed ORs of.17 (95% CI, 1.55–3.04) and 1.61 (95% CI, 1.06–2.44)or GCA in the ADPRT Ala/Ala or XRCC1 Gln/Gln geno-ype carriers, respectively, compared with noncarriers.ene-gene interaction of ADPRT and XRCC1 polymor-hisms increased the OR of GCA in a multiplicativeanner (OR for the presence of both ADPRT Ala/Ala

nd XRCC1 Gln/Gln genotypes, 6.43; 95% CI, 1.80–2.97). A supermultiplicative joint effect between theDPRT polymorphism and smoking was observed. TheRs (95% CIs) of the Ala/Ala genotype for nonsmokersnd smokers who smoked <24 or >24 pack-years were.44 (0.89–2.32), 2.00 (1.09–3.67), or 3.19 (1.59–.42), respectively (Ptrend test � .008). Conclusions: TheDPRT and XRCC1 polymorphisms confer host susceptibil-

ty to GCA, which might result from reduced ADPRT-XRCC1nteraction and attenuated base excision repair capacity.

uman genomic DNA can be damaged by environ-mental factors such as ultraviolet light, ionizing

adiation, chemical carcinogens, and certain products

enerated by normal cellular metabolism. If not properly

epaired, such DNA damage can induce mutations, ge-ome instability, and DNA strand breaks, which in turnay cause cellular malignant transformation and tumor-

genesis. Nevertheless, human cells have evolved a set ofomplex DNA repair systems that safeguard the integ-ity of genome to minimize the consequences of detri-ental damages.1 As a result, the normal functions ofNA repair proteins are critical for cells to removeamage and thus prevent carcinogenesis. The base exci-ion repair (BER) pathway, among other DNA repairystems, is basically responsible for rectifying DNA baseamage and single-strand breaks.1,2 BER includes 2 ma-or processes, ie, excision of damaged base residues andore BER reaction including strand incision at the abasicite, 1-nucleotide gap-filling reaction, and sealing of theemaining nick.1,2 It is well known that several proteinsre involved in these steps, of which adenosine diphos-hate ribosyl transferase (ADPRT) and x-ray repair cross-omplementing 1 (XRCC1) play key roles. In response toNA damage, ADPRT specifically binds to DNA strandreaks at which it is autoactivated and recruits theRCC1-Ligase III� complex to stimulate BER, andRCC1 subsequently interacts with ADPRT to recruit

ther partner proteins such as DNA polymerase � (Plo�)o execute and complete BER processes.3,4

Genetic variants have been identified for both ADPRTnd XRCC1 genes.5–8 There are at least 38 reportedariants in ADPRT (http://snp500cancer.nci.nih.gov),ne of which is a T to C transition at codon 762 of

Abbreviations used in this paper: ADPRT, adenosine diphosphateibosyl transferase; BER, base excision repair; IB, immunoblot; IP,mmunoprecipitation; PCR-RFLP, polymerase chain reaction-based re-triction fragment length polymorphism; XRCC1, X-ray repair cross-omplementing 1.© 2006 by the American Gastroenterological Association Institute

0016-5085/06/$32.00

doi:10.1053/j.gastro.2006.05.050

AtsccaXdtAtmgmfgcsXomatcn

cCfctrtmatDDahutirttmG

taW

tpf

GitHtcvhfieoeTvsGtcdcpscydm1AsdtIS

tbmfpssd

August 2006 POLYMORPHISMS AND RISK OF GASTRIC CANCER 421

DPRT located in the sixth helix of catalytic domainhat causes Val-to-Ala amino acid substitution. A recenttudy has shown that the 762Ala polymorphism is asso-iated with altered ADPRT function, with the Ala alleleontributing to significantly low poly(ADP-ribosyl)ationctivities in an allele dosage-dependent manner.5 In theRCC1 gene, at least 17 variants have been identified toate (http://snp500cancer.nci.nih.gov), and, amonghem, the G to A polymorphism at codon 399 results inrg-to-Gln amino acid change, which is situated within

he BRCT-1 region harboring the ADPRT-binding do-ain.4,7 An altered DNA repair activity has been sug-

ested to be associated with the XRCC1 399Gln poly-orphism.9–11 We have recently shown that these

unctional polymorphisms in ADPRT and XRCC1 havereat impact on susceptibility to smoking-related lungancer and esophageal cancer. Furthermore, a statisticallyignificant interaction between the ADPRT 762Ala andRCC1 399Gln polymorphisms in risk of the cancer was

bserved.8,12 Our data suggest that these genetic poly-orphisms may obstruct the interaction between ADPRT

nd XRCC1 at protein level in the BER processes andhus abolish or reduce DNA repair capacity in cellsarrying the variant alleles. However, this hypothesis hasot been examined yet at the molecular level.Gastric cardia adenocarcinoma (GCA) is a common

ancer in China as well as in the rest of the world. Inhina, this cancer is more prevalent in areas of high-risk

or esophageal cancer,13 suggesting that it may shareommon risk factors with esophageal cancer. Althoughhe integrated etiology of esophageal cancer and GCAemains to be fully clarified, several environmental fac-ors such as cigarette smoking, heavy alcohol drinking,icronutrient deficiency, dietary carcinogen exposure,

nd gastroesophageal reflux disease have been linked tohese cancers.14–17 All these factors can induce or enhanceNA damage mediated by either oxidative stress orNA-binding electrophiles, which in turn may initiate

nd/or promote tumorigenesis. However, even in theigh-risk areas, only a small fraction of exposed individ-als develop GCA in their life span, suggesting thathere may be important genetic basis rendering suchndividuals more susceptible to the cancer. Because DNAepair mechanisms play a pivotal role in preventingumorigenesis, suboptimal DNA repair capacityhroughout lifetime caused by germ-line polymorphismsight be a susceptibility factor for the development ofCA.In this report, we describe a case-control study to test

he hypothesis that the genetic polymorphisms of ADPRTnd XRCC1 contribute to host susceptibility to GCA.

e also examined the effect of these polymorphisms on t

he interaction between ADPRT and XRCC1 at therotein level in cells to explore further the underlyingunctional mechanisms.

Materials and Methods

Study Subjects

The case-control study consisted of 500 patients withCA and 1000 cancer-free controls. All subjects were biolog-

cally unrelated ethnic Han Chinese. Patients were consecu-ively recruited from January 1997 to July 2004 at the Cancerospital, Chinese Academy of Medical Sciences (Beijing). In

his study, GCA was defined as tumors arising at the gastricardia and/or gastroesophageal junction with or without in-olvement of other esophageal and/or gastric subsites. Fiveundred forty-four patients with histopathologically con-rmed GCA, regardless of age, sex, and tumor stage, werenrolled, and the response rate was 92% because of reluctancef 44 patients to participate in the study. Tumors confinedntirely to the esophageal or gastric subsites were excluded.hree hundred fifty-six patients were participants in our pre-ious molecular epidemiologic study of GCA.18 In the presenttudy, we added 144 patients, for a total of 500 patients withCA. Control subjects were cancer-free individuals living in

he Beijing region, who were selected from a communityancer-screening program for early detection of cancer con-ucted during the same time period as the patients wereollected.19 Briefly, the control subjects were selected from aool of 2800 individuals based on physical examinations. Theelection criteria include no individual history of cancer, andontrols were frequency matched to GCA patients on age (�5ears) and sex. The characteristics of 789 control subjects wereescribed previously.18 In this study, we also selected 211ore control subjects from the same database, for a total of

000 control subjects, to make a 2:1 match to GCA patients.t recruitment, informed consent was obtained from each

ubject, and each participant was then interviewed to collectetailed information on demographic characteristics and life-ime history of tobacco use. This study was approved by thenstitutional Review Board of the Chinese Academy of Medicalciences Cancer Institute and Hospital.

Polymorphism Analysis

ADPRT Val762Ala and XRCC1 Arg399Gln geno-ypes were analyzed using polymerase chain reaction (PCR)-ased restriction fragment length polymorphism (RFLP)ethods as described previously.12 All subjects were success-

ully genotyped. To ensure quality control, genotyping waserformed without knowledge of case/control status of theubjects, and a 15% random sample of patients and controlubjects was genotyped twice by different persons; the repro-ucibility was 100%.

Plasmid Construction

The complementary DNA (cDNA) encoding wild-

ype ADPRT (762Val) was generated by PCR from the human

ftvce(pmt5c7tcpXpUTaE3MMst

twUCHetMCa3SDp

wcwcpPcaui

Cwc(io4Fi(ptiauI

gtgcvOnopccpgfesswOrfS

jtjabs(

422 MIAO ET AL GASTROENTEROLOGY Vol. 131, No. 2

etal liver cDNA library (Clontech, Mountain View, CA), withhe forward primer 5=-gaagatcttcggatggcggagtcttcg-3= and re-erse primer 5=-ccgctcgagcgggtcgacttaccacagggaggt-3=, whichontain 5=-BglII and 3=-SalI sites (underlined sequences). Tonsure high-fidelity amplification, LA Taq polymeraseTaKaRa, Dalian, China) was used in the PCR. The PCRroduct was subsequently used as a template to generateutant ADPRT (762Ala) cDNA by using site-specific mu-

agenesis with aforementioned PCR primers and primers of=-caggccaaggccgaaatgcttgacaac-3= and 5=-catttcggccttggcctg-acact-3= (underlined sequences encode 762Ala). Both ADPRT62Val and 762Ala cDNA were ligated into pFlag-CMV-2 vec-or (Sigma Chemical Company, St. Louis, MO), and the resultingonstructs were designated as pFlag-CMV-2-ADPRT-762Val orFlag-CMV-2-ADPRT-762Ala. For XRCC1 constructs, 2RCC1 cDNA clones with 399Arg or 399Gln genotype, kindlyrovided by Prof. Keith W. Caldecott of University of Sussex,nited Kingdom, were used as template in PCR amplification.he PCR primers used were 5=-ccggaattcacatgccggagatccgc-3=nd 5=-gaagatcttcaggcttgcggcaccaccccata-3=, which consist of 5=-coRI and 3=-BglII sites (underlined sequences). Both XRCC199Arg and XRCC1 399Gln cDNA were ligated into pCMV-yc vector (Clontech), respectively, and designated as pCMV-yc-XRCC1-399Arg or pCMV-Myc-XRCC1-399Gln. All con-

tructs used in this study were restriction mapped and sequencedo confirm their authenticity.

Cell Culture and Transfection

Human breast cancer cell line MCF7 cells were main-ained in Dulbecco’s modified Eagle medium supplementedith 10% fetal bovine serum, 2 mmol/L L-glutamine, 100/mL penicillin, and 100 U/mL streptomycin (Invitrogen,arlsbad, CA) in a humidified, 5% CO2 incubator at 37°C.uman gastric cancer cell line MGC-803 cells and Barrett’s

sophageal adenocarcinoma cell line SEG-1 cells were main-ained in RPMI 1640 containing 10% fetal bovine serum. TheCF7 and MGC-803 cells were cotransfected with pFlag-MV-2-ADPRT-762Val or pFlag-CMV-2-ADPRT-762Aland pCMV-Myc-XRCC1-399Arg or pCMV-Myc-XRCC1-99Gln plasmids using Lipofectamine 2000 (Invitrogen), andEG-1 cells were transfected using FuGene 6 reagent (Rocheiagnostics, Shanghai, China) according to the manufacturer’srotocol.

Immunoprecipitation and Immunoblotting

Immunoprecipitation (IP) and immunoblotting (IB)ere performed as previously described20 with minor modifi-

ations. Briefly, the cells, harvested 48 hours after transfection,ere washed twice with cooled phosphate-buffered saline, and

ell lysates were prepared in lysis buffer (50 mmol/L Tris-HCl,H 7.5, 150 mmol/L NaCl, 1% Tween 20, 0.2% Nonidet-40, 10% glycerol) supplemented with protease inhibitorocktail (Roche) and phosphatase inhibitors (10 mmol/L NaFnd 1 mmol/L Na3VO4) by sonication and subsequent centrif-gation at 15,000g for 10 minutes. The supernatant was

ncubated with mouse control IgG (Santa Cruz Biochemical 7

o, Santa Cruz, CA) at 4°C for 1 hour and then were incubatedith protein A/G-agarose for another 1 hour. After a brief

entrifugation, the pellet was washed 3 times with lysis bufferdesignated IP-IgG), and the supernatant was divided equallynto 2 parts, and each was incubated with anti-Myc antibodyr anti-Flag antibody and protein A/G-agarose (Santa Cruz) at°C to immunoprecipitate Myc-tagged XRCC1 proteins andlag-tagged ADPRT proteins, respectively. The lysates andmmunoprecipitates were immunoblotted with the anti-MycClontech; 1:1000) or anti-Flag M2 (Sigma Chemical Com-any; 1:1000) antibodies after gel electrophoresis and electro-ransfer of separated proteins on nitrocellulose sheets. Afterncubation with horseradish peroxidase-conjugated secondaryntibody (Santa Cruz; 1/2000 dilution), bands were visualizedsing SuperSignal chemiluminescence kit (Pierce, Rockford,L) according to the manufacturer’s instructions.

Statistical Analysis

The �2 test was used to examine differences in demo-raphic variables, smoking, and distribution of genotypes be-ween patients and control subjects. The associations betweenenotype and risk of gastric cardia cancer were estimated byalculating odds ratios (ORs) and their 95% confidence inter-als (CIs) with unconditional logistic regression models. TheRs were adjusted for age, sex, and smoking status. Theumber of pack-years smoked was determined as an indicatorf the cumulative cigarette dose level (pack-years � cigaretteser day/20 � years smoked). Light and heavy smokers wereategorized by using the 50th percentile pack-year value of theontrol subjects as the cut-off points (ie, �24 and �24ack-years). We tested the null hypotheses of multiplicativeene-gene and gene-smoking joint effect evaluated departuresrom multiplicative joint effect models by including mainffect variables and their product terms in the logistic regres-ion model. A more than multiplicative joint effect21 wasuggested when OR11 � OR10 � OR01. The homogeneity testas done to compare the difference between smoking-relatedRs among different genotypes or between the product of

elated ORs and the joint effect OR. All analyses were per-ormed with the computer programs of Statistical Analysisystem (version 6.12; SAS Institute, Cary, NC.

Results

Subject Characteristics and GenotypeDistributions

Baseline characteristics of patients and control sub-ects are summarized in Table 1. The sex and age distribu-ions were fairly identical among patients and control sub-ects, indicating that the matching of controls to cases wasdequate. There were no statistically significant differencesetween patients and control subjects in terms of smokingtatus and levels of pack-year smoked. Genotyping resultsTable 2) showed that the allele frequencies for the ADPRT

62Ala and XRCC1 399Gln variants were 0.36 and 0.27 in

cTV1ftwc.pcn6r(r

urhgpihPtw

311ctoyrs

acpacX(aclaaM

T

S

A

S

P

a

b

l

T

A

X

a

jb

August 2006 POLYMORPHISMS AND RISK OF GASTRIC CANCER 423

ontrol subjects and 0.44 and 0.32 in patients, respectively.he frequencies of the 3 ADPRT genotypes, ie, 762Val/al, -Val/Ala, and -Ala/Ala, were 39.6%, 49.2%, and1.2%, respectively, among control subjects, which con-ormed to the Hardy–Weinberg equilibrium. The respec-ive ADPRT genotype frequencies among GCA patientsere 30.0%, 51.4%, and 18.6%, which differed signifi-

antly from those among control subjects (�2 � 22.12, P �0001), with the Ala/Ala genotype being significantly morerevalent in patients than in control subjects. The frequen-ies of XRCC1 399Arg/Arg, -Arg/Gln, and -Gln/Gln ge-otypes among control subjects were 53.2%, 40.2%, and.6%, which conformed to the Hardy–Weinberg equilib-ium but significantly differed from those among patients44.2% Arg/Arg, 46.8% Arg/Gln, and 9.0% Gln/Gln,espectively) (�2 � 11.40, P � .003).

Association Between IndividualPolymorphism and GCA Risk

The unconditional logistic regression model wassed to estimate the association between genotypes andisk of GCA (Table 2). It was observed that the odds ofaving the ADPRT 762Ala/Ala genotype in the GCAroup was 2.17 (95% CI: 1.55–3.04; P � .0001) com-ared with the 762Val/Val genotype. A significantlyncreased OR was also associated with the 762Val/Alaeterozygous genotype (OR, 1.35; 95% CI: 1.06–1.72;� .009), suggesting an allele-dose relationship (trend

est, P � .0001). Similarly, the XRCC1 399Gln allele

able 1. Frequency Distribution of Select Characteristics byCase Control Status

Variable

Patients(n � 500)

Controlsubjects

(n � 1000)P

valueaNo. (%) No. (%)

ex 1.00Male 430 (86.0) 860 (86.0)Female 70 (14.0) 140 (14.0)

ge (yrs) 1.00�40 14 (2.8) 28 (2.8)41–50 62 (12.4) 124 (12.4)51–60 165 (33.0) 330 (33.0)�60 259 (51.8) 518 (51.8)

moking statusb .14No 237 (47.4) 434 (43.4)Yes 263 (52.6) 566 (56.6)

ack-years smoked .65�24 128 (48.7) 285 (50.4)�24 135 (51.3) 281 (49.6)

Two-sided �2 test.Nonsmokers were defined as subjects who smoked �10 cigarettesifetime.

as shown to be a risk allele. Subjects carrying the t

99Gln/Gln or 399Arg/Gln genotype had an odds of.61 (95% CI: 1.06–2.44; P � .017) or 1.40 (95% CI:.12–1.75; P � .003) for developing GCA, respectively,ompared with subjects carrying the 399Arg/Arg geno-ype (trend test, P � .001), suggesting a dominant effectf this polymorphism in GCA. In the stratification anal-sis, age and smoking had no effect on the risk of GCAelated to both ADPRT and XRCC1 genotypes (data nothown).

Effects of Polymorphisms on ADPRT andXRCC1 Interaction

It has been well known that ADPRT and XRCC1ct interactively at the protein level in the BER pro-ess.3,4 We therefore examined whether the investigatedolymorphisms obstruct the interaction between ADPRTnd XRCC1 at the protein level. The expression plasmidsonsisted of Myc-XRCC1-399Arg (designed as R), Myc-RCC1-399Gln (designated as Q), Flag-ADPRT-762Val

designed as V), or Flag-ADPRT-762Ala (designed as A)nd were constructed and cotransfected into human can-er cells (designed as RV, QV, RA, and QA). Equal cellysates were immunoprecipitated with either anti-Myc ornti-Flag antibody in reciprocal coimmunoprecipitationssays (Co-IP). The Co-IP experiments were performed inCF-7 cells (Figure 1A), MGC-803 cells (Figure 1B),

able 2. Genotype Frequencies of ADPRT and XRCC1Among Cases and Control Subjects and TheirAssociation With Gastric Cardia Adenocarcinoma

Genotype

Controlsubjects

(n � 1000)Patients

(n � 500)

PtrendNo. (%) No. (%)

DPRT Val762AlaVal/Val 396 (39.6) 150 (30.0)Val/Ala 492 (49.2) 257 (51.4)Ala/Ala 112 (11.2) 93 (18.6)One Ala,

OR (95% CI)a 1.35 (1.06–1.72)Two Ala,

OR (95% CI)a 2.17 (1.55–3.04)�.0001b

RCC1 Arg399GlnArg/Arg 532 (53.2) 221 (44.2)Arg/Gln 402 (40.2) 234 (46.8)Gln/Gln 66 (6.6) 45 (9.0)One Gln,

OR (95% CI)a 1.40 (1.12–1.75)Two Gln,

OR (95% CI)a 1.61 (1.06–2.44).0009b

Data were calculated by unconditional logistic regression and ad-usted for sex, age, and smoking status.Tests for trend of odds were 2-sided and based on likelihood ratio

ests assuming a multiplicative model.

awcpciawbn53t71

sle-(8ici17apb

FR(MApmsa

424 MIAO ET AL GASTROENTEROLOGY Vol. 131, No. 2

nd SEG-1 cells (Figure 1C). Western blot analysis ofhole cell lysates showed that, under our experimental

onditions, the constructed plasmids expressed respectiveroteins, and the expression levels of each plasmid inells were similar (Figure 1I), indicating that our exper-mental model is valid. It was found that a similarmount of XRCC1-399Arg and XRCC1-399Gln proteinas specifically immunoprecipitated by anti-Myc anti-ody (Figure 1II, upper panels, lanes 2, 4, 6, and 8) butot by control IgG (Figure 1II, upper panels, lanes 1, 3,, and 7). Strikingly, interactions between XRCC1-99Arg or -399Gln and ADPRT-762Val were robust, andhe levels of the 2 XRCC1 isoforms bound to ADPRT-62Val were quite similar in the 3 cell lines (Figure

igure 1. Differential protein-protein interaction between ADPRT and) or Myc-XRCC1-399Gln (designated as Q) was respectively coexpres

designated as A) in human breast cancer cell line MCF-7 cells (deGC-803 cells (B), and esophageal adenocarcinoma cell line SEG-1DPRT variants were detected in cell lysates by immunoblotting withreimmunoprecipitated with mouse control IgG followed by immunoponoclonal antibody (III). The immunoprecipitates including IP-IgG a

ubjected to 10% SDS-PAGE and immunoblotting analysis with Myc ore identical. IgG HC, heavy chain of immunoglobin G.

AII, 1BII, and 1CII, bottom panel, lanes 2 and 4), o

uggesting that the XRCC1 399Arg¡Gln change hadittle effect on the XRCC1-ADPRT interaction. How-ver, the interactions between either XRCC1-399Arg or399Gln and ADPRT-762Ala, if any, were very weakFigure 1AII, 1BII, and 1CII, bottom panel, lanes 6 and). In reciprocal experiments, ADPRT proteins weremmunoprecipitated by the anti-Flag antibody, and theoimmunoprecipitated XRCC1 proteins were detected inmmunoblotting with anti-Myc. As shown in FigureIII, the amount of XRCC1-399Arg bound to ADPRT-62Ala was considerably reduced in MCF-7 cells andlmost absent in both MGC-803 and SEG-1 cells com-ared with that bound to ADPRT-762Vla (Figure 1III,ottom panel, lane 6 vs lane 2). Similar results were

1 variants in human cancer cells. Myc-XRCC1-399Arg (designated asith either Flag-ADPRT-762Val (designated as V) or Flag-ADPRT-762Alad as RV, QV, RA, and QA, respectively) (A), gastric cancer cell line(C) for 24 hours. Similar expression levels of different XRCC1 and

-Myc, anti-Flag, and anti-�-actin antibodies (I). The cell lysates wereitation with anti-Myc monoclonal antibody (II) and with anti-Flag M2-Myc or IP-Flag (lanes 1–8) as well as the RV lysate (lane 9) wereantibody. The experiments were repeated 3 times, and the results

XRCCsed wsignecellsanti

recipnd IPr Flag

btained for the interaction between XRCC1-399Gln

avAaX

aftgBaA6fGA�Ac

XAn

sotC9aea16stVnes

aiogoAr

T

a ted fo

T

NS

a

b

c

August 2006 POLYMORPHISMS AND RISK OF GASTRIC CANCER 425

nd ADPRT-762Ala (Figure 1III, bottom panel, lane 8s lane 4). Taken together, these results indicate that theDPRT 762Val¡Ala substitution significantly attenu-

tes the protein-protein interaction between ADPRT andRCC1.

Effects of Gene-Gene and Gene-SmokingInteraction on Risk of GCA

We next analyzed the joint effect between ADPRTnd XRCC1 genotypes in the risk of GCA (Table 3). It wasound that the cases carrying the ADPRT Ala/Ala geno-ype were also more likely to carry the XRCC1 Gln/Glnenotype than controls (1.6% vs 0.4%; P � .001).ecause the OR for subjects with the ADPRT Ala/Aland XRCC1 Gln/Gln genotypes vs subjects with theDPRT Val/Val and XRCC1 Arg/Arg genotypes, ie,.43 (Table 3), is nearly equal to the product of the ORor subjects with the ADPRT Val/Val and XRCC1 Gln/ln genotypes and the OR for subjects with the ADPRTla/Ala and XRCC1 Arg/Arg genotypes, ie, 2.23 � 2.956.58, a multiplicative joint effect between the ADPRT

la/Ala and XRCC1 Gln/Gln genotype may exist in in-reasing the risk for the development of GCA.

We further examined the joint effect of the ADPRT orRCC1 polymorphism and tobacco smoking (Table 4).mong nonsmokers, the ADPRT Ala/Ala genotype wasot associated with increased risk of GCA; however, among

able 3. Risk of Gastric Cardia Adenocarcinoma Associated

Genotypes

PatienADPRT Val762Ala XRCC1 Arg399Gln

Val/Val Arg/Arg 55Val/Val Arg/Gln 80Val/Val Gln/Gln 15Val/Ala Arg/Arg 119Val/Ala Arg/Gln 116Val/Ala Gln/Gln 22Ala/Ala Arg/Arg 47Ala/Ala Arg/Gln 38Ala/Ala Gln/Gln 8

Data were calculated by unconditional logistic regression and adjus

able 4. Risk of Gastric Cardia Adenocarcinoma Associated

Smoking status Val/Vala OR (95% CI)b Va

onsmoker 73/160 1.00 12moker 77/236 0.75 (0.50–1.11) 13�24 pack-years 40/128 0.68 (0.43–1.09) 6�24 pack-years 37/108 0.81 (0.50–1.33) 7

Number of patients/number of control subjects.Data were calculated by logistic regression, with the Val/Val genoty

P � .01, test for homogeneity between smoking-related ORs among Val/

mokers, subjects carrying the Ala/Ala genotype had an ORf 2.43 (95% CI: 1.47–4.04) that was significantly higherhan the product of the OR for smoking (OR, 0.75; 95%I: 0.50–1.11) and the OR for the genotype (OR, 1.44;5% CI: 0.89–2.32). When smoking status was addition-lly stratified by pack-year value, a multiplicative jointffect was found between the homozygous Ala/Ala genotypend various smoking doses, with the ORs of 2.00 (95% CI:.09–3.67) for �24 pack-years and 3.19 (95% CI: 1.59–.42) for �24 pack-years (Ptrend � .008) These data clearlyuggest a multiplicative joint effect between smoking andhe ADPRT polymorphism. However, the heterozygousal/Ala genotype had no such an effect (Table 4). There waso joint effect between the XRCC1 polymorphism andither smoking status or pack-years smoked (data nothown).

Discussion

It is well-known that DNA repair mechanisms playcrucial role in maintaining genomic stability and protect-

ng against carcinogenesis.1,22 Accumulating evidence dem-nstrates that impaired DNA repair capacity resulting fromerm-line polymorphisms is associated with increased riskf cancer development.23–25 We have previously shown thatDPRT 762Ala polymorphism is associated with increased

isk of lung cancer and esophageal cancer and significantly

ADPRT Genotypes by XRCC1 Genotype

. (%)Control subjects,

No. (%) OR (95% CI)a

0) 211 (21.1) 1.000) 158 (15.8) 1.91 (1.28–2.86)) 27 (2.7) 2.23 (1.09–4.53)8) 260 (26.0) 1.69 (1.17–2.45)2) 197 (19.7) 2.25 (1.53–3.29)) 35 (3.5) 2.21 (1.17–4.16)) 61 (6.1) 2.95 (1.82–4.79)) 47 (4.7) 2.96 (1.75–5.02)) 4 (0.4) 6.43 (1.80–22.97)

r sex, age, and smoking status.

ADPRT Genotypes by Smoking Status

ADPRT genotype

a OR (95% CI)b Ala/Alaa OR (95% CI)b

0 1.26 (0.88–1.79) 42/64 1.44 (0.89–2.32)2 1.05 (0.72–1.52) 51/48 2.43 (1.47–4.04)c

6 1.06 (0.69–1.64) 27/31 2.00 (1.09–3.67)c

6 1.06 (0.70–1.61) 24/17 3.19 (1.59–6.42)c

reference group and adjusted for age and sex.

With

ts, No

(11.(16.(3.0(23.(23.(4.4(9.4(7.6(1.6

With

l/Ala

2/215/281/124/15

pe as

Val and Ala/Ala genotypes.

iieoaAhtGMbc3IespsB

XcU(ppiHeAttrTcoGXpits

sG2prHaCc

(bA((rrettrrsstt25ltpSTsc

raAnciw17tpmXAtpgwmprctmm

426 MIAO ET AL GASTROENTEROLOGY Vol. 131, No. 2

nteracts with XRCC1 399Gln polymorphism to furtherncrease risk of the cancers.8,12 In the present study, wexamined whether these 2 polymorphisms have an impactn the risk of GCA. Based on the analysis of 500 patientsnd frequency-matched 1000 controls, we found that theDPRT 762Ala allele is a hazard allele, and the odds ofaving the Ala/Ala genotype in the GCA group was morehan 2. We also detected a moderately increased OR forCA that was associated with the XRCC1 399Gln allele.oreover, we observed a statistically significant interaction

etween the ADPRT and XRCC1 polymorphisms; subjectsarrying both the ADPRT 762Ala/Ala and XRCC199Gln/Gln genotypes had an OR of more than 6 for GCA.n addition, a significant gene-environment interaction wasvident between the ADPRT 762Ala/Ala genotype andmoking. These results are generally in agreement with ourrevious findings from lung cancer and esophageal cancertudies8,12 and suggest that genetic polymorphisms in theER genes are also involved in the etiology of GCA.It has been shown that interactions of ADPRT with

RCC1 as well as other partner proteins such as Pol� areritical for stimulating and executing BER processes.3,4,26

pon binding to DNA, ADPRT becomes auto-poly-ADP-ribosyl)ated, allowing itself to interact with otherroteins involved in BER. The 762Val¡Ala polymor-hism in ADPRT has been shown to have a substantialmpact on ADPRT poly(ADP-ribosyl)ation activity.5

owever, little or nothing is known so far about theffect of this polymorphism on the interaction betweenDPRT and XRCC1 at the protein level. Our data in

he present study provide evidence for the first time thathe 762Val¡Ala polymorphism in ADPRT significantlyeduces its capacity to interact with XRCC1 in cells.his functional relevance of ADPRT polymorphism isonsistent with molecular epidemiologic findings, dem-nstrating that ADPRT 762Ala as an at-risk allele forCA. Because normal interaction of ADPRT withRCC1 is necessary for stimulating and executing BERrocesses, reduced interaction of these 2 proteins maympair BER capacity and, thus, confer the host carryinghe ADPRT 762Ala allele susceptible to genotoxic in-ults and consequent cancer development.

Although there is so far no published case-controltudy on polymorphism in ADPRT and susceptibility toCA and our present study remains the first one, at leaststudies have been published regarding the XRCC1

olymorphism and risk of GCA. These 2 studies withelatively small sample size were conducted in Chinesean populations, but the results are conflicting. Shen at

l27 reported in a case-control study conducted in southhina that the XRCC1 399Gln/Gln genotype was asso-

iated with a nonsignificantly increased risk of GCA X

OR, 1.53; 95% CI: 0.98–2.39); however, subjects withoth XRCC1 399Gln/Gln or 399Arg/Gln and 194Arg/rg genotypes had a significantly higher risk for GCA

OR, 2.18; 95% CI: 1.21–3.94) and total gastric cancerOR, 1.73; 95% CI: 1.12–2.69). In contrast with thiseport, Ratnasinghe et al28 showed that in Linxian, aural county in north central China where the rates ofsophageal squamous cell carcinoma and GCA are amonghe highest in the world, subjects with at least 1 copy ofhe XRCC1 399Gln allele were associated with reducedisk of GCA (OR, 0.60; 95% CI: 0.37–0.97). The exacteason for these conflicting results obtained from theame ethnic population is unknown. However, the smallample size in these studies (n � 105 and 90, respec-ively) might cause the problem. In addition, the geno-ype frequencies among control subjects reported in thesestudies appear to be significantly different (P � .05). In00 patients and 1000 controls of a Han Chinese popu-ation, we detected a significantly increased GCA riskhat was associated with the XRCC1 399Gln polymor-hism. Our result is in agreement with that reported byhen et al27 but not that reported by Ratnasinghe et al.28

he XRCC1 Arg399Gln polymorphism has been exten-ively studied in many cancer sites, but the results areonflicting.23

Immunoprecipitation analysis in our present studyevealed that attenuated interaction between XRCC1nd ADPRT was caused by the 762Ala polymorphism inDPRT. A previous report suggested that the N-termi-al zinc finger domain (amino acids 1–202) and theentral automodification domain (amino acids 372–523)n ADPRT mediate the interaction with XRCC1,hereas the ADPRT deletion mutant (amino acids 524–014) did not associate with XRCC1.4 Although the62Ala polymorphism in ADPRT is not located withinhe suggested XRCC1-binding domain and may nothysically associate with XRCC1, the polymorphismay attenuate the interaction between ADPRT andRCC1 by reducing auto-poly(ADP-ribosyl)ation ofDPRT, which takes place in the central automodifica-

ion domain.3,4 It has been shown that XRCC1 interactsreferentially with autoribosylated ADPRT, distin-uished by a degenerate consensus motif for interactionith poly(ADP-ribose) located in the BRCT-1 do-ain.3,4 Regarding the XRCC1 Arg399Gln polymor-

hism, although it is located in the BRCT-1 regionequired for interaction with ADPRT, this amino acidhange had little effect on the ADPRT-XRCC1 interac-ion (Figure 1). This observation is inconsistent witholecular epidemiologic data, suggesting that otherechanisms underlying the association between the

RCC1 399Gln polymorphism and increased risk of

Gntp

Vgapttpio

1

1

1

1

1

1

1

1

1

1

2

2

2

2

2

2

2

2

2

EMFIB

22t

August 2006 POLYMORPHISMS AND RISK OF GASTRIC CANCER 427

CA may exist. Because ADPRT and XRCC1 are part-er repair proteins and act in the same causal pathway,he functional polymorphisms in these 2 genes are ex-ected to have a multiplicative joint effect.21

In summary, our study provides evidence that ADPRTal762Ala and XRCC1 Arg399Gln polymorphisms areenetic susceptibility factors for the development of GCA inChinese population. The ADPRT 762Val¡Ala polymor-hism results in significantly attenuated interaction be-ween ADPRT and XRCC1 proteins in vitro in cells, andhis functional alteration of ADPRT because of geneticolymorphism might disrupt cellular BER capacity, whichn turn confers the carriers susceptible to the developmentf GCA.

References1. Hoeijmakers JHJ. Genome maintenance mechanisms for prevent-

ing cancer. Nature 2001;411:366–374.2. Wood RD, Mitchell M, Sgouros J, Lindahl T. Human DNA repair

genes. Science 2001;291:1284–1289.3. Caldecott KW. XRCC1 and DNA strand break repair. DNA Repair

2003;2:955–969.4. Masson M, Niedergang C, Schreiber V, Muller S, Menissier-de

Murcia J, de Murcia G. XRCC1 is specifically associated withpoly(ADP-ribose) polymerase and negatively regulates its activityfollowing DNA damage. Mol Cell Biol 1998;18:3563–3571.

5. Lockett KL, Hall MC, Xu J, Zheng LS, Berwick M, Chuang S, ClarkPE, Cramer SD, Lohman K, Hu JJ. The ADPRT V762A geneticvariant contributes to prostate cancer susceptibility and deficientenzyme function. Cancer Res 2004;64:6344–6348.

6. Cottet F, Blanche H, Verasdonck P, Le Gall I, Schachter F, BurkelA, Murias ML. New polymorphisms in the human poly(ADP-ribose)polymerase-1 coding sequence: lack of association with longevityor with increased cellular poly(ADP-ribosyl)ation capacity. J MolMed 2000;78:431–440.

7. Shen MR, Jones IM, Mohrenweiser H. Nonconservative aminoacid substitution variants exist at polymorphic frequency in DNArepair genes in healthy humans. Cancer Res 1998;58:604–608.

8. Hao B, Wang H, Zhou K, Li Y, Chen X, Zhou G, Zhu Y, Miao X, WenT, Wei Q, Lin D, He F. Identification of genetic variants in baseexcision repair pathway and their associations with risk of esopha-geal squamous cell carcinoma. Cancer Res 2004;64:4378–4384.

9. Lunn RM, Langlois RG, Hsieh LL, Thompson CL, Bell DA. XRCC1polymorphisms: effects on aflatoxin B1-DNA adducts and glyco-phorin A variant frequency. Cancer Res 1999;59:2557–2561.

0. Abdel-Rahman SZ, El-Zein RA. The 399Gln polymorphism in theDNA repair gene XRCC1 influences the genotoxic response in-duced in human lymphocytes by the tobacco-specific nitrosamineNNK. Cancer Lett 2000;159:63–71.

1. Qu T, Morii E, Oboki K, Lu Y, Morimoto K. Micronuclei in EM9 cellsexpressing polymorphic forms of human XRCC1. Cancer Lett2005;221:91–95.

2. Zhang X, Miao X, Liang G, Hao B, Wang Y, Tan W, Li Y, Guo Y, HeF, Wei Q, Lin D. Polymorphisms in DNA base excision repairgenes ADPRT and XRCC1 and risk of lung cancer. Cancer Res2005;65:722–726.

3. Li JY, Ershow AG, Chen ZJ, Wacholder S, Li GY, Guo W, Li B, BlotWJ. A case-control study of cancer of the esophagus and gastriccardia in Linxian. Int J Cancer 1989;43:755–761.

4. Tran GD, Sun XD, Abnet CC, Fan JH, Dawsey SM, Dong ZW, Mark

SD, Qiao YL, Taylor PR. Prospective study of risk factors for

esophageal and gastric cancers in the Linxian general populationtrial cohort in China. Int J Cancer 2005;113:456–463.

5. Engel LS, Chow WH, Vaughan TL, Gammon MD, Risch HA, Stan-ford JL, Schoenberg JB, Mayne ST, Dubrow R, Rotterdam H, WestAB, Blaser M, Blot WJ, Gail MH, Fraumeni JF Jr. Populationattributable risks of esophageal and gastric cancers. J Natl Can-cer Inst 2003;95:1404–1413.

6. Ye W, Chow WH, Lagergren J, Yin L, Nyren O. Risk of adenocar-cinomas of the esophagus and gastric cardia in patients withgastroesophageal reflux diseases and after antireflux surgery.Gastroenterology 2001;121:1506–1508.

7. Mayne ST, Risch HA, Dubrow R, Chow WH, Gammon MD,Vaughan TL, Farrow DC, Schoenberg JB, Stanford JL, Ahsan H,West AB, Rotterdam H, Blot WJ, Fraumeni JF Jr. Nutrient intakeand risk of subtypes of esophageal and gastric cancer. CancerEpidemiol Biomarkers Prev 2001;10:1055–1062.

8. Miao X, Yu C, Tan W, Xiong P, Liang G, Lu W, Lin D. A functionalpolymorphism in the matrix metalloproteinase-2 gene promoter(1306C/T) is associated with risk of development but notmetastasis of gastric cardia adenocarcinoma. Cancer Res 2003;63:3987–3990.

9. Miao X, Xing D, Tan W, Qi J, Lu W, Lin D. Susceptibility to gastriccardia adenocarcinoma and genetic polymorphisms in methyl-enetetrahydrofolate reductase in an at-risk Chinese population.Cancer Epidemiol Biomarkers Prev 2002;11:1454–1458.

0. Zhang L, Xing G, Tie Y, Tang Y, Tian C, Li L, Sun L, Wei H, Zhu Y, HeF. Role for the pleckstrin homology domain-containing protein CKIP-1in AP-1 regulation and apoptosis. EMBO J 2005;24:766–778.

1. Brennan P. Gene-environment interaction and etiology of cancer:what does it mean and how we measure it? Carcinogenesis2002;23:381–387.

2. Dixon K, Kopras E. Genetic alterations and DNA repair in humancarcinogenesis. Semin Cancer Biol 2004;14:441–448.

3. Goode EL, Ulrich CM, Potter JD. Polymorphisms in DNA repairgenes and associations with cancer risk. Cancer Epidemiol Bi-omarkers Prev 2002;11:1513–1530.

4. Berwick M, Vineis P. Markers of DNA repair and susceptibility tocancer in humans: an epidemiologic review. J Natl Cancer Inst2000;92:1536–1537.

5. Frank SA. Genetic predisposition to cancer—insights from popu-lation genetics. Nat Rev Genet 2004;5:764–772.

6. Leppard JB, Dong Z, Mackey ZB, Tomkinson AE. Physical andfunctional interaction between DNA ligase III� and poly(ADP-Ribose) polymerase 1 in DNA single-strand break repair. Mol CellBiol 2003;23:5919–5927.

7. Shen H, Xu Y, Qian Y, Yu R, Qin Y, Zhou L, Wang X, Spitz MR, WeiQ. Polymorphisms of the DNA repair gene XRCC1 and risk ofgastric cancer in a Chinese population. Int J Cancer 2000;88:601–606.

8. Ratnasinghe LD, Abnet C, Qiao YL, Modali R, Stolzenberg-So-lomon R, Dong ZW, Dawsey SM, Mark SD, Taylor PR. Polymor-phisms of XRCC1 and risk of esophageal and gastric cardiacancer. Cancer Lett 2004;216:157–164.

Received October 7, 2005. Accepted May 11, 2006.Address requests for reprints to: Dongxin Lin, MD, Department of

tiology and Carcinogenesis, Cancer Institute, Chinese Academy ofedical Sciences, Beijing 100021, China. ; fax: (86) 10-67722460 or

uchu He, PhD, Department of Genomics and Proteomics, Beijingnstitute of Radiation Medicine, Beijing Proteome Research Center,eijing, China. e-mail: dlin@public.bta.net.cnSupported by State Key Basic Research Program grant

004CB518701; National “863” High Technology Project grants002BA711A06, 2002BA71A02-5, and 2004AA221100; and Na-ional Natural Science Foundation grant 30321003.

X.M., X.Z., and L.Z. contributed equally to this work.