DNA Repair Gene Polymorphisms and Risk of Pancreatic Cancer

Pancreatic cancer is the fourth leading cause of cancer death for both men and women in this country (1). It is a highly aggressive malignancy that shows extensive genomic instability and aneuploidy (2). Identification of genetic factors, environmental exposures, and gene environment interactions that contribute to pancreatic cancer development is crucial for the primary prevention of this disease.

About 10% of patients with pancreatic cancer have a family history of this disease (3). Some of the familial aggregation can be explained by relation to inherited cancer syndromes, such as familial atypical multiple-mole melanoma, Peutz-Jeghers syndrome, hereditary breast-ovarian cancer, hereditary pancreatitis, familial adenomatous polyposis, and hereditary nonpolyposis colorectal cancer (4, 5). Some of the genes that are responsible for these cancer syndromes are involved in DNA repair and cellular response to DNA damage. However, whether these genes contribute to the development of sporadic pancreatic cancer remains unknown. The known and suspected environmental risk factors for sporadic pancreatic cancer include cigarette smoking, obesity, diabetes, chronic pancreatitis, and dietary factors (6–8). The genetic susceptibility factors for sporadic pancreatic cancer have been investigated in a few case-control studies, most of which have examined common gene polymorphisms in carcinogen metabolism (9, 10) and DNA repair (11–15). For example, studies conducted by us and by other investigators on DNA repair have shown a significant main effect of XPF gene (11), an interaction of XRCC1 with either MGMT or APE1 gene (12), or interaction of XPD (13), XRCC1 (12, 14), or XRCC2 gene (15) with cigarette smoking on the risk of pancreatic cancer. These observations support a role of genetic variability in DNA repair in pancreatic cancer and request further investigations on many of the important but unexplored genes involved in various DNA repair pathways.

The base excision repair and DNA single strand break repair mechanisms are the major repair systems that are involved in the processing of oxidative DNA lesions (16, 17). DNA double-strand breaks are the most lethal DNA lesions caused by ionizing radiation or some chemical agents. DNA double-strand breaks can also occur as a consequence of DNA replication error. Because DNA double-strand breaks lead to chromosome breaks and genomic instability, many genes involved in the repair of double-strand breaks are well-known tumor suppressors (18, 19). We have previously shown that single nucleotide polymorphisms (SNP) of base excision repair and double-strand break repair genes, such as OGG1, XRCC1, APEX1, POLB, ATM, RAD54L, and RECQL, were significantly associated with the clinical outcome and overall survival of patients with pancreatic cancer who received chemoradiation (20–22). The relation of some of these genes, such as XRCC1, XRCC2, XRCC3, and APEX1, to the risk of pancreatic cancer has been investigated in our case-control study (12, 14). The current study further investigate nine remaining previously clinically investigated SNPs of the hOGG1, LIG3, LIG4, POLB, ATM, RAD54L, and RECQL genes in a large-scale, case-control study. The main effects of these genes and their potential interactions with smoking, alcohol, and diabetes on the risk of pancreatic cancer were analyzed by logistic regression.

Study population. The study design and data collection methods have been previously described in detail (8, 23). Briefly, in this hospital-based case-control study, 734 non-Hispanic White patients with pathologically confirmed primary pancreatic ductal adenocarcinoma and 780 healthy controls were consecutively enrolled at The University of Texas M. D. Anderson Cancer Center from the year 2000 through the year 2007. All study participants were U.S. residents and were able to communicate in English. Control subjects were recruited from healthy spouses, friends, and nonblood relatives of patients with various types of cancers other than pancreatic cancer, and members of the control group were frequency matched to cases by age at enrollment (±5 y), sex, and race. The response rate was 80.6% for cases and 76.9% for controls. Information on cigarette smoking, alcohol consumption, medical history, family history of cancer, and other risk factors was collected by personal interview, and a blood sample for genotyping was collected from each participant at the time of enrollment. Cumulative smoking was calculated in pack-years [pack-years = (packs per day) × (years smoked)]. Alcohol consumption was calculated in terms of milliliters of ethanol consumed daily, with 12.0 oz of beer, 4.0 oz of wine, and 1.5 oz of hard liquor each considered to be equivalent to ∼12.0 mL of ethanol. Family history of cancer among first-degree relatives was collected. Body mass index (BMI, kg/m²) was calculated based on the self-reported weight and height at age 34 to 39 in 455 cases and 466 controls because this information was collected only after January 2004 in our study. Written informed consent was obtained from each study participant for interview and a blood sample. The study was approved by the Institutional Review Board of M. D. Anderson Cancer Center.

DNA extraction and genotyping. Peripheral blood mononuclear cells were collected from freshly drawn blood by Ficoll-Hypaque (Amersham Pharmacia Biotech) density gradient centrifugation and stored at −80 °C. DNA was extracted with the use of a FlexiGene DNA kit (Qiagen) and a Maxwell16 automated system (Promega) and stored at 4°C for immediate use.

Genotyping was initially conducted using the Masscode technique by BioServe Biotechnologies, Ltd. and, later, the Taqman diallelic discrimination method in our laboratory. The reference numbers, gene locus, chromosome location, nucleotide change, amino acid change, and minor allele frequency of the SNPs examined in this study are described in Table 1. These SNPs were selected based on their previously reported associations with either risk of cancer or patient survival. All SNPs except RAD54L 154C>T and POLB-2133 T>C had a minor allele frequency of >10% among non-Hispanic Whites. Probes and oligonucleotides were obtained from Applied Biosystems using the Assay-by-Design product. The reactions were prepared by using 2× Taqman Universal Master Mix, 40× SNP Genotyping Assay Mix, DNase-free water, and 10 ng genomic DNA in a final volume of 5 μL per reaction. The PCR amplification was done using the ABI Prism 7900 HT sequence-detector. About 5% of the samples were analyzed in duplicate and 100% consistency was achieved.

Statistical analysis. All statistical analyses were done using SPSS 15.0 (SPSS Inc.) and STATA 9.0 (Stata Corp.) software programs. P values of <0.05 were indicative of statistical significance. The Pearson's χ² test was used to compare the distribution of categorical variables and genotype frequencies between cases and controls. Odds ratios (OR) and their associated 95% confidence intervals (95% CI) were estimated by using an unconditional logistic regression analysis with adjustment for smoking (never, ≤20 pack-years, or >20 pack-years), alcohol (never, ≤60 mL of ethanol/day, or >60 mL of ethanol/day), diabetes (yes or no), and family history of cancer among first-degree relatives (yes or no). Because diabetes could be a manifestation of pancreatic cancer, individuals with recent diabetes onset, i.e., within 2 y before the cancer diagnosis for cases and before recruitment to the study for controls were excluded from some of the analyses. Because information on BMI was missing from study participants recruited in the early stage of the investigation, BMI was not included in the multivariate model. For detection of possible interactions between genotypes and smoking, alcohol, diabetes, or BMI, individuals without the risk factor (e.g., never smoked) and the non-at-risk genotype were used as the reference group and adjusted ORs for never smokers with the at-risk genotype (OR₁₀), smokers with non-at-risk genotype (OR₀₁), and smokers with the at-risk genotype (OR₁₁) were estimated using unconditional logistic regression. An OR₁₁ greater than the sum of OR₁₀ + OR₀₁ or greater than the product of OR₁₀ X OR₀₁ indicates a more than additive or more than multiplicative effect, respectively. The cross-product term of genotype with smoking, alcohol, diabetes, and BMI was generated with the use of the logistic regression model correspondingly. The significance of the interaction term was tested by using a likelihood ratio test, with the full model containing the interaction term, the main effect of the genotype, and the exposure variable and reduced model lacking the interaction term. For any statistically significant association we observed, we estimated the false-positive report probability using the methods described by Wacholder et al. (24). The false-positive report probability value for noteworthiness was set as 0.2.

Characteristics of the study subjects. The distribution of demographics, risk factors, and genotypes between cases and controls are described in Table 2. The mean ± SD ages of cases and controls were 62.2 ± 9.6 and 62.0 ± 9.6 years, respectively (P = 0.673). The cases and controls were well matched by age and sex. As previously described (23, 25), history of diabetes, family history of cancer, heavy smoking, heavy alcohol drinking, and overweight were all associated with increased risk of pancreatic cancer in this study population.

Main effect of genotype. The 9 SNPs were successfully amplified in 95.8% to 99.1% of the patients and controls. All SNPs except one, OGG1 T2657C (χ² = 6.73, P < 0.05), followed the Hardy-Weinberg equilibrium in both cases and controls (data not shown). There were no statistically significant differences in genotype distributions between cases and controls for seven of the nine SNPs (Table 3). The LIG3 G-39A AA homozygous mutant was associated with a significantly reduced risk of pancreatic cancer (OR, 0.23; 95% CI, 0.06-0.82; P = 0.024) after adjusting for age, sex, smoking, alcohol, diabetes, and family history of cancer. On the other hand, ATM D1853N AA variant was significantly associated with an increased risk of pancreatic cancer (OR, 2.55; 95% CI, 1.08-6.00; P = 0.032), with adjustment for other confounders. The estimated false-positive report probability was 0.136 for LIG3 G-39A and 0.166 for ATM D1853N, given a prior probability of 25%. Both are below the threshold of 0.20 indicating noteworthiness.

Interaction of genotype with known risk factors. Next, we examined the potential interactions between genotype and known risk factor for pancreatic cancer in this study population, such as cigarette smoking, alcohol intake, diabetes, and BMI. Because of the low frequency of the homozygous variants, they were combined with the heterozygous group in this analysis. The exposure variables used were as follows: never smoker versus ever smoker; nondrinker and light drinker (<60 mL/day) versus heavy drinker (>60 mL/day); diabetes (no versus yes); and BMI (≤25 versus >25 kg/m²). As shown in Table 4, there were no significant interactions of genotypes with smoking, alcohol, and BMI. There was a borderline significant interaction of POLB T-2133C genotype with smoking (P_interaction = 0.051). A statistically significant interaction of ATM D1853N or LIG4 C54T genotype with diabetes was observed in the analysis excluding recent onset diabetes. For example, when the group of nondiabetics with the ATM D1853N GG genotype w compared with other groups, ORs (95% CI) of 0.96 (0.74-1.24), 1.32 (0.89-1.95), and 3.23 (1.47-7.12) were obtained for nondiabetics with the GA/AA, diabetics with the GG, and diabetics with the GA/AA genotypes, respectively (P_interaction = 0.032). The OR (95% CI) was 0.91 (0.71-1.17), 1.11 (0.73-1.69), and 2.44 (1.34-4.46) for nondiabetics carrying the LIG4 CT/TT genotype, diabetics with the CC genotype, and diabetics carrying the CT/TT genotype, respectively, compared with nondiabetics carrying the CC genotype (P_interaction = 0.02). A weak interaction of the OGG1 T2657C genotype with diabetes was observed (P _interaction = 0.049). However, this interaction became statistically nonsignificant (P _interaction = 0.076) after excluding individuals with recent onset diabetes. RECQL A159C and RAD54L C154T genotype also showed a nonsignificant interaction with diabetes when all study participants were included in the analysis.

In this large case-control study, we examined SNPs of a number of DNA repair genes in association with risk of pancreatic adenocarcinoma. We observed a significant main effect of the homozygous variants of the LIG3 G-39A and ATM D1853N genotypes, and a statistically significant interaction of the ATM D1853N and LIG4 C54T genotype with diabetes on the risk of pancreatic cancer. These findings support our hypothesis that genetic variations in DNA repair modify the risk of pancreatic cancer.

DNA joining enzymes play an essential role in the maintenance of genomic integrity and stability. Three mammalian genes encoding DNA ligases, LIG1, LIG3, and LIG4, have been identified and each has distinct functional significance. LIG3 participates in base excision repair and DNA singles strand break repair by forming a stable complex with XRCC1. The LIG3 haplotype has been associated with radiation sensitivity in a study of breast cancer patients (26). LIG3 SNPs have been associated with increased risk of lung cancer (27) and esophageal cancer (28), but no association was found in other studies of cancer of the lung (29, 30), bladder (31), colon (32), and breast (33). The current study is the first to show a significant protective effect of the LIG3 G-39A AA genotype on risk of pancreatic cancer. However, this SNP is located in the intron region of the gene and the functional significance of this genetic variation is unknown. The frequency of the homozygous variant is very low so the observed association could be by chance alone. Whether this SNP is in linkage with other functionally significant SNPs of the gene needs further investigation. LIG4 is essential for V(D)J recombination and DNA double-strand break repair through nonhomologous end joining. Polymorphisms of this gene have been related to significantly increased risk of developing glioma (34), multiple myeloma (35), and lymphoblastic leukemia (36), as well as to the survival of patients with breast cancer (37). The LIG4 C54T SNP that showed a significant interaction with diabetes in modifying the risk of pancreatic cancer in the current study is located at 3′ untranslated region. Although DNA sequences in this region do not translate into proteins, the 3′untranslated region may contain sequence motifs crucial for the regulation of transcription, mRNA stability, and cellular location of the mRNA or the binding of microRNA (38). It is conceivable that LIG4 gene variant may confer a deficient repair of DNA damage caused by diabetes-associated elevated level of oxidative stress, in turn increase the risk of pancreatic cancer.

ATM is an important cell cycle checkpoint kinase that plays a critical role in cellular response to DNA damage and in maintaining genome stability. Germline mutations in the ATM gene result in the rare genomic instability syndrome ataxia telangiectasia, which is characterized by elevated cancer risk. ataxia telangiectasia heterozygote or common SNPs of the ATM gene have been associated with increased risk of developing breast cancer (39–41). In the current study, we have shown a significant main effect of the ATM D1853N homozygous variant on increased risk of pancreatic cancer. The same SNP was previously associated with increased risk of second primary tumors among breast cancer patients (42). Because the homozygous variant AA genotype had a very low frequency (1% in the controls), the possibility that its association with increased risk of pancreatic cancer was due to chance alone cannot be excluded. Notably, a possible synergistic effect of this SNP with diabetes was also observed in this study population. The OR of pancreatic cancer was 4.17 for diabetics carrying the A allele versus 2.08 for diabetics carrying the G allele compared with nondiabetics. Because diabetes could be a manifest of pancreatic cancer, we did analysis among those with a >2 years of diabetes duration to reduce the problem of reversal causality; and the interaction of ATM D1853N genotype with diabetes remained statistically significant. It is conceivable that diabetes increases oxidative stress and deficiency in the defending DNA repair genes would confer a greater chance for tumor development via impaired cellular response to accumulative DNA damage.

The OGG1 gene is a base excision repair gene that removes an oxidative DNA lesion, such as 7,8-dihydro-8-oxo-guanine, from DNA. Because 7,8-dihydro-8-oxo-guanine leads to a high degree of DNA mispairing, decreased OGG1 activity could lead to a higher frequency of mutation and could possibly increase the cancer risk of an individual under oxidative stress. Among the numerous DNA repair genes that have been investigated in cancer studies, the OGG1 genotype has been frequently associated with altered risk of human cancers (43, 44). We have previously observed a significantly reduced overall survival time for patients carrying the OGG1 C-315G (aka S326C) GG homozygous variant genotype (22). The current study observed a weak interaction of the OGG1 C-315G CC/CG genotype with diabetes in increased risk of pancreatic cancer. Together these observations suggest that the CC/CG genotype may confer relatively little enzyme activity, thus causing an increased risk of pancreatic cancer. However, neither the cancer association nor the functional significance of the OGG1 genotype is conclusive (43, 44). Furthermore, our observations need to be replicated in other study populations.

Our study has several merits as well as limitations. Except OGG1 (12), none of the genes investigated in the current study has previously been examined in association with risk of pancreatic cancer. We, for the first time, showed a possible role of genes involved in the repair of DNA strand breaks (LIG3 and LIG4) or in cellular response to DNA damage (ATM) in modifying the risk of pancreatic cancer. Although the current study has a relatively large sample size, the power to detect the main genotype effect in low-frequency homozygous mutants and the interaction of genotype with other risk factors is still limited. So the possibility that some of the observations were false discoveries associated with multiple testing cannot be excluded. In addition, the study was conducted in a single tertiary referral hospital; results from this study population may not be generalized to the U.S. population. Some of the genes investigated in this study have previously been associated with patient survival, the association of genotype and risk of pancreatic cancer could be biased if our study missed a lot of patients that were succumbed to this fatal disease rapidly. We also admit that the number of genes and SNPs selected were very limited and the functional significance of the SNPs was largely unknown. We may have missed many important genes and SNPs on the same DNA repair pathways. Future large studies with systematic selection of genes and SNPs are required to provide sufficient power to reveal the true associations between the rare variant genotypes and the risk of pancreatic cancer.

No potential conflicts of interest were disclosed.

We thank all the personnel involved in patient recruitment and data collection, and Kathryn B. Carnes and Sue Moreau for editing the manuscript.