Association of superoxide dismutases and NAD(P)H quinone oxidoreductases with prognosis of patients with breast carcinomas

Authors


Abstract

Associations of transcript levels of oxidative stress-modifying genes SOD2, SOD3, NQO1 and NQO2 and their functional single nucleotide polymorphisms (SNPs) rs4880, rs1799895, rs2536512, rs699473, rs1800566 and rs1143684 with prognosis of breast cancer patients were studied. SNPs were assessed by allelic discrimination in a cohort of 321 breast cancer patients from the Czech Republic. Transcript levels were determined by real-time polymerase chain reaction (PCR) with absolute quantification in tumor and adjacent non-neoplastic control tissues. Both genotypes and transcript levels were then compared with available clinical data on patients. Patients carrying low activity allele Leu in NQO2 rs1143684 had a greater incidence of stage 0 or I disease (i.e., better prognosis) than patients with the Phe/Phe genotype. This association was more evident in patients without expression of progesterone receptors (p = 0.031). Patients carrying the Thr allele in SOD3 rs2536512 SNP had a significantly greater incidence of tumors expressing estrogen receptors than patients carrying the Ala/Ala genotype (p = 0.007). SOD3 transcript level was significantly higher in grade 1 or 2 tumors than in grade 3 tumors (p = 0.006). Patients carrying T allele in SOD3 rs699473 SNP had significantly poorer progression-free survival (PFS) than patients carrying the CC genotype (p = 0.038). The same applied to the subgroup of patients treated by hormonal regimens (p = 0.021). Patients carrying the high activity Ala/Ala genotype in SOD2 (rs4880) had significantly poorer PFS than Val allele carriers in the group treated by cyclophosphamide but not hormonal regimens (p = 0.004). Our results suggest that NQO2, SOD2 and SOD3 may significantly modify prognosis of breast cancer patients and that their significance should be further characterized.

Breast cancer is the most common cancer in women. In 2008, about 1,383,523 new cases of invasive breast cancer were diagnosed worldwide.1 The necessity of treating such a large number of patients calls for efficient tools that can be used for subgrouping patients according to estimated prognosis. A different spectrum of treatment modalities with diverse mechanisms of action and adverse effects could then be offered to various prognostic groups. Besides well-established classical prognostic factors such as tumor size, nodal status, grading and expression of hormonal receptors, numerous biological molecules are under investigation as potential prognostic biomarkers in breast carcinomas.

Excess of oxidative stress, mediated by reactive oxygen species, may cause cellular deregulation leading to cell apoptosis,2 proliferation or tumor promotion.3 Alkylation and topoisomerase poisoning present the major mechanisms of action of cyclophosphamide and anthracyclines, respectively. Oxidative stress represents an additional cytotoxic mechanism.4, 5 One of the initial molecules of oxidative stress, superoxide anion radical, is formed by the univalent reduction of triplet-state molecular oxygen. This process is mediated by enzymes such as NAD(P)H oxidases and xanthine oxidase or non-enzymatically by redox-reactive compounds such as the semiubiquinone compound of the mitochondrial electron transport chain.6 NAD(P)H:quinone oxidoreductases (NQO1 and NQO2) protect cells from oxidative damage by catalyzing reduction of carcinogenic quinone compounds to their hydroquinone forms.7 Ahn et al. suggested that deletion of NQO2 plays a differential role in the tumor necrosis factor (TNF) signaling pathway by downregulation of cell survival signals and by upregulation of TNF-induced apoptosis.8 NQO2 was also shown to reduce estrogen orthoquinones, indicating that it could be a novel target for prevention of breast cancer initiation.9 NQO1 (as well as NQO2) is also involved in the protection of a “gatekeeper,” TP53, from 20S proteasomal degradation.10–12 Single nucleotide polymorphisms (SNPs) in exon 6 of NQO1 (Pro187Ser, rs1800566) and exon 3 of NQO2 (Phe47Leu, rs1143684) were associated with lower enzyme activities.13, 14 A recent study suggested that the genotype Ser/Ser in NQO1 SNP rs1800566 is a strong prognostic and predictive factor in breast cancer.15 Fagerholm et al. proposed that NQO1 influences the outcome of epirubicin treatment through at least three mechanisms: the TP53 and TNF-NF-kB pathways and direct detoxification of reactive oxygen species.15 Our study at both mRNA and protein levels did not find additional evidence about the mechanism of NQO1-TP53 association.16 Thus, more complex mechanism may be behind the reported association.

Manganese superoxide dismutase (SOD2) converts the superoxide anion radical into hydrogen peroxide and oxygen in mitochondria and plays a key role in protecting cells from oxidative damage.17 The SOD2 SNP (Val16Ala, rs4880) in gene sequence leads to a conformational change in the helical structure of the protein. The Val allele of the enzyme has lower efficiency of transport into mitochondria.18 A recent study suggested that higher antioxidant activity due to SNP rs4880 in SOD2 may lead to poorer survival after cyclophosphamide-containing breast cancer chemotherapy,19 and these results were partly confirmed by another study.20 Patients with alleles related to a higher antioxidant activity experienced significantly less grade 3–4 neutropenia but had poorer progression-free survival (PFS) than women with low-activity genotypes.20 The extracellular superoxide dismutase EC-SOD (SOD3) has Cu and Zn in the catalytic center and is highly expressed in blood vessels, heart, lungs, kidney, placenta and extracellular fluids.21 Of all three human SODs tested in pancreatic tumor xenograft nude mice, direct injections of the SOD3 expressing AdEcSOD vector had the greatest effect in inhibiting in vivo tumor growth and increasing survival.22 The most studied Gly allele in SOD3 SNP Arg231Gly (rs1799895) leads to 10-times higher concentration of SOD3 in plasma.23 The more frequent SOD3 SNPs (Ala58Thr, rs2536512 and −1257C>T, rs699473) have unknown functional effects. No report on association of SOD3 with prognosis of breast carcinomas has been published so far.

This study intended to evaluate a possible association between important modifiers of oxidative stress and routinely used prognostic factors in breast cancer. Functional SNPs in NQO1 (Pro187Ser, rs1800566), NQO2 (Phe47Leu, rs1143684), SOD2 (Val16Ala, rs4880) and unexplored SNPs in SOD3 (Ala58Thr, rs2536512; Arg231Gly, rs1799895 and −1257C>T, rs699473) as well as transcript levels in tumor tissues were followed in order to assess their potential use as biomarkers of breast cancer prognosis. The genotype–phenotype relationship was also addressed.

Material and Methods

Subjects

Blood samples were obtained from 321 incident breast cancer patients consecutively diagnosed in three faculty hospitals in Prague during the period between February 2000 and December 2006. From this cohort, 120 samples of mammary carcinomas and 51 paired adjacent control tissues without morphological signs of carcinoma were also collected. Tissue samples were collected during surgery and snap-frozen in liquid nitrogen. Cryostat sections were stained by hematoxylin and eosin and reviewed by a staff pathologist. Five-micrometer cryostat sections were prepared for isolation of total RNA. The presence of tumor cells in the sample was histologically verified in the first and in the last section of a row. The sections cut in the layers between the aforementioned histological controls were used for total RNA isolations. To increase the content of tumor cells in samples, the surrounding non-neoplastic tissue (connective and adipose tissue) was removed before sectioning. Samples of normal adjacent breast tissue were taken in areas remote from a grossly apparent tumor tissue in resection specimens. In the majority of cases, the distance from the tumor was greater than 10 mm. After deep freezing of this tissue, histological sections were evaluated microscopically to confirm that samples were tumor free. In some cases, excision from the tumor sample was available only in a native state (non fixed in standard neutral buffered formaldehyde). Some non-tumor samples were excluded because infiltration by tumor cells was found after microscopic evaluation of the frozen section. In some samples, the number of non-neoplastic cells in the adjacent tissue was not satisfactory for isolation of nucleic acid (only adipose or fibrous tissues were present in these tissue samples).

Histological classification of carcinomas was performed according to standard diagnostic procedures.24 The following data on patients were retrieved from medical records: age at diagnosis, menopausal status, personal history, family history (number of relatives affected by breast cancer, ovarian cancer or other malignant diseases), stage, tumor size, histological type and grade of the tumor, expression of estrogen and progesterone receptors and PFS. Expression of estrogen and progesterone receptors was assessed according to the published procedure25 with the commonly used 10% cut-off value. Disease progression was determined by the same technique in all the patients followed up, i.e., radiological assessment (computer tomography, abdominal ultrasonography, chest x-ray, mammography or bone scan), and clinically in case of skin metastases. Locoregional relapse was histologically confirmed. Tumor marker elevation without any clinical or radiological signs was not considered as disease progression. All patients were asked to read and sign informed consent in accordance with the requirements of the Ethical Commission of the National Institute of Public Health in Prague.

DNA isolation and genotyping

Blood samples were collected during the diagnostic procedures using tubes with K3EDTA anticoagulant. Genomic DNA was isolated with BioSprint 15 DNA Blood Kit (Qiagen, Hilden, Germany) using a KingFisher mL instrument and the protocol supplied by the manufacturer (Thermo Fisher Scientific Inc., Waltham, MA). SNPs in NQO1 (rs1800566), NQO2 (rs1143684), SOD2 (rs4880) and SOD3 (rs2536512, rs1799895 and rs699473) were determined by TaqMan Drug Metabolism Genotyping Assays (Assay ID: C_2091255_30 for NQO1; C_8774861_1 for NQO2; C_8709053_10 for SOD2; C_2668728_10, C_2307506_10 and C_2668727_1 for SOD3, respectively) obtained from Applied Biosystems (Foster City, CA) using real-time polymerase chain reaction (PCR) in a RotorGene 6000 instrument (Corbett Research, Sydney, Australia). Reaction mixtures for real-time PCR contained 5 μl of 2x TaqMan Genotyping Master Mix (kit no.4371355, Applied Biosystems), 0.25 μl of TaqMan Drug Metabolism Genotyping Assay and 4.75 μl of DNA template diluted to concentration of 0.7 ng/μl. Cycling parameters were initial denaturation at 95°C for 10 min, followed by 45 cycles consisting of denaturation at 95°C for 15 sec and annealing at 60°C for 60 sec for NQO1, NQO2, SOD2, rs2536512 and rs1799895 in SOD3 or 58°C for 60 sec for rs699473 in SOD3. The non-template control consisted of a reaction tube in which water was used in place of the DNA sample. Ten percent of randomly selected samples were reanalyzed with 100% concordance of results.

Preparation of standards for quantitative real-time PCR

Fragments spanning all coding sequences of NQO1 (825 bp), NQO2 (696 bp) and SOD2 (863 bp) and exons 1–2 (208 bp) of SOD3 and exons 3–5 (738 bp) of peptidylprolyl isomerase A (PPIA) were subcloned into vector pDONR201 using GatewayTM Cloning Technology (Invitrogen, Carlsbad, CA). The entry clones were propagated in Escherichia coli DH5α Maximum Efficiency Cells (Invitrogen) and isolated by Plasmid Midi Kit (Qiagen). Concentrations and purity of plasmids were determined spectrophotometrically. Sequences of primers and PCR conditions for cloning are available upon request.

Quantification of mRNA transcript levels

Total RNA was isolated from tissue sections using Trizol Reagent (Invitrogen) according to the procedure recommended by the manufacturer and stored at −80°C. The quality of RNA was verified by horizontal agarose gel electrophoresis. RNA quantity was assessed by UV spectrophotometry on a Cary 300 spectrophotometer (Varian, Palo Alto, CA). cDNA was synthesized using 0.5 μg of total RNA in 20 μl of final volume by help of RevertAidTM First Strand cDNA Synthesis Kit with random hexamer primers according to the manufacturer's protocol (MBI Fermentas, Vilnius, Lithuania) and stored at −80°C. The quality of cDNA in terms of DNA contamination was then confirmed by PCR amplification of ubiquitin C fragment discriminating between product from cDNA (190 bp) and genomic DNA (1,009 bp) as described previously.26 All cDNA samples that were free of DNA contamination (absence of 1,009 bp band in sample incubated without reverse transcriptase) were further analyzed. Subsequently, quantitative real-time PCR was performed in RotorGene 6000 (Corbett Research, Sydney, Australia) and TaqMan Gene Expression Assays were used (NQO1, kit no.Hs01045994_m1; NQO2, kit no.Hs01056948_m1; SOD2, kit no.Hs00167309_m1; SOD3, kit no.Hs00162090_m1; Human PPIA Endogenous Control, kit no. 4333763F). PPIA was used as a reference gene. The real-time PCR reaction mixtures contained 10 μl of 2x TaqMan Universal PCR Master Mix (kit no.4304437) for NQO1, NQO2 and PPIA or 10 μl of ABSolute QPCR ROX Mix (ABgene) for SOD2 and SOD3, 1.0 μl of FAM-labeled probe and primers set (TaqMan Gene Expression Assay specified above), 5 μl of diluted cDNA template and water to a final volume of 20 μl. Cycling parameters were initially held at 50°C for 2 min and initial denaturation at 95°C for 10 min, followed by 50 cycles consisting of denaturation at 95°C for 15 sec and annealing/extension at 58°C for 60 sec for PPIA, 56°C for 50 sec for NQO1 and NQO2 or 60°C for 60 sec for SOD2 and SOD3. Fluorescence was acquired after each extension step. The non-template control contained water instead of cDNA. Negative cDNA synthesis controls (RNA transcribed without reverse transcriptase) were also employed to reveal possible carry-over contamination. Each real-time PCR run contained serially diluted concentrations of standard plasmid DNA (NQO1, NQO2, SOD2 range 102–106 copies/reaction; SOD3, range 102–106 and 5.106 copies/reaction; PPIA, range 5.102–5.106 copies/reaction where 100 copies/reaction were set as limit of detection) for generation of the calibration curve. Transcript levels were analyzed using RotorGene v6 software (Corbett Research). Absolute quantification with external standard curves (efficiency above 90% and r2 > 0.99) generated from five decimal dilutions of the respective plasmid was performed. NQO1, NQO2, SOD2 and SOD3 transcript levels (target) were normalized to levels of PPIA as internal control (copiestarget/copiescontrol). Each sample was assessed in duplicate and the mean value was used for further analyses. Samples with more than 15% variation between duplicates were reanalyzed.

Statistical analyses

For statistical analyses, transcript levels of all established genes were logarithmically transformed to normalize value distribution. Transformed levels were then analyzed by parametric ANOVA test toward categorized values (genotype and categorical clinicopathological data on patients). Transcript levels of all target genes were correlated with size of the tumor or age at diagnosis using bivariate regression. Genotypes were evaluated as homozygote for ancestral allele vs. homozygote for nonancestral allele, homozygote for ancestral allele vs. heterozygote or homozygote for ancestral allele vs. nonancestral allele (dominant model) or homozygote for nonancestral allele vs. ancestral allele (recessive model). Associations between categorized values as genotype and clinicopathological data listed below were analyzed using two-sided Fisher's Exact test. The tested clinical and pathological variables were as follows: menopausal status (pre- vs. post- or perimenopausal), stage, lymph node metastasis (N0 vs. N1-3), histological type (invasive duct vs. other invasive carcinoma) and grade and expression of estrogen and progesterone receptors (negative vs. positive). The additive genetic model was assessed by logistic regression as well. PFS was calculated by the Kaplan–Meier method and the log-rank test was used for comparisons. PFS was defined as the time elapsed between surgical treatment and disease progression or death from any cause. Patients lost to follow-up and patients with stage IV disease (n = 2) were excluded from PFS analyses. p values are departures from two-sided tests. A p value of less than 0.05 was considered statistically significant. All statistical analyses were performed using SPSS v15.0 software (SPSS Inc., Chicago, IL). The study had 80% power to detect OR = 2.0 or higher at the 5% type I error rate when considering the lowest minor allele frequency of 17% (NQO1, rs1800566) for analyses of menopausal status, histological type and grade. The study had 80% power to detect OR = 1.8 or higher for analyses of stage, lymph node status and expression of hormonal receptors under the same considerations. QUANTO (version 1.2.4) software was used to calculate the statistical power of the study.

Results

Study population characteristics and SNPs

We collected blood samples from patients with breast carcinoma (n = 321). Clinicopathological data on all patients are presented in Table 1. SNPs in oxidative stress modifying genes were assessed in all available DNA samples (Table 2). With the exception of rs4880 in SOD2, the genotype distribution and nonancestral allele frequencies of the studied SNPs did not significantly differ from Caucasian populations (CEU sample) genotyped previously.27 However, the distribution of genotypes in SOD2 rs4880 in our study did not differ from a recently published Norwegian cohort of a similar size (Ref.19, n = 329). None of the followed SNPs deviated from Hardy-Weinberg equilibrium at p < 0.05 level of significance (data not shown). A total of 23.7% of all patients (n = 76) had a positive family history of breast and/or ovarian cancer. The genotype distribution in the group of patients with a positive family history did not differ from the group of patients with no history of breast and/or ovarian cancers (Pearson chi-square test, data not shown). The median and mean times of follow-up were 43 and 47 months (with the range of 3–110 months), respectively.

Table 1. Clinicopathological characteristics of patients involved in the study
inline image
Table 2. NQO1, NQO2, SOD2 and SOD3 genotype distributions
inline image

NQO1, NQO2, SOD2 and SOD3 transcript levels

Transcript levels were detected by real-time PCR in 120 samples of tumor tissues and 51 paired samples of adjacent non-neoplastic tissue (Fig. 1). Transcript levels of NQO2, SOD2 and SOD3 significantly differed between tumor and non-neoplastic tissues (Fig. 1). NQO2 and SOD2 were upregulated in the tumor tissue. The opposite was true for SOD3, i.e., strong downregulation in tumor tissue samples. The average transcript levels in tumors followed this trend: NQO1>NQO2>SOD2>SOD3. The following trend NQO1>SOD2>NQO2>SOD3 was observed in non-neoplastic tissues. Raw transcript levels did not show normal value distribution and therefore logarithmic transformation was performed prior to statistical analyses.

Figure 1.

NQO1, NQO2, SOD2 and SOD3 transcript levels in tumor (n = 120) and non-tumor (n = 51) tissues. Transcript levels were assessed by quantitative real-time PCR with absolute quantification as described in the Material and Methods section. Differences in logarithmically transformed transcript levels between tumor () and non-tumor (▭) samples were evaluated by two-sided paired T-test for 51 samples for which paired tissues were available. In one patient, the NQO1 and NQO2 transcript levels could not be determined.

No significant association between the studied SNPs and transcript levels was found by ANOVA test using logarithmically transformed transcript levels (results not shown).

Associations between transcript levels, SNPs and clinicopathological data

The functional SNP rs1143684 in NQO2 was significantly associated with disease stage and expression of progesterone receptors. Patients with stages 0 or I (less advanced disease) had a greater incidence of the NQO2 nonancestral Leu allele than those with more advanced stages II–IV (p = 0.033, Table 3), and this association was confirmed by the additive model (p = 0.031, Table 3). Patients with tumors expressing progesterone receptors had a greater incidence of the NQO2 nonancestral Leu allele than those without expression of progesterone receptors (p = 0.028, Table 3). However, analysis of the additive model did not confirm this association (p = 0.053, Table 3). Expression of progesterone receptors was significantly associated with stage (p = 0.032). Subgroup analyses were therefore performed. In the group of patients without expression of progesterone receptors, those with stages 0 or I had a significantly greater incidence of the NQO2 nonancestral Leu allele than those with stages II–IV (p = 0.029, additive model). No association was found in patients with expression of progesterone receptors. The missense SNP rs2536512 in SOD3 was significantly associated with expression of estrogen receptors. Patients with tumors expressing estrogen receptors carried the nonancestral Thr allele with significantly higher frequency than those without expression (p = 0.007, Table 3). This association was also significant in patients with tumors expressing both estrogen and progesterone receptors (p = 0.018, Table 3). Both associations were confirmed by the additive model (Table 3). No other association was observed in the analyses of SNPs in patients divided by clinicopathological data.

Table 3. Associations between SNPs and clinicopathological data
inline image

SOD3 transcript level was significantly higher in grade 1 or 2 tumors (mean ± SE, 13.4 ± 2.4 copies of SOD3/PPIA/μg RNA*1000, n = 77) than in grade 3 tumors (5.6 ± 1.9, n = 28) (ANOVA with logarithmic normalization, p = 0.006). No association between age at diagnosis, menopausal status, tumor size, histological type and SNPs or transcript levels was found (results not shown).

Progression-free survival analysis

For PFS analysis, patients were divided into groups according to their logarithmically transformed transcript levels. Groups were divided into 50th percentiles and quartiles. This grouping was then used for the stratification of PFS analysis. Patients were also divided according to regimens received (cyclophosphamide-containing regimens: yes vs. no; anthracycline-containing regimens: yes vs. no; hormonal regimens: yes vs. no; radiotherapy: yes vs. no). PFS of patients did not significantly differ according to the type of therapy (data not shown). Tumor type (invasive ductal vs. other) and grade did not significantly modify PFS. Lymph node status (p < 0.001), stage (p < 0.001) and expression of both receptors (p = 0.011) significantly modified the PFS of the whole group of patients. Transcript levels alone did not significantly modify PFS in either the whole patient group (n = 101) or in patients stratified according to their therapy (data not shown). SNP rs699473 in SOD3 significantly modified the PFS of all patients (n = 217, Fig. 2a, p = 0.038, log-rank test). Other studied SNPs did not significantly modify PFS (data not shown). No significant association of the studied SNPs with PFS in patients with node-positive (n = 88) or node-negative (n = 118) disease was observed (data not shown). No significant association of the studied SNPs with PFS in patients expressing both receptors in their tumors (n = 94) was found. In the group of patients without expression of receptors (n = 59), SNP rs4880 in SOD2 significantly modified the PFS in the dominant model (p = 0.001, Fig. 3a). In stratified analyses, SNPs did not significantly modify PFS in patients treated by anthracyclines, cyclophosphamide or by radiotherapy (data not shown). In the group of patients treated by hormonal regimens (containing tamoxifen or aromatase inhibitors), the PFS was significantly modified by SOD3 SNP rs699473 in the dominant model (n = 179, Fig. 2b, p = 0.021, log-rank test). In patients treated by both hormonal regimens and chemotherapy (n = 59), PFS was significantly modified by SOD3 SNP rs699473 in the recessive model (p = 0.026, log-rank test, data not shown). In patients treated by regimens containing cyclophosphamide without hormonal therapy (n = 30) the PFS was significantly modified by SOD2 rs4880 SNP in the dominant model (Fig. 3b, p = 0.004, log-rank test).

Figure 2.

(a) Association between PFS of all studied patients with breast carcinoma and SNP rs699473 in SOD3. Kaplan–Meier survival curves for patients carrying CC (solid line, n = 30), CT (dashed line, n = 106) and TT (dotted line, n = 80) genotypes (p = 0.038, log-rank test). (b) Association between PFS of patients with breast carcinoma and SNP rs699473 in SOD3 in group of patients treated by hormonal regimens. Kaplan–Meier survival curves for patients carrying CC (

equation image

, n = 24) and CT or TT (…, n = 155) genotypes (p = 0.021, log-rank test).

Figure 3.

(a) Association between PFS of patients with breast carcinoma and SNP rs4880 in SOD2 in group of patients with negative expression of estrogen and progesterone receptors. Kaplan–Meier survival curves for patients carrying Val allele (

equation image

, n = 44) and Ala/Ala genotype (…, n = 15) (p = 0.001, log-rank test). (b) Association between PFS of patients with breast carcinoma and SNP rs4880 in SOD2 in the group of patients treated by cyclophosphamide-containing regimens. Kaplan–Meier survival curves for patients carrying Val allele (

equation image

, n = 22) and Ala/Ala genotype (…, n = 8) (p = 0.004, log-rank test). All patients in this group had negative expression of estrogen and progesterone receptors and did not receive hormonal therapy.

Discussion

In humans, oxidative stress is involved in many diseases, e.g., atherosclerosis, myocardial infarction or Parkinson's disease, but short-term oxidative stress may also be important in prevention of aging.28 Reactive oxygen species are beneficial for cell signaling and mediate inflammation used by the immune system as a way to attack and kill pathogens. Oxidative stress has a quite ambivalent role in cancer. Reactive oxygen species influence apoptosis,2 proliferation or tumor promotion.3 Anthracyclines and cyclophosphamide generate oxidative stress during cancer therapy.5, 29 Antioxidants and oxidative stress-modifying enzymes may thus interact with their antitumor effect. Although a number of studies have explored the associations between genes coding enzymes modifying oxidative stress and breast cancer risk (susceptibility), only a few of them have studied their effects on prognosis, survival or therapy outcome of patients with breast cancer. The majority of these studies was limited to following up a single gene or analysis of SNPs and haplotypes in a number of selected candidate genes.15, 19, 30, 31 Our study combined the assessment of functional SNPs and transcript levels of candidate genes (NQO1, NQO2, SOD2 and SOD3) modifying levels of reactive oxygen species with an analysis of the association of these molecular factors with major prognostic factors and PFS. For assessment of transcript levels we developed highly sensitive methods based on real-time PCR with absolute quantification which is superior to relative or semiquantitative approaches. This approach has already been used in our previous studies on the role of ABCB1 and NQO1 in the prognosis of breast cancer patients.16, 25

Transcript levels of NQO2 and SOD2 in this study were significantly upregulated in tumors of the mammary gland in comparison with non-neoplastic tissues. However, analysis of clinicopathological data did not reveal any significant association of NQO1, NQO2 or SOD2 transcript levels with prognostic markers or PFS. Thus, transcript levels of these genes do not appear to modify prognosis of breast cancer patients.

In our previous study addressing “knock-out” SNP and transcript and protein levels, we have concluded that the role of NQO1 in human mammary gland carcinogenesis does not seem to be directly associated with classical clinicopathological factors (Ref.16, n = 105). This result was confirmed by the lack of association of clinical data with both NQO1 transcript levels (n = 120) and the knock-out SNP rs1800566 (n = 321) in this study. NQO1 was shown to be a strong prognostic and predictive factor in breast cancer patients receiving anthracyclines, and interaction with TP53 status was reported.15 Our previous study did not reveal direct cosegregation between TP53 and NQO1 protein expression in breast tumors and non-neoplastic tissues.16 The present study failed to confirm the association between PFS and rs1800566 in patients treated by anthracyclines. However, it could be due to the considerably lower sample size in comparison with Fagerholm et al.15NQO1 transcript level also did not modify the PFS in any of the stratified analyses.

We hypothesized that NQO2 status may also contribute to the progression of breast cancer because stress-induced NQO1 and NQO2 transiently stabilize TP53, leading to protection against adverse effects of stressors.12 Carriers of the low activity allele Leu in NQO2 rs1143684 SNP in this study were at moderately lower risk of developing more advanced disease (stages II–IV) or disease with poor prognosis (tumors without expression of progesterone receptors) than carriers of the normal activity genotype Phe/Phe. The observed association was refined by subgroup analyses. Patients without expression of progesterone receptors had more frequent occurrence of stage 0 or I than stage II–IV disease in comparison with patients whose tumors expressed progesterone receptors. Carriers of the NQO2 Leu allele in the group of patients without expression of progesterone receptors were at significantly lower risk of developing more advanced disease. The above findings suggested a possible link between NQO2 and prognosis of breast cancer in the subgroup of patients without expression of progesterone receptors. Neither PFS analysis nor assessment of NQO2 transcript levels helped to explain this phenomenon. Our numerous experiments with detection of NQO2 protein expression by immunohistochemistry in formalin-fixed and paraffin-embedded tissues failed due to the high level of unspecific staining (unpublished data). Currently there are no commercially available antibodies suitable for immunohistochemistry of human NQO2 in formalin-fixed and paraffin-embedded tissues. Thus, we cannot add mechanistic evidence to support the observed association of NQO2 rs1143684 SNP with prognosis of breast cancer patients. Bearing in mind the issue of possible false positive results due to multiple testing, our present observations should be treated with caution. Negative results on association of rs1143684 SNP in NQO2 with breast cancer PFS comply with the recently published study.32

SOD2 function can be modulated by estrogen and TP53, and thus it may be highly relevant for breast cancer progression and treatment.33, 34 Patients with the high activity Ala allele in SOD2 rs4880 SNP treated by cyclophosphamide had poorer breast cancer survival than patients with the low activity genotype Val/Val.19, 20 In our study, SOD2 rs4880 SNP did not significantly modify the PFS of patients treated by cyclophosphamide (n = 89). However, after exclusion of patients who were also treated by hormonal regimens (i.e., tamoxifen or aromatase inhibitors), the effect of this SNP was revealed. Carriers of the high activity genotype Ala/Ala in SOD2 rs4880 SNP treated by regimens containing cyclophosphamide but not hormonal regimens had significantly poorer PFS than carriers of the low activity Val alleles (p = 0.004). This result was further confirmed by analysis of patients divided by expression of estrogen and progesterone receptors. In the group of patients without expression of receptors (i.e., candidates for chemotherapy) SNP rs4880 in SOD2 significantly modified the PFS in the same way as in the cyclophosphamide-treated group. Our observation is in agreement with the above cited studies. Due to the small group size (n = 30), our observation should be confirmed by meta-analysis together with already published studies. Contrary to the study based on immunohistochemically detected levels of SOD2 (MnSOD35), the transcript levels of SOD2 followed in our study did not associate with any of the analyzed prognostic factors. Due to various transcriptional and post-translational events, transcript levels do not necessarily represent protein level or activity. However, recently, Sgambato et al.36 also reported the lack of association between SOD2 protein levels evaluated by immunohistochemistry and breast cancer PFS and overall survival in a study similarly sized (n = 134) with ours. We did not find an association between SOD2 transcript levels and PFS. Because of the relatively short mean and median follow-up time of patients, the observed associations between SOD2 SNP and PFS could be modified by the progression-free surviving fraction of patients, particularly among the hormonally treated group. A larger controlled clinical trial is needed to ultimately answer the question of utility of SOD2 as a marker of prognosis of breast cancer patients.

The importance of SOD3 (formerly ECSOD) for prognosis of breast cancer patients has not been studied to date. Thus, our study brings completely novel results. A recent study using cell models of tumors of the mammary gland suggested that overexpression of SOD3 may be a promising strategy for enhancing the efficacy of heparin/LMWH (low molecular weight heparin) by inhibiting heparanase. This finding may theoretically have implications for treatment of breast cancer patients.37 We have observed the following trend in SOD3 expression: tumor grade 3 < tumor grade 1 or 2 < non-neoplastic tissue. Downregulation of the SOD3 expression in lung tumors was previously reported, and authors have suggested that it can have potential effects on extracellular regulation of multiple factors that regulate angiogenesis and invasion.38, 39 SOD3 also blunts oxidant-dependent VEGF expression and experiments with its overexpression in C57BL/6 mice carrying implanted melanoma B16-F1 tumor cells have suggested that tumor-related oxidative stress may facilitate tumor vascularization and thus promote tumor growth.40 Direct injections of the SOD3 expressing the AdEcSOD vector had the greatest effect (of all SODs tested) in inhibiting in vivo pancreatic tumor growth and increasing the survival of xenograft nude mice.22 Our data thus support the previously envisaged role of SOD3 in tumorigenesis. This effect seems to be generalized as lung tumors, melanoma and pancreatic and breast tumors have quite distinct etiologies and phenotypes. Patients with expression of estrogen (or both estrogen and progesterone) receptors in their tumors carried the nonancestral Thr allele in SOD3 rs2536512 SNP with significantly greater frequency than those without expression. Estrogen was shown to modulate SOD3 activity through mRNA stabilization in human circulating monocytes.41 However, the functional importance of this SNP and another studied SOD3 rs699473 SNP is unknown and there are no published studies on the association of these SNPs with the prognosis of patients with any type of carcinoma. Moreover, our study did not find any association between SOD3 transcript levels and PFS. A frequently studied SNP in SOD3, the rs1799895, is functional,23 but surprisingly we did not find any carriers of the nonancestral allele in the studied patient group (n = 321). Most interestingly, the SOD3 SNP rs699473 in intron 1 significantly modified the PFS of all patients in our study, and its effect became more apparent in the group of patients treated by hormonal regimens. These findings together with the revealed association of SOD3 transcript levels with grade suggest that SOD3 may be important for the prognosis of breast cancer patients, although further studies are needed to characterize its potential role.

No significant association between the followed transcript levels and the studied SNPs in either tumor or non-neoplastic tissues was found. It seems obvious that although functional SNPs were mostly followed (in NQO1, NQO2 and SOD2), either unknown haplotypes or methylation status influence the genotype–phenotype relationship more profoundly not only in tumors but also in tissues adjacent to the carcinoma (non-neoplastic tissues). It is the aim of our future studies to address the importance of these factors in the prognosis of breast cancer patients.

Study power considerations suggested that subgroup analyses might have underestimated some potentially important effects. Correction for multiple testing was not applied and therefore some of the observed associations might be false. Future studies should reevaluate the results presented here in larger independent sample sets.

In conclusion, our study shows that oxidative stress-modifying enzymes present a promising area of investigation of prognostic markers of breast carcinoma. Associations identified by us, however, are quite complex. This fact is well demonstrated by the highly important discovery of the hitherto little studied SOD3, as compared to the lack of evidence for the role of the frequently followed NQO1.

Ancillary