The etiology of esophageal squamous cell carcinoma (ESCC) has been shown to be associated with genetic and certain environmental factors that produce DNA damage. Base excision repair (BER) genes are responsible for repair of DNA damage caused by reactive oxygen species and other electrophiles and therefore are good candidate susceptibility genes for ESCC. We first screened eight BER genes for new and potential functional polymorphisms by resequencing 27 DNA samples. We then identified and genotyped for important tagging single nucleotide polymorphisms (SNPs) in a case-control study of 419 patients with newly diagnosed esophageal cancer and 480 healthy controls by frequency matching on age and sex. The association between genotypes and ESCC risk was estimated by unconditional multivariate logistic regression analysis, and stepwise regression procedure was used for constructing the final logistic regression model. We identified 129 SNPs in the eight BER genes, including 18 SNPs that cause amino acid changes. In the final model, 4 SNPs, including 2 in the coding regions (ADPRT Val762Ala and MBD4 Glu346Lys) and others in noncoding regions (LIG3 A3704G and XRCC1 T-77C), remained as significant predictors for the risk of ESCC. The adjusted odd ratios were 1.25 [95% confidence interval (CI) 1.02–1.53] for the ADPRT 762Ala allele, 1.25 (95% CI 1.02–1.53) for the MBD4 346 Lys allele, 0.78 (95% CI 0.63–0.97) for the LIG3 3704G allele, and 1.38 (95% CI 1.01–1.89) for the XRCC1–77C allele. In addition, we observed a significant gene-gene interaction between XRCC1 Gln399Arg and ADPRT Val762Ala. The results suggest that the polymorphisms in five BER genes may be associated with the susceptibility to ESCC in a Chinese population.
Cancer of the esophagus ranks among 10 most frequent cancers in the world, with a marked regional variation in incidence and mortality worldwide (1) . Northern China is one of the regions where esophageal squamous cell carcinoma (ESCC) is common, and there are ∼250,000 patients newly diagnosed with ESCC each year in China, accounting for more than half of the world’s cases (2) . Although the integrated etiology of ESCC remains to be fully elucidated, accumulative epidemiological evidence suggests that tobacco smoking, heavy alcohol drinking, micronutrient deficiency, and dietary carcinogen exposure may cause the disease (3, 4, 5, 6) . All these factors can induce or enhance DNA damage mediated by either oxidative stress or DNA-binding electrophiles, which in turn may initiate and/or promote carcinogenesis. However, to safeguard the integrity of genome and prevent the detrimental consequences of DNA damage, humans have developed a complex set of DNA repair systems. Defects in DNA repair have been demonstrated to be a critical mechanism in human carcinogenesis (7) . In addition, accumulating evidence suggests that suboptimal DNA repair capacity caused by genetic polymorphism is associated with increased cancer risk (8, 9, 10, 11) . Thus, ESCC is probably associated with not only environmental factors but also individuals’ suboptimal DNA repair capacity.
Base excision repair (BER) is one of the important DNA repair pathways against DNA damage resulted from many insults, including altered metabolism, reactive oxygen species, and methylating and deaminating agents (7 , 12, 13, 14) . BER has two major two steps: excision of damaged base residues and core BER reaction (7 , 14) . Briefly, a battery of glycosylases, each dealing with a relatively narrow, partially overlapping spectrum of lesions, feeds into a core reaction by releasing the modified base and creating abasic sites; the core BER reaction is initiated by strand incision at the abasic site by the APEX endonuclease; and DNA polymerase β performs a one-nucleotide gap-filling reaction and removes the 5′-terminal baseless sugar residue via its lyase activity. This is then followed by sealing of the remaining nick by the XRCC1-ligase3 complex (14) . It is the so-called short-patch repair that performs a one-nucleotide gap-filling reaction, which is the dominant mode of the repair activity in mammals. The gap of 2–10 bases was filled by the long-patch repair mode, which involves DNA polβ, polδ/ε, and proliferating cell nuclear antigen in repair synthesis, FEN1 endonuclease in removal of the displaced DNA flap, and DNA ligase 1 in gap sealing. Enzymes from the BER pathway can also rectify single-strand interruptions in DNA. ADP-ribosyltransferase (ADPRT), as well as XRCC1, temporarily binds to single-strand interruptions in DNA and may act to recruit repair proteins (7 , 14) . Knockout mouse models showed inactivation of BER core proteins (i.e., APEX1, POLB, and XRCC1) induces embryonic lethality, highlighting the pivotal roles of these genes (7) . Recently, a knockout mouse model of MBD4, a glycosylase, revealed that although MBD4 inactivation does not by itself cause cancer predisposition in mice, it may alter the mutation spectrum in cancer cells and therefore contribute to the cancer-predisposition phenotype (15 , 16) . Until last year, inherited deficiencies in the BER pathway had not been causally linked to any human genetic disorders (7) . It has recently been found that the bi-allelic mutations in a DNA glycosylase gene, MUTYH, may lead to an autosomal recessive syndrome of adenomatous colorectal polyposis with a high colorectal cancer risk (17, 18, 19, 20) . Thus, common polymorphisms of BER genes are plausible candidates that may contribute to susceptibility to cancer, including ESCC according to the Common Disease-Common Variant hypothesis (21) .
Several studies have in fact found associations between genetic polymorphisms in some BER genes such as OGG1 and XRCC1 and risk of certain cancers, including ESCC, and the results are encouraging (Refs. 10 , 11 ; reviewed in Ref. 22 ). However, most of the previous studies were designed to analyze individual gene, which obviously has limitations to elucidate the effect of the entire BER pathway. To comprehensively investigate the roles of the polymorphisms in the BER genes in the development of ESCC and identify potential genetic markers for ESCC in Chinese populations, we conducted this study by using a candidate-gene association approach. We first screened genomic regions of eight selected BER genes for potential functional single nucleotide polymorphisms (SNPs) and other genetic markers in genomic DNA samples from 27 Chinese individuals. We selected the identified SNPs based on their functional potentials and pairwise linkage disequilibrium and then genotyped the selected SNPs in 419 esophageal cancer patients and 480 control subjects. Finally, we estimated the frequencies of genotypes and haplotypes of the selected SNPs studied in this Chinese population and evaluated their associations with risk of ESCC.
MATERIALS AND METHODS
We first obtained the candidate genes in the BER pathway (13) and their genomic sequences from the published database of the National Center for Biotechnology Information. We used annotations in the database for all known exons, intron-exon boundaries, untranslated region, and 5′-flank regions of eight selected genes, i.e., MUTYH, MBD4, OGG1, ADPRT, APEX1, LIG3, POLB, and XRCC1. We designed the primers for the target regions in each gene with the web-based software Primer 3.0. 5 We amplified and purified DNA samples from 27 apparently normal Chinese individuals. These samples included 54 chromosomes, providing at least a 95% confident level to detect alleles with frequencies > 5%. We then sequenced the PCR products by using ABI PRISM Dye Terminator Sequencing kits (Applied Biosystems, Foster City, CA) and loading the samples onto an ABI 3700 sequencer. Finally, we used the PolyPhred program to identify SNP candidates that were then confirmed by two independent observers. We further confirmed these SNP positions and individual genotypes by using reamplifying and resequencing the SNP site from the opposite strand.
To assess frequencies and linkage disequilibrium of the identified SNPs, we genotyped for 37 selected SNPs by resequencing of another 96 samples from the apparently normal controls and did not observe significant differences in the allele frequencies between the 27 sample-set and the 96 sample-set, except for 1 SNP in APEX1 and 3 in XRCC1. However, the linkage disequilibrium of the pairwise SNPs was stronger in the 96 sample-set than in the 27 sample-set.
The case-control study consisted of 419 patients with ESCC and 480 population controls. All subjects were unrelated ethnic Han Chinese and residents in Beijing and the surrounding regions. Case patients were recruited between July 1999 and December 2001 at the Cancer Hospital, Chinese Academy of Medical Sciences (Beijing, China). All patients with histologically confirmed ESCC were enrolled with a response rate of 94%, and there was no sex and age restriction. The exclusion criteria included previous cancer and previous chemotherapy or radiotherapy. Most case patients were participants in a molecular epidemiological study of esophageal cancer described previously (10) . In the current study, we extended the sample size of ESCC patients to 419. Control subjects were cancer-free individuals and randomly selected from a nutritional survey database conducted in the same regions during the same time period of case collection, with a response rate of 89%. The characteristics of control subjects were described previously (10) . The selection criteria for control subjects included no individual history of cancer and frequency matching to case patients on sex and age (±5 years). In the present study, we genotyped 480 control subjects because some DNA samples used in the previous study were run out. At recruitment, informed consent was obtained from each subject, and personal data from each participant regarding demographic characteristics such as sex and age and related risk factors, including tobacco smoking, were collected by questionnaire. This study was approved by the Institutional Review Board of the Chinese Academy of Medical Sciences Cancer Institute.
Genomic DNA was extracted from blood samples of all control subjects and most of case patients. Twenty eight percent of DNA samples from case patients were isolated from surgically resected normal tissues adjacent to the tumor of the esophagus. Genotypes were analyzed using PCR-based RFLP or tetra-primer amplification refractory mutation system (23) . The SNPs, which have the sequences recognized by restriction endonuclease, were determined by using PCR-RFLP or determined otherwise by using tetra-primer amplification refractory mutation system-PCR. The primers for amplification refractory mutation system were designed by using a web-based program. 6 All PCRs were performed in plate thermal cyclers (GeneAmp PCR System 9700; PE Applied Biosystems). The amplification for RFLP genotyping was performed with a 15-μl reaction mixture containing 10 ng of DNA, 0.1 μm of each primer, 0.2 mm deoxynucleoside triphosphates, 1.5 mm MgCl2, and 0.5 units of DNA polymerase with 1× reaction buffer (Takara Biotech Co. Ltd., Dalian City, China), and the amplification for amplification refractory mutation system was accomplished with a 10-μl reaction mixture containing 10 ng of DNA, 0.1 μm of each primer, 0.2 mm deoxynucleoside triphosphates, 1.5 mm MgCl2, 0.4 units of HotStar Taq with 1× buffer and 1× Q-solution (Qiagen, Chatsworth, CA). The primers and the restriction enzymes used in the present study are listed in the supplement.
To ensure that the observed polymorphisms were specific and not the results of experimental variation, the results were confirmed by repeating 10% of the samples and by directly sequencing 10% of the specimens. All results were 100% concordance.
On the basis of the observed frequencies of the selected SNPs, we estimated their haplotypes by using the Bayesian statistical method (24) . We used the LDA software to calculate linkage disequilibrium index (D′ and r2; Ref. 25 ). For each polymorphism, deviation of the genotype frequencies in the control subjects from those expected under Hardy-Weinberg equilibrium was assessed by using the standard χ2 test. We compared the distributions of genotypes and haplotypes between case patients and control subjects by using the Monte Carlo approach in which the simulations were performed 100,000 times (26) . We estimated the cancer risk associated with the alleles, genotypes, and haplotypes as odds ratios (ORs) and 95% confidence intervals (CIs) by using unconditional logistic regression with adjustment for age, sex, and smoking status. We also fitted the multi-SNP-adjusted logistic model with both forward and backward stepwise regression methods. The joint effect of genes participating in BER was evaluated by including the number of high-risk genotypes in the final logistic regression model. The haplotype risks were estimated as ORs and 95% CIs by haplotype-based logistic regression (27) . All statistical tests were two-sided, and the probability level of <0.05 was used as the criterion of significance. The Monte Carlo simulations were performed by using Curtis’ program CLUMP and the logistic regression by using STATA 7.0.
We chose eight genes involved in the BER pathway: three coding glycosylases (MUTYH, MBD4, and OGG1) and five coding core proteins (ADPRT, APEX1, LIG3, POLB, and XRCC1; Table 1 ⇓ ). Overall, we had explored the total lengths spanned 15 kb in the coding regions and 42 kb in the noncoding regions. The total number of SNPs identified in these eight genes was 129 (1 polymorphism/442 bp), which included 18 SNPs that cause amino acid substitutions of the proteins, and each gene had at least one coding SNP. The average number of polymorphisms/kb of DNA was 1.7 in the coding regions and 2.5 in the noncoding. All genes, except for LIG3 and POLB, had nonsynonymous SNPs with minor allele frequencies > 3% but <50% identified in 27 individuals and were used for additional association studies.
Association with Individual SNP.
We genotyped for 21 selected SNPs in the eight BER genes in all 419 ESCC case patients and 480 control subjects. None of the genotype distributions in the control subjects differed significantly from those expected from Hardy-Weinberg equilibrium (data not shown).
Six SNPs in OGG1, APEX1, and POLB showed no significant difference between ESCC patients and control subjects (Tables 2 ⇓ and 3 ⇓ ). All adjusted ORs were not different from 1.0, with upper 95% confidence limits of ≤1.57 and lower limits of ≥0.63. However, the genotype distribution of the coding SNP (Glu346Lys) in MBD4 differed between ESCC case patients and control subjects, and subjects with the Lys/Lys genotype had a significantly increased risk of developing ESCC compared with those with the Glu/Glu genotype (OR 1.64, 95% CI 1.08–2.48; Table 2 ⇓ ). The Lys allele may have a recessive effect because the heterozygotes had a slightly but not significantly increased risk.
The genotype frequencies of both ADPRT G44108C and Val762Ala polymorphisms were significantly different between case patients and control subjects. The ADPRT G44108C polymorphism was found to be in strong linkage disequilibrium with the nonsynonymous Val762Ala polymorphism. The alleles of both ADPRT 762Ala and 44108C were associated with an increased risk of ESCC compared with the respective major allele (Table 3) ⇓ . In contrast, the GG genotype of LIG3 A3704G polymorphism had a significantly protective effect, with the OR being 0.53 (95% CI 0.31–0.92; Table 3 ⇓ ) compared with the AA or AG genotype. The XRCC1-77T allele was also associated with a borderline increased risk of ESCC (Table 3) ⇓ .
Association with Multiple SNPs.
In multiple SNPs’ association analysis, we firstly eliminated the statistically nonsignificant SNPs (P > 0.10) in the individual SNP association analysis. Thus, only 5 SNPs (ADPRT Val762Ala, LIG3 G3704A, MBD4 Glu346Lys, MUTYH G7369A, and XRCC1 T-77C) were selected by additional stepwise logistic regression procedures using P = 0.10 in which the three nongenetic covariates (age, sex, and smoker) were remained. Although the ADPRT G44108C polymorphism was significantly associated with risk of ESCC in the individual SNP analysis, it was eliminated because it was in strong linkage disequilibrium with the ADPRT Val762Ala, which is likely to be a functional polymorphism. The results of forward and backward stepwise logistic regression procedures were consistent: 4 SNPs (ADPRT Val762Ala, LIG3 G3704A, MBD4 Glu346Lys, and XRCC1 T-77C) retained in the final logistic model (Table 4) ⇓ . Environmental factor smoking was also the main risk factor in this model, and in the presence of smoking in the same model, genetic factors appeared to have a relatively moderate effect on risk of ESCC. Compared with those subjects with no high-risk genotypes (i.e., the none group as the reference), the ORs for one and two high-risk genotypes were not significant elevated, but the risks of ESCC associated with three and four high-risk genotypes (∼12% among the controls) were significantly increased (Table 5) ⇓ . We noticed that the smokers with the ADPRT 762 Ala/Ala and XRCC1-77 C/C genotypes had higher risks of ESCC than nonsmokers, and the protective effects of LIG3 3704A/A were greater in nonsmokers (Table 6) ⇓ .
Association with Haplotypes.
For the eight genes in this study, seven genes (but not POLB) had >1 SNP genotyped in ESCC cases and control subjects. We estimated haplotype frequencies in the case patients and control subjects and haplotype-specific ORs by the haplotype-based logistic regression method. For statistic advantage, we combined all haplotypes that had an allele frequency of <0.05. Table 7 ⇓ summarizes the association between the haplotypes and risk of ESCC. We observed significant differences between case patients and control subjects in the distributions of MUTYH and LIG3 haplotypes. The GAG haplotype of MUTYH was associated with a substantially reduced risk of ESCC (OR 0.60, 95% CI 0.45–0.79) compared with the TGA haplotype, whereas the low-frequency haplotype was associated with an increased risk (OR 2.01, 95% CI 1.31–3.08). The GG haplotype of LIG3 appeared to be associated with a significantly decreased risk of ESCC as we had shown in the individual SNP association analysis. We also observed an increased risk of ESCC associated with the CA haplotype of MBD4 (OR 1.23, 95% CI 1.01–1.50) compared with the most common CG haplotype. In addition, a reduced risk of ESCC was associated with the GTCA haplotype of ADPRT (OR 0.70, 95% CI 0.55–0.90) compared with the GCGA haplotype that carried both 40336C and 44108G alleles, which had been shown to be associated with increased risk in the individual SNP association analysis.
Analysis of Gene-Environment and Gene-Gene Interactions.
We did not observe any significant interaction between smoking and SNPs in all SNPs in the selected genes, except for 3 SNPs in OGG1. The interaction between smoking and the heterozygote of OGG1 T-1911C was significant (P = 0.011), and the interaction between smoking and the OGG1 326CC or 9295GG genotype was borderline significant (P = 0.088 and 0.077, respectively; data not shown). The gene-gene interaction analysis was performed on the pairwise coding SNPs in XRCC1-ADPRT and XRCC1-APEX1 because there are physical and functional interactions between these proteins (7) . A significant interaction between XRCC1 Gln399Arg and ADPRT Val762Ala polymorphisms was observed, and the combined homozygous genotypes (i.e., the minor alleles of both genes) exhibited the highest risk (OR 7.91, 95% CI 1.63–38.39) compared with the combined major alleles (Table 8) ⇓ .
In the present study, we investigated the association between risk for development of ESCC and genetic polymorphisms in eight BER genes by using a more practical candidate-gene approach (28) . First, we identified potentially functional polymorphisms and genetic markers in the candidate genes by resequencing the genes from a subset of 27 subjects. The SNPs that cause amino acid substitutions were then examined primarily in the case-control study because they have the potential to be of functional significance. We also tested some noncoding SNPs of suitable frequency because any potentially functional polymorphisms, even if not directly assayed, are likely to be in linkage disequilibrium (LD) with these noncoding SNPs. Although we did not type other 30 common SNPs (with a minor allele frequency >10% but <50%) that we found in our population, most of these variants are in strong LD with the SNPs we typed so they can be evaluated indirectly. The power to detect an association by LD is dependent on the closeness in frequency between the functional polymorphisms and the observed marker allele or haplotype in LD with it. Generating haplotypes with multiple polymorphisms also can reduces the possibility that a true association will be missed.
The average number of SNPs/kb of DNA identified in our samples is slightly less than that obtained from published studies of other large-scale SNPs screening, which may be because of different size and ethnicity of populations included (29) . Apparently, the amino acid substitutions and other common SNPs (minor allele frequency > 10%) in ADPRT, APEX1, LIG3, POLB, and XRCC1 reported in other populations were also identified with similar allele frequencies in this Chinese population (30) . However, it is not the case in the rare variants alleles because most of rare SNPs (minor allele frequency < 5%) we identified are unique in this Chinese population. Recently, two multigenic studies on variants in DNA double-strand break repair genes and breast cancer susceptibility identified three (XRCC2, XRCC3, and LIG4) of seven selected double-strand break repair genes and two (Ku70 and XRCC4) of five selected double-strand break repair genes, respectively, that are associated with increased risk of breast cancer (31 , 32) . In the present study, we identified five of eight selected BER genes associated with risk of ESCC, and a more pronounced risk was found in those with several high-risk genotypes. Taken together, these findings suggest that it is necessary to study SNPs of multiple genes in this kind of association studies because 1 SNP may not play a major role in the etiology of cancer. This study did not include some genes involved in the long patch subpathway, which may affect the risks of ESCC. Their role in the etiology of ESCC warrants future studies.
We found that some SNPs in two glycosylase genes (MBD4 and MUTYH) and three core genes (ADPRT, LIG3, and XRCC1) were associated with risk for the development of ESCC. To the best of our knowledge, this is the first study that has investigated the five BER genes (MBD4, MUTYH, ADPRT, LIG3, and POLB) simultaneously in cancer susceptibility in a Chinese population. The final logistic regression model identified four important polymorphisms: ADPRT Val762Ala and MBD4 Glu346Lys, as putative functional SNPs, and LIG3 intron A3704G and XRCC1 5′-UTR T-77C, as genetic markers. Our results on the joint effects of high-risk genotypes suggest that individuals with three and more high-risk genotypes had a high risk of ESCC compared with those one or two high-risk genotypes. The ESCC risk associated with smoking in the final model is also consistent with results of previous epidemiological studies in Chinese and other ethnic populations (3 , 33) . The stratification analysis by smoking status suggested that three high-risk genotypes in core BER reaction had a great effect on the risk in smokers.
Our findings are biologically plausible. ADPRT is an abundant nuclear protein that functions as a molecular nick sensor and is important for genetic stability and for cellular resistance to ionizing radiation and alkylating agents (34 , 35) . It also appears to strongly influence chromatin architecture, gene expression, and induction of cell death (32) . Transient binding of ADPRT to DNA breaks may serve a signaling role to initiate DNA repair reactions by recruiting other proteins (36) . Previous studies showed that the activity of ADPRT in mononuclear leukocytes was significantly different between cancer patients and health controls, suggesting that its activity might affect the susceptibility to cancers (37) . In the present study, we found that an increased risk of ESCC was associated with the ADPRT nonsynonymous SNP (Val762Ala), which is located in the sixth helix of catalytic regulatory domain. ADPRT is highly conserved, especially at amino acids comprising structure motifs and functional domains (34) . The 762Val is conserved in homologous genes of the mouse and rat (34) . Although the functional relevance of the amino acid substitution (Val762Ala) caused by T40336C mutation is not known yet, its association with increased risk of ESCC suggested that this polymorphism may diminish BER capacity in individuals with the 762Ala allele.
MBD4 G3226A is another SNP that causes amino acid substitution (Glu346Lys), associated with increased risk of ESCC, and retained in our final risk model. MBD4 functions as a mismatch-specific DNA glycosylase active on thymine, uracil, and 5-fluorouracil paired with guanine (38, 39, 40, 41) . The glycosylase activity of MBD4 prefers substrates containing a G:T mismatch within methylated or unmethylated CpG sites. It has been shown that MBD4 also plays an important role in genomic surveillance and apoptosis by interacting with the mismatch repair/tumor suppressor protein MLH1 and Fas-associated death domain protein (42) . On the other hand, MBD4 is frequently mutated in many types of human cancer such as colorectal, gastric, endometrial, and pancreatic cancers, and it is considered as a candidate tumor suppressor gene (43, 44, 45) . Furthermore, animal experiments showed that knockout mice (mbd4−/−) had increased frequency of spontaneous carcinomas compared with heterozygous (mbd4+/−) and wild (mbd4+/+) mice (15 , 16) . In the present study, we found a recessive effect of MBD4 Glu346Lys polymorphism on risk of ESCC, which is consistent with the findings in the mouse model (15 , 16) , and the common natures of tumor suppressor gene. Although this SNP causes a neutral substitution that does not lie in any of the known functional domains, two-charges change of this protein may affect its interaction with damaged DNA. Overall, the results from previous laboratory studies support our molecular epidemiological finding.
XRCC1 Gln399Arg has been extensively studied in many cancer sites, but the results are conflicting (reviewed in Ref. 22 ). We failed to find significant association between the Gln399Arg genotype alone and risk of ESCC in the individual SNP association analysis, which is consistent with our previous finding (10) . However, it is interesting to note that there was a significant interaction between the XRCC1 399Arg/Arg genotype and the ADRPT 762Ala/Ala genotype on risk of ESCC; subjects having both these homozygous genotypes had a ∼8-fold increased risk compared with those having the other genotypes. The Gln399Arg is located in BRCT-1 region of the XRCC1, which harbors the ADPRT binding domain (33 , 46) . A recent study (47) showed that the rapid activation of ADPRT is required to assemble and stabilize nuclear foci at sites of oxidative DNA damage by recruiting the molecular scaffold protein XRCC1. Because of functional interaction between XRCC1 and ADPRT, one may expect that the polymorphisms in these two genes would have a synergic effect. Taken together, these data strongly support our findings in the current study showing an association between the polymorphisms in XRCC1/ADPRT and risk of ESCC.
Although there was no single MUTYH polymorphism retained in final statistic model, we did observe a significant association between the MUTYH haplotype and reduced risk of ESCC, with the GAG haplotype having reduced risk by half. These results are also biologically plausible because it has recently been discovered that bi-allelic mutations in MUTYH lead to an autosomal recessive syndrome of adenomatous colorectal polyposis and high susceptibility to colorectal cancer (17, 18, 19, 20) . However, we also observed a 2-fold increased risk of ESCC associated with the minor haplotypes of MUTYH. The reason for this reverse direction of the association between different haplotypes in the same gene is not immediately evident. One suggestive explanation is that this effect is secondary to LD with a yet unidentified but tightly linked ESCC locus. The LIG3 intron A3704G and XRCC1 5′UTR T-77C identified in the final risk model was also likely caused by linkage disequilibrium with other functional SNPs close to the loci yet to be identified. Additional studies, especially functional analysis of the polymorphisms, would be required to address these hypotheses.
In conclusion, this current study provides a substantial evidence for the first time that genetic polymorphisms in five BER genes were associated with risk of developing ESCC in a Chinese population. These polymorphisms influence susceptibility to ESCC in different genetic models of allele-dose effect, recessive effect, and gene-gene or gene-smoking interaction. Our study is among few that have examined the association between SNPs in multiple genes from a specific DNA repair pathway and the complex disease. Because BER is a protective mechanism against DNA damage caused by endogenous and exogenous DNA-damaging agents, these polymorphisms may also act as genetic susceptibility factors for other types of human cancers.
We thank Professors Yan Shen and Zhijian Yao in Chinese National Human Genome Center at Beijing for supports on sequencing and Dr. Keyue Ding for constructive discussion on LD analysis.
Grant support: National “863” High Technology Project Grants 2001AA224011 (F. He) and 2002BA711A06 (D. Lin) and National Natural Science Foundation Grant 39990570 (D. Lin).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Requests for reprints: Fuchu He, Laboratory of Systems Biology, Beijing Institute of Radiation Medicine, Beijing 100850, China. Phone: 86-10-681-71208; Fax: 86-10-682-14653; E-mail:
↵5 Internet address: http://www-genome.wi.mit.edu/cgi-bin/primer/primer3_www.cgi.
↵6 Internet address: http://cedar.genetics.soton.ac.uk/public_html/primer1.html.
- Received February 4, 2004.
- Revision received March 22, 2004.
- Accepted March 30, 2004.
- ©2004 American Association for Cancer Research.