We analyzed the survival curves for the three polymorphisms in a group of 470 breast cancer patients with follow up. The characteristics of these patients are detailed in Table 1. Of the 470 patients included in the study, 159 patients progressed and 88 died. The Kaplan-Meier analysis shows a trend to association with overall survival (OS) for the polymorphisms rs1052133 on the OGG1 gene and rs207454 on the XDH gene.

The survival analysis for rs1052133 shows a trend (p-value = 0.058, borderline significance) with an OS increase of 15.02 months in the patients with genotype CC + CG versus the patients with the GG genotype (CC + CG: 113.00 months [95% CI 103.64 to 122.35 months] vs. GG: 97.98 months [95% CI 93.52 to 102.45 months]) (Figure 1A). Similarly, the polymorphisms rs207454 were associated with the risk of breast cancer in a recessive model. The genotype CC predisposes to the illness, and it also shows a lower overall survival in the Kaplan-Meier analysis (AA + AC: 99.67 months [95% CI, 95.62 to 103.72 months] vs. CC: 76.90 months [95% CI 39.60 to 114.20 months]) with a difference between medians of 22.77 months (p = 0.076) (Figure 1B).

The polymorphism rs3736729 on the GCLC gene shows a significant association with the DFS and OS (Table 2). If we compare the three possible genotypes (CC, AA and AC) with the time to first recurrence, the association is statistically significant (p-value = 0.015). The AC genotype shows a lower DFS median time compared to the other genotypes (AC: 78.62 months [95% CI, 71.07 to 86.17 months] vs. AA: 90.74 months [95% CI, 82.74 to 98.74 months] and CC: 93.24 months [95% CI, 83.98 to 102.49 months]). The association is even more powerful (p = 0.004) when we compare the heterozygote vs. homozygote genotypes (AC: 78.62 months [95% CI, 71. 07 to 86.17 months] vs. AA/CC: 91.77 months [95% CI, 85.71 to 97.82 months]). The analysis of OS shows a similar tendency with a borderline significance (p = 0.05). We did not find any relationship between the drug or administration treatment followed by the patients and the outcome. The analysis of association between polymorphisms and the characteristic of the patients shows significance for the tumor size and the hormonal receptor status (data not shown). Consequently, we performed the analysis of survival depending on these variables.

The polymorphisms rs1052133 and rs207454 show a trend toward significance in the OS analysis when the patients were ER+ (p = 0.077 and p = 0.053, respectively) (Figure 2A,B). The analysis, as a result of the progesterone receptor (PGR) status, shows an association only in the case of the rs207454, with a significant value in the time of DFS (p-value = 0.024) and a trend in the OS (p = 0.062). Both cases present the lowest survival time in patients with the CC genotype and PGR− (Figure 2C,D).

The genotype of the polymorphism rs3736729 presents the highest association with ER− patients in comparison to the total breast cancer cases analyzed. The data for DFS showed that the patients with the AC genotype, and in the absence of ER expression, present the worse median survival (65.01 months [95% CI, 53.59 to 76.43 months], p-value = 0.001) compared to patients with the other genotypes. The difference in survival time between these groups is 24.62 months. Consequently, the OS was reduced in the ER− vs. overall group for the AC genotype (82.99 months [95% CI, 71.32 to 94.66 months] vs. 93.02 months [95% CI, 85.77 to 100.26 months]). The decreased survival is 18.64 months. Figure 3 shows the survival curves according to the GLCL genotype of ER− patients. There were no significant associations between genotype and either DFS or OS among patients with ER+. The ten year DFS and OS for each rs3736729 genotype by tumor ER and PGR status and overall group are shown in Table 2.

The analysis of association between PGR and rs3736729 shows a relationship exclusively with the PGR− subgroup of patients, where the AC genotype again corresponds to the worst outcome (64.82 months [95% CI, 53.97 to 75.67 months]. However, the significance is not as good as that for the ER− subgroup (p = 0.001 vs. p = 0.032, respectively). The combination of PGR− and ER− shows a DFS median survival with lower significance than for ER− alone (p = 0.005 vs. p = 0.001, respectively). These data suggest a possible contribution of the AC genotype to DFS in patients according to their ER status. A similar conclusion may be reached in the OS study for ER− and PGR− vs. ER−(p = 0.017 vs. p = 0.012, respectively). The analysis of the polymorphism rs1052133, rs207454 and rs3736729 with the tumor size does not show a significant relationship.

The influence of many polymorphisms in the predisposition to breast cancer has been described in the literature. However, the relationship between these genetic variants and outcomes in breast cancer patients is more limited. To best understand the role of the three significant SNPs in breast cancer, in this work, we decided to analyze the correlation of breast cancer patient survival with the different genotypes.

Our results suggest an association between the rs207454 polymorphism on XDH (encoding an enzyme involved in the oxidative metabolism of purines) [32] and survival of breast cancer patients. The OS in the total population shows a trend toward significance. However, the analysis in the subgroup of PGR patients presents a significant association between the genotype and the time to relapse of the illness when using a recessive model (AA + AC vs. CC). These data suggest that the rs207454 genotype could be a marker of DFS in the subgroup of patients in which PGR is absent.

Our data suggest an association between the polymorphisms rs3736729 on the GCLC gene and the progression of breast cancer patients compatible with an over-dominant model, where the genotype AC gives the worst prognosis in patients with breast cancer. These results indicate that individuals heterozygous for this polymorphism have a remarkably poor prognosis vis-à-vis the evolution of their breast cancer. The most common models adopted in statistics are the dominant and the recessive ones. Nevertheless, over-dominance is observed in many human diseases. Molecular heterosis is common in humans and may occur in up to 50% of gene associations. Heterosis in the SLC6A4 serotonin transporter gene, HTR2A, ESR1, TIGR and DRD2 genes has been described. A possible mechanism by which heterosis could exert a biological effect is at the level of protein subunit interaction. The heterosis could produce allosteric changes that affect protein function and result in a more (or less) efficient protein than the wild-type [33]. Our hypothesis is that the heterozygote genotype of rs3736729 could decrease the efficacy of the GLCL enzyme and, as a result, diminish the antioxidant capacity of each person.

The sharp association between the rs3736729 genotype on GCLC and the absence of ER suggest the existence of a possible mechanism of interrelation between the nuclear receptor and genes of the stress oxidative pathway. Indeed, 17-β-estradiol induces production of antioxidative enzymes, including gamma-glutamylcysteine synthetase (g-GCS/GCLC). The effects of 17-β-estradiol are mediated mostly through ER, which belong to the nuclear receptor superfamily and functions as a ligand-induced transcription factor. Therefore, one of the important ways to increase the expression of GCLC is through the activation of the ER receptor. Furthermore, the enzyme GCLC, which plays a critical function in the protection against reactive oxygen species, is also essential in the process of detoxification of xenobiotics. The exposure of cells to these agents results in a significant increase of the enzyme levels by the transcriptional up-regulation of the genes (GCLC and GLCM) encoding the two subunits [34,35]. This bibliographic evidence suggests that ER− patients would have less anti-oxidative potential as a consequence of the ER expression reduction. Consequently, the expression levels of GCLC induced by ER would decrease. The combination of this fact with a possible unfavorable rs3736729 genotype (AC) could potentiate the early recidive of this group of patients. This hypothesis seems reasonable, since several studies suggest that oxidative stress contributes significantly to cancer progression by the alteration of the redox control of the cell cycle [36,37]. We postulate that the genotype of rs3736729 could determine different outcomes in ER− breast cancer patients, just as the genotype of rs207454 could influence the time of disease free survival in PGR− patients. This is critical to the identification of subpopulations for which the current therapy is insufficient, and new clinical approximations are required. Several therapeutic agents commonly used in a clinical setting, such as radiation therapy and chemotherapeutic agents, exert antitumor effects through the increased formation of reactive oxygen species (ROS). Nowadays there are in development pharmacological agents (considered dirty drugs) that modulate cellular redox through pleiotropic interactions. Ongoing studies are testing the activity of these agents as combinatorial drugs consistent with the induction of deviation from the redox homeostasis that sensitizes cancer cells to the cytotoxic effects of chemotherapeutic agents. The benefit provided by redox chemotherapeutics depends on the genotypic and phenotypic profiling of the patients, as the case may be for the different rs3736729 or rs207454 genotypes.

Our study has several potential limitations to consider, such as the possible interaction with other polymorphisms located in these genes, but not included in this study. Also, some SNPs are located in non-coding regions. However, variations in intronic structure have been proposed as an influence in cancer via the regulation of gene expression, gene splicing or mRNA stability. It is also possible that these polymorphisms are in linkage disequilibrium with other functional polymorphisms that may affect breast cancer outcome. Despite these considerations, our work, as far as we know, is the first study that shows a relationship between polymorphisms on the GCLC and XDH genes and the survival outcome in ER− and PGR− breast cancer patients, respectively. The results obtained in this work seem to sustain the hypothesis of the participation of oxidant and antioxidant genes in breast cancer. However, we consider further studies necessary to confirm this result in independent population groups and to test for possible therapeutic alternatives.

The present study included a population-based cohort of 470 women diagnosed with primary breast cancer at the Clinic Hospital of Valencia (Spain). The recruitment was done between 1983 and 1998, and the subjects were adult, Caucasian females (mean age at diagnosis of 54.1 years, range 20.5–86.5) and residents in the area of influence of the Hospital. Recurrences were determined by oncologists, and patient death was always confirmed by death certificates. All the participants in the study were informed and gave their written informed consent to participate in the study. The characteristics and follow-up of patients was routinely updated in the data bases by trained Hospital personnel. The mean follow-up was 52.44 months.

The blood samples of the 470 patients were collected over a period of six months prior to the start of this study. Genomic DNA was extracted from blood samples using the DNeasy tissue kit from Qiagen (Izasa, Madrid, Spain). A final elution volume of 100 μL was established. DNA quantity was measured by absorbance at 260nm using a NanoDrop spectrophotometer, and DNA purity was evaluated by measurement of the 260/280 absorbance ratios. Each DNA sample was stored at −20 °C until analysis. Genotyping analysis was performed by SNPlex (genotyping system based on ligation assay/polymerase chain reaction technology; Applied Biosystems, Foster City, CA, USA) according to the manufacturer’s protocol.

The selection of polymorphisms to study (rs3736729 (GCLC), rs207454 (XDH) and rs1052133 (OGG1)) was based on previous data from our group that has associated genetic variants of these polymorphisms with a predisposition to breast cancer.

The SPSS statistical package (version 17.0) was used to analyze, by regression methods, the association between polymorphisms and clinical variables. In addition to the genetic information, the following characteristics of the patients were included in the analysis: age, hormonal receptor status (ER and PGR), Her-2 overexpression, menopausal status, histological status, grade, tumor size, nodule status, metastases, type of treatment (adjuvant or neoadjuvant), drug of treatment (anthracyclines, tamoxifen and others) and radiotherapy. The non-parametric Kruskal–Wallis method was used for testing the association between quantitative variables and polymorphisms genotypes (Table 3). The Cox regression analysis was used to adjust by variables. Imputation of missing values was performed by multiple regression. Disease-free survival (DFS) and overall survival (OS) curves were plotted according to the Kaplan-Meier method, and differences between groups were assessed using the log-rank test. DFS was defined as time (months) from diagnostic to earliest recurrence or to last contact with the patient, while OS was defined as time (months) from diagnosis to death caused by breast cancer. Other causes of death were considered as censored observations and also contribute to estimate the survival. Survival analysis was carried out for the complete patient sample, as well as subgroups, according to ER and PR status. The follow up times were censored at 10 years. In all tests, p-values equal to or lower than 0.05 were considered statistically significant.