The Association of Polymorphisms in Base Excision Repair Genes with Ovarian Cancer Susceptibility in Chinese Women: A Two-Center Case-Control Study

Base excision repair (BER) acts upon the most important mechanism of the DNA repair system, protecting DNA stability and integrity from the mutagenic and cytotoxic effects. Multiple researches have indicated that single-nucleotide polymorphisms (SNPs) in the BER-related gene may be associated with the susceptibility of ovarian cancer. However, the results are controversial. In this two-center case-control study, 19 potentially functional SNPs in six BER-related genes (hOGG1, APE1, PARP1, FEN1, LIG3 and XRCC1) was genotyped in 196 ovarian cancer cases and 272 cancer-free controls. And, their associations with ovarian cancer risk were assessed by unconditional logistic regression analyses. We found that PARP1 rs8679 and hOGG1 rs293795 polymorphisms were associated with a decreased risk of ovarian cancer under dominant model (adjusted OR=0.39, 95% CI=0.17-0.90, P=0.026; and adjusted OR=0.36, 95% CI=0.13-0.99, P=0.049, respectively). Stratification analysis demonstrated that this association was more pronounced in the subgroups of lower BMI and patients with early menarche and serous carcinoma. Moreover, LIG3 rs4796030 AA/AC variant genotypes performed an increased risk of ovarian cancer under recessive model (adjusted OR=1.54, 95% CI=1.01-2.35, P=0.046), especially in the subgroups of higher BMI, early clinic stage and the carcinoma at the left. These results suggested that PARP1, hOGG1 and LIG3 polymorphisms might impact on the risk of ovarian cancer. However, more researches with larger and different ethnic populations are warranted to support our findings.


Introduction
Ovarian cancer is the main cause of death in gynecological malignancy according to global cancer statistics in 2018, with an estimated 184,799 deaths worldwide [1]. The common risk factors of that result in extremely poor 5-year survival, such as asymptomatic characteristics and the lack of early detection strategies, have become an immediate problem to be solved for ovarian cancer [2,3]. Thus, ovarian cancer remains a great burden and challenge for the affected family and public health.
Numerous studies have indicated that the risk of ovarian cancer was related to BRCA1 and BRCA2 genes. Marc D. et al. [4] found that BRCA1/2 mutations may ultimately impact ovarian cancer Ivyspring International Publisher mortality. Adrianna et al. [5] demonstrated significantly increased APC rs11954856 and rs351771 frequencies in Polish women with ovarian cancer. However, several genetic studies published recently had suggested that the risk of ovarian cancer was related to the mutations and polymorphic variants in DNA repair genes [6,7]. DNA repair mechanism mainly include direct repair, base excision repair (BER), nucleotide excision repair (NER) and mismatch repair (MMR) [8]. The primary DNA damage repair pathway, BER, is responsible for the modification of the most common forms of DNA damage, such as spontaneous hydrolytic, oxidative and alkylative lesions and single-strand breaks [9]. Although considerable genetic studies demonstrated that the deficiency of BER related genes would influence on ovarian cancer risk, the real relationship between both of them remained controversial [10].
BER processes can be divided into the following three basic steps: damage recognition/strand scission, gap tailoring and gap filling/nick ligation [11]. Based on different types of base lesions, either of two subways (short patch or long patch BER) will be initiated by corresponding lesion-specific DNA glycosylase [12]. 8-oxoguanine glycosylase (hOGG1) acts as one of the most important bifunctional glycosylases, which cannot only excise the substrate base but also incises DNA 3' to the damage site via an intrinsic AP lyase activity [13]. Meanwhile, apurinic endonuclease 1 (APE1) removes the resulting 3'-obstructive termini and prepares for next polymerization and/or nick ligation [14]. Differing from the classical initial step, however, the singlestrand break repair (SSBR) depends on poly (ADPribose) polymerase 1 (PARP1) to recognize the damage and recruit other protein to the damage site [15]. In addition, flap endonuclease (FEN1) was also included in our study for its gap-tailoring properties in the DNA ends [16]. DNA ligase 1 (LIG1) and the complex of DNA ligase 3 (LIG3)/X-ray repair cross-complementing 1 (XRCC1) exhibit similar performance of nick ligation, but we only selected LIG3/XRCC1 as the investigated biomarker for its unique character of collecting proteins at the damaged site [17,18]. In the current study, we analyze the associations between 19 SNPs of six BER-related genes (hOGG1, APE1, PARP1, FEN1, LIG3 and XRCC1) and ovarian cancer risk by genotyping method in 196 patients and 272 controls.

Patients and controls
In this hospital-based case-control study, 196 ovarian cancer patients were newly diagnosed and histopathologically confirmed in two different institutions of the Second Affiliated Hospital and Yuying Children's Hospital and the First Affiliated Hospital of Wenzhou Medical University (WMU) from February 2007 to March 2018. Additionally, we selected 272 cancer-free women who participated in the routine physical examination of the Second Affiliated Hospital and Yuying Children's Hospital of WMU. All of them were frequency-matched to cases on age (± 2 years) and race/ethnicity, and reconfirmed without any personal tumor history or family tumor history. All people included in the present research had signed a written informed consent. The demographic data and environmental exposure factors, such as age, BMI, menarche, menopause, number of pregnancies, were collected from each participant. Clinical and pathological information were further extracted from the patients' electronical database, including pathologic classification (WHO 2014), histology grade, FIGO stage (International Federation of Gynecology and Obstetrics), tumor size (the largest tumor diameter of the primary tumor) and anatomic neoplasm subdivision. However, the patients who were diagnosed without histology grade would be categorized as a single group of "undermined". And, some patients with some rare pathological classification were included in the "others" pathological group. The research was approved by the Second Affiliated Hospital and Yuying Children's Hospital of WMU. Our study was conducted following the Declaration of Helsinki, and participants or guardians were required to sign informed consent forms. Blood samples were obtained from cases before receiving radiotherapy or chemotherapy.
For all cases, the DNA genomic was extracted from paraffin-embedded tissue using TIANquick FFPE DNA Kit (Qiagen Inc., Valencia, CA), while the DNA genomic of all controls were extracted from the peripheral blood specimens using the TIANamp Blood DNA Kit (TianGen Biotech Co. Ltd.). DNA purity and concentration were measured by a UV spectrophotometer (Nano Drop Technologies, Inc., Wilmington, DE).
According to standard protocols, the genotyping of 19 selected SNPs was performed with a TaqMan real-time PCR assay in ABI Prism 7900HT genetic detection system (Applied Biosystems Inc., USA) [20][21][22]. To confirm the accuracy of the genotyping, 5% samples were randomly selected as positive controls (repeat samples) and negative controls (without DNA template). As a result, the genotyping success rate reached 96.8%, and the results of duplicated samples were 100% concordant.

Statistical analysis
All statistical tests were performed by SAS software (Version 9.4; SAS Institute, Cary, NC, USA). Differences in demographic (e.g., age and BMI) and other covariates (e.g., menarche, menopause and number of pregnancies status) between patients and controls were evaluated by Pearson's χ 2 test. To assess the associations of these 19 SNPs with ovarian cancer risk, the adjusted odds ratios (ORs) and 95% confidence intervals (CIs) were calculated for dominant (AB+BB vs. AA) and recessive models (BB vs. AB+AA) by unconditional logistic regression analyses. A and B respectively represent different alleles (wild and mutant) at this location. Additionally, stratification analysis was performed to further evaluate the significant associations existed in which demographic or clinic covariates groups. Deviation from Hardy-Weinberg Equilibrium (HWE) in the controls was assessed by a goodness-of-fit χ 2 test. All P value was two-sided with a significant level < 0.05.

Population characteristics
In the current study, we enrolled 196 ovarian cancer patients with an average age of 50.01 ± 12.51 years and 272 cancer-free controls with an average age of 50.48 ± 11.98 years. The demographic and clinical characteristics of all participants were shown in Supplemental Table S2. There were no significant differences between cases and controls in age (P=0.570), BMI (P=0.947), menarche (P=0.072), menopause (P=0.421) and number of pregnancies (P=0.448). To discuss the pathological classification, 132 (67.4%), 27 (13.8%) and 15 (7.7%) cases were subdivided into serous, mucinous and clear cell carcinoma, respectively, while 22 (11.2%) cases were included into "others" because of the limited availability of tumor samples. Regarding anatomic neoplasm subdivision, 27 (13.8%), 74 (37.8%) and 95 (48.5%) ovarian cancers occurred in the left, right and both sides, respectively. In term of tumor size, 27 (13.8%) patients reported with a tumor of < 5cm, 74 (37.8%) with a tumor of 5-10 cm and 95 (48.5%) with a tumor of >10cm. According to the FIGO stage, 96 (49.0%) and 100 (51.0%) cases distributed into the stage of I+II and III+V, respectively. Moreover, we classified all patients into four different histological grades, 14 (7.1%) in G1, 20 (10.2%) in G2, 114 (58.2%) in G3+G4 and 17 (8.7%) in borderline tumor, but an exception of 31 cases (15.8%) were into the groups of "undetermined" due to the lack of information.

Association between the SNPs of BER-related genes and risk of ovarian cancer
Genotype distributions of these 19 SNPs are shown in Table 1

Stratification analysis
In order to further explore these significant genetic associations existing in which kinds of ovarian cancer patients, we performed a stratification analysis by age, BMI, menarche, menopause and number of pregnancies status, as well as some clinic-related factors like pathologic classification, histology grade, FIGO stage, tumor size and anatomic neoplasm subdivision (

Discussion
The evidence derived from numerous studies had testified the important role of BER in the prevention of mutation accumulation [23,24]. According to different base lesions, BER can trigger corresponding subways and assemble a series of repair complex at the site of DNA lesions [25]. Compared with other DNA repair system, the core components of BER are quite conserved across most species. As is mentioned above, the whole repair processes are simplified as three distinct phases, and the completion of each step is based on the scaffold protein of XRCC1/LIG3 and PARP1 [26]. hOGG1, one of the most important bi-functional glycosylases, is initiated by some oxidative base lesions [13]. The study reported by Xie et al. found that the hOGG1 deficiency animals would increase tumor predisposition, especially for the lung and ovarian tumors and lymphomas [27]. Similarly, with the increasing of spontaneous mutant frequencies in the APE1 heterozygous mice, the organism also suffered many deleterious effects brought by oxidative stress, including cancer [28,29]. Although FEN1 was the only one endonuclease included in the present study, its potential role in cancer development was fully demonstrated by a heterozygous animal model [30].
Numerous investigations have tried to uncover the genetic association of ovarian cancer in these six genes; however, no firm conclusions could be drawn from them. For example, hOGG1 rs1052133 as one of the hottest spots in the recent years has been researched extensively in ovarian cancer. According to the studies of Michalska and Chen, hOGG1 rs1052133 C>G polymorphism was considered as an unfavorable factor for the susceptibility of ovarian cancer [31,32], but a totally different result was discovered in our research. Arc and reached the same conclusion with us and suggested that there was no association between both of them [33]. Likewise, a negative association was observed between XRCC1 rs25487 G>A and ovarian cancer risk in Serbian women [34], while this finding was unable to repeat in another study by Khokhrin and ours [35]. In fact, these contradictory conclusions were caused by several factors, of which the relatively limited sample size and different genetic backgrounds a probably the most critical reasons.
In this current hospital-based case-control study, we not only explored the impact of 19 potentially functional SNPs in six BER-related genes on ovarian cancer risk, but also conducted an analysis of gene-environmental interactions in a certain number of cases and controls from two different institutions.  The genotyping results indicated that the PARP1 rs8679 and hOGG1 rs293795 A>G polymorphisms were associated with the decreased ovarian cancer risk under a dominant model, while LIG3 rs4796030 A>C polymorphism with an increased ovarian cancer risk under a recessive model. Additionally, we noticed that the protective effect of PARP1 rs8679 A>G polymorphism was more pronounced among the groups of BMI < 24, menarche age ≤ 14 and serous carcinoma. Moreover, the negative effect of LIG3 rs4796030 A>C polymorphism significantly associated with the patients in the groups of BMI ≥ 24, clinical stages III/IV and tumor at the left. The preliminary results obtained here indicated that compared with patients of BMI ≥ 24, patients of BMI < 24 had a more prominent protective effect of genetic variation, and their negative effects were relatively small. These results showed that healthy lifestyle also played an important role in the development of cancer. The role of genetic variation was also related to the pathological type and clinical stage of cancer.
Although we found two novel related SNPs associating with the risk of ovarian cancer, some inherent limitations still should be listed to discussion. Firstly, the limited number of cases and controls would cause a deficiency of statistical power to some extent. Secondly, many relative risk factors of ovarian cancer, such as oral contraceptive, CA125 and breast feeding, are not fully considered due to lack of individual information. Thirdly, only 19 SNPs of six BER-related genes were investigated in the present study, which was insufficiently to explain the biological relationship of BER with the risk of ovarian cancer. Fourthly, we noticed that the genotype frequencies of FEN1 rs174538 G>A deviate from HWE, which indicates that our research may exist the problem sampling bias to some extent. Finally, these new findings in our research were not confirmed in vitro and in vivo experiences, and the survival analysis for SNPs was not explored further.
In conclusion, our study suggested that the PARP1 and hOGG1 polymorphisms might correlate to ovarian cancer susceptibility. However, more comprehensive studies with larger independent cohorts should be provided to verify the relationship between these significant genetic variations in PARP1 and hOGG1 gens and ovarian cancer risk.