An Analysis of Polymorphisms Within the Wnt Signaling Pathway in Relation to Ovarian Cancer Risk in a Polish Population

The patients include 258 women with histologically determined ovarian carcinoma according to the International Federation of Gynecology and Obstetrics (FIGO). They were inducted into the study from the University Hospital, Clinic of Gynecological Surgery and Chair of Gynecologic Oncology at Poznań University of Medical Sciences (Poznań, Poland). Histopathological classification, including the stage, grade, and tumor type, was performed by an experienced pathologist (Table 1). The control group included 282 unrelated healthy female volunteers who were matched by age to the female cancer patients. Controls were selected during medical examination at the University Hospital, Clinic of Gynecological Surgery at Poznań University of Medical Sciences (Table 1). Written informed consent was obtained from all participating individuals. The procedures of the study were approved by the Local Ethical Committee of Poznań University of Medical Sciences. All women with ovarian cancer and controls were Caucasian from the Wielkopolska area of Poland.

Genomic DNA was isolated from peripheral blood leukocytes by salt extraction.

All patients and controls with ovarian cancer were tested for the three most common BRCA1 mutations (5382incC, C61G, 4153delA) affecting the Polish population using the LightCycler^® 480 system with HybProbe assays (Roche, Indianapolis, IN, USA). In addition, the patient group was tested for the presence of the most common BRCA2 mutation (5946delT) using high-resolution melting curve analysis (HRM) [Supplemental Table 2]. Information on HybProbe probe sequences is available upon request. The DNA samples were then genotyped for the nine SNPs in CTNNB1, APC, and AXIN2 (Supplemental Table 1 and Supplemental Fig. 1A, Fig. 1B, and Fig. 1C). SNPs were selected with the use of the genome browsers of the International HapMap Consortium (http://​www.​hapmap.​org/​index.​html.​en), University of California Santa Cruz (UCSC; http://​genome.​ucsc.​edu), and dbSNP database (http://​www.​ncbi.​nlm.​nih.​gov/​projects/​SNP/​). SNPs were selected based on functional significance, location in different LD blocks, and minor allele frequency (MAF) >0.1 in the Caucasian population. Genotyping of the CTNNB1 rs4533622, rs2953, APC rs351771, and AXIN2 rs4074947, rs3923087, rs2240308 SNPs was performed by HRM on the LightCycler^® 480 system (Roche Diagnostics, Mannheim, Germany). Genotyping of the APC rs11954856, rs459552, and AXIN2 rs7224837 SNPs was performed by PCR, followed by the appropriate restriction enzyme digestion [PCR–restriction fragment length polymorphism (PCR-RFLP)] according to the manufacturer’s instructions (Fermentas, Vilnius, Lithuania). Primer sequences and conditions for HRM and PCR-RFLP analyses are presented in Supplemental Table 2. Genotyping quality was evaluated by repeated genotyping of 10 % randomly selected samples.

For each SNP, the Hardy-Weinberg equilibrium (HWE) was assessed by Pearson’s goodness-of-fit Chi-square (χ²) statistic. The differences in the allele and genotype frequencies between cases and controls were determined using standard χ² or Fisher tests. The odds ratio (OR) and associated 95 % confidence intervals were also calculated. The data were analyzed under recessive and dominant inheritance models. For the additive inheritance model, SNPs were tested for association with ovarian cancer using the Cochran–Armitage trend test. To adjust for the multiple testing, the Single Nucleotide Polymorphism Spectral Decomposition (SNPpD) method was used [33]. Pair-wise LD between selected SNPs was computed as both D′ and r ² values using HaploView 4.0 software (Broad Institute, Cambridge, MA, USA). Haplotype analysis was performed using the UNPHASED 3.1.5 program with the following analysis options: all window sizes, full model, and uncertain haplotype [34]. Haplotypes with a frequency below 0.01 were set to zero. The p values for both global and individual tests of haplotype distribution between cases and controls were determined. Statistical significance was assessed using the 1,000-fold permutation test.

High-order gene–gene interactions among all tested polymorphic loci were studied by the multifactor dimensionality reduction (MDR) approach (MDR version 2.0 beta 5). A detailed explanation on the MDR method has been described elsewhere [35]. Based on the obtained testing balanced accuracy and cross-validation consistency values, the best statistical gene–gene interaction models were established. A 1,000-fold permutation test was used to assess the statistical significance of MDR models (MDR permutation testing module 0.4.9 alpha).

In patients with ovarian cancer, 30 carriers of the most common BRCA1 mutations were identified. The BRCA1 5382insC mutation was identified in 22 individuals, C61G in five individuals, and 4153delA in three individuals (Supplemental Table 3). None of the patients was a carrier of the BRCA2 5946delT mutation. Moreover, none of the controls had the tested BRCA1/BRCA2 nucleotide variants.

The distribution of CTNNB1, APC, and AXIN2 genotypes did not display deviation from HWE between patients and control groups (p > 0.05). The number of genotypes, ORs, and 95 % confidence interval calculations for the nine CTNNB1, APC, and AXIN2 polymorphisms are shown in Table 2.

The lowest p values of the trend test (p _trend) were observed for the APC rs351771 and rs11954856 SNPs in patients with ovarian cancer (p _trend = 0.006 and p _trend = 0.007, respectively) [Table 2]. Moreover, we observed that, in a dominant inheritance model, the APC rs11954856 SNP is associated with an increased risk of ovarian cancer development [OR = 2.034 (95 % CI 1.302–3.178); p = 0.002]. We also found significant allelic differences for the APC rs351771 SNP (p = 0.006) between patients and controls. The experiment-wide significance threshold required to keep Type I error rate at 5 % was 0.006 (effective number of independent marker loci: 8.388). There was no association of the APC rs11954856 SNP (Supplemental Table 4) or other tested SNPs (not shown) with any histologic subtypes of ovarian cancer. Furthermore, none of the other eight CTNNB1, APC, and AXIN2 polymorphisms displayed significant association with ovarian cancer development either in dominant, recessive, or additive inheritance models (Table 2).

Haplotype analysis of the studied CTNNB1, APC, and AXIN2 polymorphisms did not reveal SNP combinations associated with the risk of ovarian cancer development (Table 3). The lowest global p = 0.025 was observed for haplotypes composed of the APC rs11954856 and rs351771 SNPs (Table 3). However, these results were not statistically significant when permutations were used to generate empiric p values. The empiric 5 % quintile of the best p value after 1,000 permutations was 0.03904 for CTNNB1, 0.00748 for APC, and 0.00888 for AXIN2 haplotypes. The CTNNB1 rs4533622 and rs2953 SNPs were in perfect LD (Supplemental Table 5). The SNPs situated in distinct regions of APC and AXIN2 were either in strong or weak pairwise LD. This was calculated from the control samples, and had D′ ranges of 0.401–0.988 for APC SNPs and 0.002–1.000 for AXIN2 SNPs (Supplemental Table 5).

Exhaustive MDR analysis evaluating two- to four-loci combinations of all studied SNPs for each comparison did not demonstrate statistical significance in predicting susceptibility to ovarian cancer development (Table 4). The best combination of possibly interactive polymorphisms was observed for rs4533622 of CTNNB1, rs11954856 of APC, and rs7224837 and rs2240308 for AXIN2 (testing balanced accuracy = 0.5719, cross validation consistency of 10 out of 10, permutation test p = 0.068).