Chemotherapy response in patients with primary breast cancer is difficult to predict and the role of host genetic factors has not been thoroughly investigated. We hypothesized that polymorphisms in oxidative stress (OS)-related genes, including estrogen–quinone metabolizing enzymes NQO2 and GSTM1-5, may influence disease progression and treatment response. In this prospective observational study, nineteen polymorphisms tagging known variations in candidate genes were genotyped and analyzed in 806 patients with primary breast cancer. Three functional polymorphisms, which were shown to affect gene expression levels in experiments in vitro and ex vivo, modified the effect of chemotherapy on disease-free survival. There were significant interactions between chemotherapy and individual polymorphisms or combined genotypes (designated as genetic score). Patients harboring high genetic score had a 75% reduction in the hazard of disease progression compared with patients with low genetic score when no chemotherapy was administered (HR = 0.25, 95% CI: 0.10–0.63, P = 0.005); however, they received much less survival benefit from adjuvant chemotherapy compared with patients with low genetic score when chemotherapy was administered (HR = 4.60 for interaction, 95% CI: 1.63–13.3, P = 0.004). These findings were validated in another population (n = 339). In conclusion, germline polymorphisms in OS-related genes affect chemotherapy sensitivity in breast cancer patients. Although reduced OS levels might prevent breast cancer progression, they probably compromise the effectiveness of adjuvant chemotherapy. Our findings also indicate that host-related factors must be considered for individualized chemotherapy. Cancer Res; 72(2); 408–19. ©2011 AACR.

It is well established that estrogens and their metabolites play critical roles in breast carcinogenesis. One of the most important mechanisms involved in this process is oxidative stress (OS) generated by estrogen quinones (1). Cells have approaches to reduce redundant quinines or semiquinones, such as reducing them by quinone oxidoreductases (NQO) and clearing them by glutathione S-transferases (GST). Of note, the estrogen–quinone metabolizing enzymes are not only specifically involved in breast carcinogenesis but are also related to the detoxification of reactive oxygen species (ROS; ref. 1).

The NQO family consists of 2 members, NAD(P)H:quinone oxidoreductase (NQO1) and NRH:quinine oxidoreductase (NQO2), both of which catalyze the detoxification of quinines and protect against OS (2). NQO1 was recently identified as a strong prognostic and predictive factor in breast cancer (3). The NQO1-deficient phenotype is a defective anthracycline response. In contrast, the role of NQO2 in breast cancer prognosis, as well as in chemotherapy response, is still unclear. Besides the NQO family, GSTs have been shown to have important roles in protection against ROS by conjugating with glutathione. GST detoxification pathways are active in normal breast tissue and GSTM and GSTP are the predominant enzymes in the breast (4). The genes of the GSTM family are arranged in a tandem of 5′-GSTM4-M2-M1-M5-M3-3′ (5). GSTM1 is of particular interest because it possesses a null polymorphism that results in a complete absence of GSTM1 enzyme activity. Our previous studies concluded that genetic variants in NQO2 and GSTM1-5 are related to breast cancer risk to different extents (6, 7).

Thus far, genetic alterations in somatic tumor cells have been shown to be correlated with prognosis, but the effects of genetic variations are less well understood (8–10). We hypothesize that genes modifying susceptibility to breast cancer may also influence disease progression and treatment response; variations in estrogen–quinone metabolizing genes involved in OS are good candidates. Notably, the relationship between OS and breast cancer prognosis is somewhat complicated. On one hand, increasing evidence has implied that OS may facilitate tumor cell migration, invasion, and metastasis through multiple mechanisms, including activation of the MAPK–HSP27 pathway, activation of matrix metalloproteinases, and inhibition of anti-proteases (11). A recent study also uncovered a novel mechanism in which OS could affect tumor microenvironment and increase the migratory properties of stromal fibroblasts, which in turn potentiate breast cancer dissemination (12). On the other hand, most chemotherapy regimens exert their cytotoxic effects through apoptosis which is mainly mediated by ROS and concomitant OS (13, 14), indicating that increased OS levels might enhance the effectiveness of adjuvant chemotherapy. Therefore, the influence of variations in OS-related genes on breast cancer progression and prognosis might depend on whether a patient underwent chemotherapy or not.

To our knowledge, although adjuvant chemotherapy could effectively reduce the risk of recurrence and mortality for women with operable disease, only a fraction of patients benefit from it. It is difficult to predict the response of patients to chemotherapy. Many somatic factors such as nodal status, tumor size, and expression of hormone receptors and HER2 are routinely used to determine prognosis and response to specific therapies (15). However, the roles of host-related factors, for example, inherited genetic factors, have not been thoroughly investigated. In this study, we investigate whether women harboring genetic variations in estrogen–quinone metabolizing genes involved in OS experience different disease progression, and we explore the effect of these variations on chemotherapy response, which is determined by the survival after diagnosis of breast cancer in the adjuvant setting.

Patients

This study was approved by the Ethical Committee of the Shanghai Cancer Center of Fudan University, and each participant signed an informed consent document. This prospective observational study was initiated in 2004. All patients with malignant breast cancer tumors who were willing to participate in the study were enrolled. For each participant, clinicopathologic and treatment data were recorded, disease outcome was followed up, and a blood sample was collected. From January 2004 to January 2008, we recruited approximately 1,036 unrelated patients with pathologically confirmed primary breast cancer in the Shanghai Cancer Center. Genotyping of the NQO2 and GSTM1-5 genes was done in 2008–2009 (6, 7, 16). Patients selected for the present analysis fulfilled the following inclusion criteria: (i), female patients diagnosed with unilateral invasive breast cancer; breast carcinoma in situ (with or without microinvasion) were excluded; (ii), pathologic examination of tumor specimens was carried out in the Department of Pathology in our hospital; (iii), patients with operable tumor and without any evidence of metastasis at diagnosis; (iv), patients not receiving neoadjuvant systemic therapy (chemotherapy and/or hormone therapy) or preoperative irradiation; (v), patients harboring a rare 16-bp insertion allele of a triallelic polymorphism in the NQO2 gene were also excluded because of the indefinite function of that allele; (vi), with follow-up data for at least 3 months; (vii), not treated with anti-HER2 therapy, for example, trastuzumab.

Of the 1,036 unrelated patients who were originally enrolled in the prospective observational study, 38 had bilateral breast cancer, and 158 of the 998 unilateral patients were DCIS. In the 840 unilateral patients with invasive cancer, 34 patients were further excluded due to not fulfilling other inclusion criteria. As a result, 806 patients were included in this study as the test set. The preoperative evaluation and examination has been described elsewhere (17). The basic information of patients is shown in Table 1. Determination of estrogen receptor (ER), progesterone receptor (PR), and HER2 status was done by pathologists in the Department of Pathology in our hospital. Most of, but not all, patients with equal HER2 protein expression (immunohistochemistry 2+) were also selected to have a FISH test for HER2 gene amplification. This was done according to established procedures that have been described elsewhere (18, 19). Because the data for tumor grade were lacking in many cases, we did not include this variable in our analysis.

Table 1.

Characteristics of breast cancer patients in 2 sets

VariableTest set (n = 806, %)Validation set (n = 339, %)P
Age (continuous) Median (ranges) 50 y (23–87) 43 y (19–85) <0.001b 
Follow-up time Median (ranges) 52 mo (3–86) 50 mo (3–86) <0.001b 
Age ≤50 y 49.4 64.3 <0.001 
 >50 y 50.6 35.7  
 Missing data 0.0 0.0  
Menopausal status Premenopausal 56.6 71.4 <0.001 
 Postmenopausal 43.4 28.6  
 Missing data 0.0 0.0  
Lymph nodes Negative 56.5 56.9 0.612 
 Positive 41.9 39.5  
 Missing data 1.6 3.5  
Tumor size ≤2 cm 53.2 51.3 0.825 
 >2 cm 43.2 40.4  
 Missing data 3.6 8.3  
ER Negative 34.1 42.8 0.002 
 Positive 64.1 52.8  
 Missing data 1.7 4.4  
PR Negative 41.4 49.9 0.002 
 Positive 56.8 45.4  
 Missing data 1.7 4.7  
HER2/neu Negative 75.4 65.5 0.051 
 Positive 22.0 25.7  
 Missing data 2.6 8.8  
Chemotherapya No 28.0 27.1 0.756 
 Yes 72.0 72.9  
 Missing data 0.0 0.0  
Endocrine therapy No 30.8 35.4 0.034 
 Yes 67.5 57.8  
 Missing data 1.7 6.8  
VariableTest set (n = 806, %)Validation set (n = 339, %)P
Age (continuous) Median (ranges) 50 y (23–87) 43 y (19–85) <0.001b 
Follow-up time Median (ranges) 52 mo (3–86) 50 mo (3–86) <0.001b 
Age ≤50 y 49.4 64.3 <0.001 
 >50 y 50.6 35.7  
 Missing data 0.0 0.0  
Menopausal status Premenopausal 56.6 71.4 <0.001 
 Postmenopausal 43.4 28.6  
 Missing data 0.0 0.0  
Lymph nodes Negative 56.5 56.9 0.612 
 Positive 41.9 39.5  
 Missing data 1.6 3.5  
Tumor size ≤2 cm 53.2 51.3 0.825 
 >2 cm 43.2 40.4  
 Missing data 3.6 8.3  
ER Negative 34.1 42.8 0.002 
 Positive 64.1 52.8  
 Missing data 1.7 4.4  
PR Negative 41.4 49.9 0.002 
 Positive 56.8 45.4  
 Missing data 1.7 4.7  
HER2/neu Negative 75.4 65.5 0.051 
 Positive 22.0 25.7  
 Missing data 2.6 8.8  
Chemotherapya No 28.0 27.1 0.756 
 Yes 72.0 72.9  
 Missing data 0.0 0.0  
Endocrine therapy No 30.8 35.4 0.034 
 Yes 67.5 57.8  
 Missing data 1.7 6.8  

aApproximately 100% and 85% of the patients that were treated with chemotherapy received cyclophosphamide-containing and anthracycline-based regimens, respectively.

bCompared by student t test and other P values by Pearson's χ2 test.

Postoperative recurrence risk category was mainly determined according to St. Gallen consensus. The choice of chemotherapy depended on the risk category: patients with moderate recurrence risk underwent cyclophosphamide, doxorubicin/epirubicin, and 5-Fu (CAF) regimen; patients with low risk underwent cyclophosphamide, methotrexate, and 5-Fu (CMF) regimen, or AC regimen; and patients with high risk would receive taxane-containing regimens [AC followed by paclitaxel (P), or CAF followed by docetaxel (T), or TAC]. Approximately 100% and 85% of the patients that were treated with chemotherapy received cyclophosphamide-containing and anthracycline-based regimens, respectively. All of the patients with positive hormone receptor status were recommended to take tamoxifen or aromatase inhibitors (only for postmenopausal women) for 5 years. The strategy of systemic treatments was updated according to the St. Gallen consensus (15, 20, 21).

We also validated the significant polymorphisms from the first set of tests in an independent population with mainly familial/early-onset breast cancer cases (ref. 22; Table 1). Since 2000, the Shanghai Cancer Hospital has conducted a multicenter hospital-based gene mutation screening project to gain a full understanding of the contribution of germline mutations of high-penetrance genes to hereditary and early-onset breast cancer in the Han Chinese population (22). Approximately 600 cases coming from 5 medical centers in China had been collected between 2000 and 2008, and an additional 150 new cases were recruited in 2008–2009 in our hospital. The eligibility criteria have been described elsewhere. Using early-onset/familial breast cancer patients as a validation set is acceptable because that, first, although the etiology of sporadic and familial/early-onset breast cancers is probably different, the clinical administration of adjuvant chemotherapeutic regimen does not differ between these 2 diseases according to either NCCN breast cancer guideline or St. Gallen early breast cancer consensus (15); second, prognosis of breast cancer seems to be not affected by family history (23); third, evidence has shown that in patients with BCRA1/2-negative familial breast cancer (in this study, the familial patients were also non-BRCA1/2 carriers), the objective response rate of neoadjuvant chemotherapy is comparable with that of sporadic breast cancer patients (24, 25). We finished genotyping of NQO2 and GSTM1-5 in approximately 400 unrelated familial and early-onset cases from southeast China (mainly Shanghai City and its surrounding regions) in 2008–2009 (6, 7, 16). In this study, we selected 339 patients who fulfilled the inclusion criteria described above.

DNA/RNA preparation

Extraction, preservation of genomic DNA and mRNA, and general PCR were done as previously reported (22). We also collected 40 pairs of tissue samples including normal breast (>3 cm away from tumor), peritumor (1–2 cm away from tumor), and cancer tissue; each pair was collected from the same patient.

Selection of genetic variants and genotyping

Selection of genetic variations of NQO2 and GSTM1-5 has been described elsewhere (6, 7). The 19 polymorphisms analyzed are listed in Table 2. Detailed information of selection of genetic variants is shown in Supplementary Materials and Methods. Single-nucleotide polymorphisms (SNP) were genotyped on the 12-plex SNPstream platform (Beckman Coulter Inc.; ref. 26). The genotyping work was carried out by the Chinese National Human Genome Center (Shanghai), and the call rates varied from 93.2% to 99.6%. The genotyping of the GSTM1 gene-deletion variant and the I-29/D polymorphism have been reported elsewhere (6, 16). In the validation study, the SNP rs2071002 (+237A>C) was genotyped using a PCR-RFLP–based assay (6). The samples were assayed in a 96-well PCR plate with a positive control consisting of a DNA sample with known heterozygous genotype. Two research assistants (K-D.Y and L.F) independently examined the gel pictures and repeated the assays if they did not reach a consensus on the genotype. An adequate quantity of restriction enzyme was used to completely cleave PCR amplicons. Samples with inconsistent outcomes in 2 independent tests were directly sequenced. In addition, 10% of the samples were randomly selected for repeated RFLP analysis for both of the polymorphisms, and the results were 100% concordant.

Table 2.

Impact of genotypes of NQO2 and GSTM1-5 on disease prognosis

Overall (n = 806)No chemotherapy (n = 226)Chemotherapy (n = 580)
Genetic variantLocationGenotypenHRa (95% CI)PbHRc (95% CI)PbHRc (95% CI)Pb
PaPcPc
NQO2-rs2070999 Promoter GG 433 Reference 0.164 Reference 0.625 Reference 0.077 
  AG+AA 363 0.84 (0.57–1.23) 0.365 1.72 (0.74–4.01) 0.208 0.69 (0.44–1.08) 0.108 
NQO2-I-29/D Promoter I-29/I-29 525 Reference 0.447 Reference 0.024 Reference 0.644 
  I-29/D+D/D 264 0.87 (0.58–1.30) 0.499 0.34 (0.12–0.99) 0.049 1.06 (0.68–1.66) 0.794 
NQO2-rs2071002 5′-UTR AA 332 Reference 0.510 Reference 0.0018 Reference 0.308 
  AC+CC 434 0.82 (0.56–1.20) 0.308 0.24 (0.10–0.55) 0.0007 1.18 (0.75–1.85) 0.469 
NQO2-rs1143684 Exon, Phe47Leu TT 355 Reference 0.003 Reference 0.162 Reference 0.009 
  CT+CC 444 1.68 (1.14–2.47) 0.009 1.40 (0.62–3.18) 0.417 1.75 (1.12–2.74) 0.015 
NQO2-rs4149367 Exon Ser135Ser CC 540 Reference 0.968 Reference 0.060 Reference 0.377 
  CT+TT 259 0.94 (0.63–1.38) 0.740 0.44 (0.15–1.30) 0.139 1.10 (0.71–1.71) 0.662 
NQO2-rs1885298 Intron GG 609 Reference 0.419 Reference 0.507 Reference 0.479 
  GT+TT 189 0.86 (0.54–1.36) 0.520 0.51 (0.16–1.64) 0.261 0.94 (0.56–1.57) 0.805 
NQO2-rs9501910 Intron GG 318 Reference 0.014 Reference 0.088 Reference 0.066 
  CG+CC 433 1.63 (1.09–2.44) 0.016 1.73 (0.71–4.21) 0.230 1.58 (1.01–2.50) 0.047 
GSTM4-rs542370 Promoter TT 410 Reference 0.756 Reference 0.440 Reference 0.952 
  CT+CC 379 0.92 (0.63–1.33) 0.659 0.75 (0.33–1.70) 0.487 1.51 (0.97–2.35) 0.066 
GSTM4-rs1010167 5′-UTR GG 717 Reference 0.331 Reference 0.126 Reference 0.982 
  CG+CC 38 0.87 (0.32–2.41) 0.795 N.A. 0.982 1.16 (0.42–3.21) 0.776 
GSTM4-rs560018 Intron TT 726 Reference 0.048 Reference 0.412 Reference 0.064 
  CT+CC 73 1.93 (1.13–3.32) 0.017 1.83 (0.61–5.53) 0.284 2.19 (1.18–4.04) 0.013 
GSTM4-rs535537 Intron GG 743 Reference 0.058 Reference 0.158 Reference 0.156 
  AG+AA 59 0.41 (0.13–1.28) 0.125 N.A. 0.977 0.51 (0.16–1.61) 0.247 
GSTM2-rs655315 Intron GG 418 Reference 0.731 Reference 0.814 Reference 0.844 
  AG+AA 375 0.86 (0.59–1.25) 0.439 0.60 (0.27–1.34) 0.215 0.97 (0.63–1.47) 0.880 
GSTM1-Null/Present Whole gene Null 456 Reference 0.023 Reference 0.009 Reference 0.285 
  Present 343 0.60 (0.41–0.88) 0.009 0.33 (0.14–0.75) 0.008 0.79 (0.51–1.22) 0.282 
GSTM5-rs3754446 Promoter CC 385 Reference 0.196 Reference 0.457 Reference 0.338 
  AC+AA 409 0.67 (0.46–0.98) 0.039 0.35 (0.14–0.85) 0.020 0.75 (0.49–1.15) 0.184 
GSTM5-rs4970773 Intron GG 344 Reference 0.217 Reference 0.797 Reference 0.212 
  CG+CC 421 0.69 (0.47–1.02) 0.060 0.45 (0.18–1.12) 0.085 0.70 (0.45–1.08) 0.109 
GSTM5-rs11807 3′-UTR TT 589 Reference 0.892 Reference 0.544 Reference 0.527 
  CT+CC 209 0.83 (0.54–1.29) 0.426 0.98 (0.40–2.44) 0.971 0.78 (0.47–1.28) 0.320 
GSTM3-rs7483 Exon, Val224Ile TT 462 Reference 0.273 Reference 0.669 Reference 0.401 
  CT+CC 337 0.80 (0.55–1.18) 0.258 0.58 (0.26–1.27) 0.174 0.84 (0.54–1.32) 0.455 
GSTM3-rs1332018 5′-UTR AA 555 Reference 0.062 Reference 0.158 Reference 0.255 
  AC+CC 221 0.75 (0.48–1.19) 0.226 0.65 (0.27–1.58) 0.346 0.74 (0.43–1.27) 0.282 
GSTM3-rs4970737 Promoter GG 374 Reference 0.929 Reference 0.791 Reference 0.873 
  CG+CC 376 0.88 (0.59–1.30) 0.508 0.62 (0.26–1.48) 0.285 0.93 (0.59–1.45) 0.740 
Overall (n = 806)No chemotherapy (n = 226)Chemotherapy (n = 580)
Genetic variantLocationGenotypenHRa (95% CI)PbHRc (95% CI)PbHRc (95% CI)Pb
PaPcPc
NQO2-rs2070999 Promoter GG 433 Reference 0.164 Reference 0.625 Reference 0.077 
  AG+AA 363 0.84 (0.57–1.23) 0.365 1.72 (0.74–4.01) 0.208 0.69 (0.44–1.08) 0.108 
NQO2-I-29/D Promoter I-29/I-29 525 Reference 0.447 Reference 0.024 Reference 0.644 
  I-29/D+D/D 264 0.87 (0.58–1.30) 0.499 0.34 (0.12–0.99) 0.049 1.06 (0.68–1.66) 0.794 
NQO2-rs2071002 5′-UTR AA 332 Reference 0.510 Reference 0.0018 Reference 0.308 
  AC+CC 434 0.82 (0.56–1.20) 0.308 0.24 (0.10–0.55) 0.0007 1.18 (0.75–1.85) 0.469 
NQO2-rs1143684 Exon, Phe47Leu TT 355 Reference 0.003 Reference 0.162 Reference 0.009 
  CT+CC 444 1.68 (1.14–2.47) 0.009 1.40 (0.62–3.18) 0.417 1.75 (1.12–2.74) 0.015 
NQO2-rs4149367 Exon Ser135Ser CC 540 Reference 0.968 Reference 0.060 Reference 0.377 
  CT+TT 259 0.94 (0.63–1.38) 0.740 0.44 (0.15–1.30) 0.139 1.10 (0.71–1.71) 0.662 
NQO2-rs1885298 Intron GG 609 Reference 0.419 Reference 0.507 Reference 0.479 
  GT+TT 189 0.86 (0.54–1.36) 0.520 0.51 (0.16–1.64) 0.261 0.94 (0.56–1.57) 0.805 
NQO2-rs9501910 Intron GG 318 Reference 0.014 Reference 0.088 Reference 0.066 
  CG+CC 433 1.63 (1.09–2.44) 0.016 1.73 (0.71–4.21) 0.230 1.58 (1.01–2.50) 0.047 
GSTM4-rs542370 Promoter TT 410 Reference 0.756 Reference 0.440 Reference 0.952 
  CT+CC 379 0.92 (0.63–1.33) 0.659 0.75 (0.33–1.70) 0.487 1.51 (0.97–2.35) 0.066 
GSTM4-rs1010167 5′-UTR GG 717 Reference 0.331 Reference 0.126 Reference 0.982 
  CG+CC 38 0.87 (0.32–2.41) 0.795 N.A. 0.982 1.16 (0.42–3.21) 0.776 
GSTM4-rs560018 Intron TT 726 Reference 0.048 Reference 0.412 Reference 0.064 
  CT+CC 73 1.93 (1.13–3.32) 0.017 1.83 (0.61–5.53) 0.284 2.19 (1.18–4.04) 0.013 
GSTM4-rs535537 Intron GG 743 Reference 0.058 Reference 0.158 Reference 0.156 
  AG+AA 59 0.41 (0.13–1.28) 0.125 N.A. 0.977 0.51 (0.16–1.61) 0.247 
GSTM2-rs655315 Intron GG 418 Reference 0.731 Reference 0.814 Reference 0.844 
  AG+AA 375 0.86 (0.59–1.25) 0.439 0.60 (0.27–1.34) 0.215 0.97 (0.63–1.47) 0.880 
GSTM1-Null/Present Whole gene Null 456 Reference 0.023 Reference 0.009 Reference 0.285 
  Present 343 0.60 (0.41–0.88) 0.009 0.33 (0.14–0.75) 0.008 0.79 (0.51–1.22) 0.282 
GSTM5-rs3754446 Promoter CC 385 Reference 0.196 Reference 0.457 Reference 0.338 
  AC+AA 409 0.67 (0.46–0.98) 0.039 0.35 (0.14–0.85) 0.020 0.75 (0.49–1.15) 0.184 
GSTM5-rs4970773 Intron GG 344 Reference 0.217 Reference 0.797 Reference 0.212 
  CG+CC 421 0.69 (0.47–1.02) 0.060 0.45 (0.18–1.12) 0.085 0.70 (0.45–1.08) 0.109 
GSTM5-rs11807 3′-UTR TT 589 Reference 0.892 Reference 0.544 Reference 0.527 
  CT+CC 209 0.83 (0.54–1.29) 0.426 0.98 (0.40–2.44) 0.971 0.78 (0.47–1.28) 0.320 
GSTM3-rs7483 Exon, Val224Ile TT 462 Reference 0.273 Reference 0.669 Reference 0.401 
  CT+CC 337 0.80 (0.55–1.18) 0.258 0.58 (0.26–1.27) 0.174 0.84 (0.54–1.32) 0.455 
GSTM3-rs1332018 5′-UTR AA 555 Reference 0.062 Reference 0.158 Reference 0.255 
  AC+CC 221 0.75 (0.48–1.19) 0.226 0.65 (0.27–1.58) 0.346 0.74 (0.43–1.27) 0.282 
GSTM3-rs4970737 Promoter GG 374 Reference 0.929 Reference 0.791 Reference 0.873 
  CG+CC 376 0.88 (0.59–1.30) 0.508 0.62 (0.26–1.48) 0.285 0.93 (0.59–1.45) 0.740 

Abbreviations: N.S., not significant; UTR, untranslated region; N.A., not applicable.

Pa and Pc: multivariate adjusted by the Cox risk proportion model.

aAdjusted for age (y), lymph node status (positive or negative), tumor size (≤2 cm or >2 cm), ER (positive or negative), PR (positive or negative), HER2 (positive or negative), chemotherapy (yes or no), and endocrine therapy (yes or no). HR with its 95% CI is calculated by the Cox risk proportion model.

bP: unadjusted.

cAdjusted for age, lymph node status, tumor size, ER, PR, HER2, and endocrine therapy.

Plasmid constructs, cell culture, transient transfection, and luciferase assays

The pGL3-Basic reporter vector from Promega was used to construct luciferase reporter plasmids using standard recombination techniques, as previously described (6). Briefly, promoter regions of NQO2 (−537 to +529 bp, the transcriptional start site is designated as +1) containing haplotype I-C and D-C were cloned from individual DNA samples and were then cloned into the pGL3-Basic vector to generate pGL3-I-C and pGL3-D-C constructs, respectively. Wang and colleagues have indicated that approximately −500 to +300 bp of NQO2 promoter region is sufficient for basal expression of NQO2 (27). A site-directed mutagenesis kit (Stratagene) was used to generate pGL3-I-A and pGL3-D-A plasmids. The abbreviations I, D, C, and A in above pGL3 constructs denote 29-bp insertion allele of I-29/D polymorphism, 29-bp deletion allele of I-29/D polymorphism, +237C allele of rs2071002, and +237A allele of rs2071002, respectively. A human Sp1 expression vector was constructed using the pcDNA 3.1 Directional TOPO Expression Kit (Invitrogen). All constructs were verified by direct sequencing before use.

A human immortal normal breast epithelial cell line (HBL-100) was obtained from the American Type Culture Collection (ATCC). Liquid nitrogen stocks were made upon receipt and maintained until the start of each study. Morphology and doubling times were also recorded regularly to ensure maintenance of phenotypes. Cells were used for no more than 3 months after being thawed. Thus, the cell line has been tested and authenticated by ATCC and maintained in our laboratory for less than 3 months, during which all experiments were conducted. Cells were grown in complete medium consisting of Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal calf serum in a humidified, 5% CO2 incubator at 37°C. For the luciferase experiments, cells were transfected with 500 ng of plasmid DNA (4 haplotype vectors or pGL3-Basic as a negative control) and cotransfected with 10 ng of pRL-SV40 as a control for transfection efficiency, with or without cotransfection of 1.0 μg of the Sp1 expression vector. Transfections were done using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. Luciferase activity was measured on a VeritasTM Microplate luminometer (Turner BioSystems) using the Dual-Luciferase Reporter Assay System Kit (Promega). Each experiment was conducted in triplicate at least 3 times. Luciferase units were calculated using the formula Firefly luciferase units/Renilla luciferase units. Fold increase was reported by defining the activity of the empty pGL3-Basic vector as one.

Real-time PCR

Real-time PCR with SYBR Green fluorescent-based assay (TaKaRa) was done in a fluorescence temperature cycler (Opticon, MJ Research) using the standard curves method (28). All samples were done in triplicate at least 3 times. cDNA-specific primers were designed using Primer Premier 5.00 (Premier Biosoft International). The following primers were used for gene expression detection: NQO2 sense, 5′-GAAACCCACGAAGCCTACA-3′, and antisense, 5′-CAGCACCCTATCCATCCAG-3′ (153 bp); glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used for normalization.

Survival analysis, prognosis modeling, and receiver operating characteristics curve

The categories analyzed for disease-free survival (DFS) were the first recurrence of disease at a local, regional, or distant site; the diagnosis of contralateral breast cancer; and death from any cause. All of these categories listed above were considered DFS events. Patients with study end date and loss of follow-up were considered to be censored. Survival curves were determined using the Kaplan–Meier method and compared by the log-rank test (univariate analysis). HR for disease progression and 95% CIs were calculated by the Cox risk proportion model. Multivariate analysis was carried out using the Cox risk proportion model (method: backward stepwise, likelihood ratio). Timescale of follow-up time used for Cox proportional hazards model is “month.”

We used multivariate logistic regression to construct the prediction model for DFS events. The aim of the model was to predict the risk of occurrence of disease events of an individual woman using individual clinicopathologic data, with or without personal genetic information. For a feasible modeling procedure, we divided the patients into 2 groups. One group experienced relapse during the follow-up period, whereas the other group did not. All of the cases selected for modeling should be followed up for at least 6 months, and a few cases with insufficient follow-up time were thus excluded from the modeling analysis. Because we observed an interaction between chemotherapy and genotypes of estrogen–quinone metabolizing genes, we carried out modeling using independent variables [age (years), lymph node status (positive or negative), tumor size (≤2 cm or >2 cm), ER (positive or negative), PR (positive or negative), HER2 (positive or negative), and endocrine therapy (yes or no), with or without genetic factor] in the nonchemotherapy group. The probability of disease progression was estimated by the formula “eL/(1+eL)”, in which the value of L was derived by multivariate logistic regression analysis (method: backward stepwise, likelihood ratio).

To further evaluate the accuracy of the prediction model, we employed receiver operating characteristics (ROC) curves and calculated the area under the curve (AUC) with its 95%CIs. The ROC curve shows the relation between sensitivity and false-positive rate (1-specificity) of a given test across all possible threshold values that define the positivity of a disease or condition. In ROC analysis, the independent variable was disease outcome (occurrence of disease events or not), and the classification variable is probability of disease progression, which was calculated using the formula eL/(1 + eL).

L for the traditional model is:

formula

L for the combined model is:

formula

Statistical analysis

Tests of association were conducted using Pearson χ2 test. Student t test was used to compare continuous variables between 2 groups. A P value of less than 0.05 (2-sided) was considered to be significant. Statistical analysis was done using Stata/SE 10.0 (Stata) and SPSS 12.0 (SPSS).

Association of candidate gene genotypes with DFS

We studied the association of the genotypes of 19 genetic variants with DFS in the dominant model (major homozygous vs. heterozygous + minor homozygous; Table 2). Analyses of a few variants were unavailable in the recessive model (major homozygous + heterozygous vs. minor homozygous) or in the additive model because of rare numbers of minor homozygous (data not shown). In the overall population, 4 polymorphisms (2 in NQO2, 1 in GSTM4, and the GSTM1-Null/Present polymorphism) showed significant associations with DFS in univariate analysis. All of these polymorphisms were still related to DFS after adjustment for clinicopathologic factors. However, after conservative Bonferroni correction of multiple comparisons, none of the polymorphisms remained significant.

Association of genotype with DFS is modified by adjuvant chemotherapy

We further investigated the effect of chemotherapy on the association between genotypes and DFS (Table 2). Interestingly, in patients not treated with chemotherapy, genotypes with minor alleles of the 3 polymorphisms, NQO2-I-29/D (Fig. 1A), NQO2-rs2071002 (Fig. 1B), and GSTM1-Null/Present (Fig. 1C), showed significantly better DFS than their major counterparts, although these associations were not observed in the chemotherapy group (Supplementary Fig. S1A–S1C). In contrast, the variant genotype of NQO2-rs1143684 was significantly correlated with poorer DFS compared with the wild-type genotypes in the chemotherapy group, but this effect was not observed in the nonchemotherapy group. After multivariate adjustment, all of the 4 polymorphisms still reached significant P values.

Figure 1.

Effects of genetic variants on DFS according to adjuvant chemotherapy in primary breast cancer without adjuvant chemotherapy for NQO2-I-29/D (A), rs2071002 (B), and GSTM1-null/present (C). P value tested by log-rank test.

Figure 1.

Effects of genetic variants on DFS according to adjuvant chemotherapy in primary breast cancer without adjuvant chemotherapy for NQO2-I-29/D (A), rs2071002 (B), and GSTM1-null/present (C). P value tested by log-rank test.

Close modal

Interaction between genetic variations and resistance to chemotherapy

The results presented above strongly suggest the presence of an interaction. Multivariate analysis of interaction was done in 2 steps. In the first step, the Cox regression model included established prognostic factors but not genotypes (see details in note of Table 3). We identified that positive lymph node status (HR = 2.85, 95% CI: 1.83–4.46, P < 0.0001), large tumor size (HR = 2.36, 95% CI: 1.56–3.57, P < 0.0001), positive ER (HR = 0.40, 95% CI: 0.27–0.59, P < 0.0001), positive HER2 (HR = 1.76, 95%: CI: 1.20–2.58, P = 0.004), and using chemotherapy (HR = 0.40, 95% CI: 0.24–0.65, P = 0.0005) were significant independent factors for DFS after multivariate adjustment. In the second step, interaction between each genetic variation and chemotherapy was investigated, along with adjustment for the factors identified in the first step (factors with P < 0.10). The interaction between genotypes and chemotherapy had strong impacts on DFS (Table 3). For instance, an interaction implies that patients with the variant allele of NQO2-rs2071002 received only 25% (1/4.02, 4.02 is the HR of interaction between chemotherapy and rs2071002 genotype) of the benefit from adjuvant chemotherapy compared with patients with the wild-type genotypes. Interestingly, NQO2-I-29/D, NQO2-rs2071002, and GSTM1-Null/Present not only interact with chemotherapy but tend to be independent prognostic factors if their interactions to chemotherapy are adjusted.

Table 3.

Multivariate Cox model (DFS) including interaction of genotypes with adjuvant chemotherapy

Genetic variantFactorsaPHR (95% CI)
NQO2-I-29/D Genotype (AA vs. Aa+aa) 0.062 0.38 (0.14–1.09) 
 Chemo. (no vs. yes) <0.001 0.30 (0.17–0.51) 
 Interaction, Chemo*Genotype 0.073 2.87 (0.91–9.06) 
NQO2-rs2071002 Genotype (AA vs. Aa+aa) 0.003 0.30 (0.14–0.66) 
 Chemo. (no vs. yes) <0.001 0.20 (0.10–0.37) 
 Interaction, Chemo*Genotype 0.003 4.02 (1.61–10.0) 
NQO2-rs1143684 Genotype (AA vs. Aa+aa) 0.670 1.21 (0.49–2.98) 
 Chemo. (no vs. yes) <0.001 0.40 (0.24–0.67) 
 Interaction, Chemo*Genotype 0.009 1.68 (1.14–2.48) 
NQO2-rs9501910 Genotype (AA vs. Aa+aa) 0.023 1.59 (1.07–2.37) 
 Chemo. (No vs. Yes) 0.002 0.42 (0.25–0.72) 
 Interaction, Chemo*Genotype 0.915 0.95 (0.36–2.49) 
GSTM1-Null/present Genotype (AA vs. Aa+aa) 0.001 0.25 (0.10–0.56) 
 Chemo. (no vs. yes) <0.001 0.21 (0.11–0.40) 
 Interaction, Chemo*Genotype 0.015 3.38 (1.27–8.97) 
Genetic score Genotype (0–1 vs. 2–3) 0.005 0.25 (0.10–0.63) 
 Chemo. (no vs. yes) <0.001 0.22 (0.12–0.40) 
 Interaction, Chemo*Genotype 0.004 4.60 (1.63–13.3) 
Genetic variantFactorsaPHR (95% CI)
NQO2-I-29/D Genotype (AA vs. Aa+aa) 0.062 0.38 (0.14–1.09) 
 Chemo. (no vs. yes) <0.001 0.30 (0.17–0.51) 
 Interaction, Chemo*Genotype 0.073 2.87 (0.91–9.06) 
NQO2-rs2071002 Genotype (AA vs. Aa+aa) 0.003 0.30 (0.14–0.66) 
 Chemo. (no vs. yes) <0.001 0.20 (0.10–0.37) 
 Interaction, Chemo*Genotype 0.003 4.02 (1.61–10.0) 
NQO2-rs1143684 Genotype (AA vs. Aa+aa) 0.670 1.21 (0.49–2.98) 
 Chemo. (no vs. yes) <0.001 0.40 (0.24–0.67) 
 Interaction, Chemo*Genotype 0.009 1.68 (1.14–2.48) 
NQO2-rs9501910 Genotype (AA vs. Aa+aa) 0.023 1.59 (1.07–2.37) 
 Chemo. (No vs. Yes) 0.002 0.42 (0.25–0.72) 
 Interaction, Chemo*Genotype 0.915 0.95 (0.36–2.49) 
GSTM1-Null/present Genotype (AA vs. Aa+aa) 0.001 0.25 (0.10–0.56) 
 Chemo. (no vs. yes) <0.001 0.21 (0.11–0.40) 
 Interaction, Chemo*Genotype 0.015 3.38 (1.27–8.97) 
Genetic score Genotype (0–1 vs. 2–3) 0.005 0.25 (0.10–0.63) 
 Chemo. (no vs. yes) <0.001 0.22 (0.12–0.40) 
 Interaction, Chemo*Genotype 0.004 4.60 (1.63–13.3) 

NOTE: Multivariate analysis of interaction was done in 2 steps. In the first step, the Cox regression model included established prognostic factors (age, lymph node status, tumor size, ER, PR, HER2, chemotherapy, and endocrine therapy) but not genotypes. The first step identified that lymph node status (P < 0.0001), tumor size (P < 0.0001), ER (P < 0.0001), HER2 (P = 0.004), and chemotherapy (P = 0.0005) were significant independent factors for DFS. As well, age tends to be significant (P = 0.08). In the second step, interactions between each genetic variant (in the dominant model) and chemotherapy were investigated along with adjustment for those factors (with P < 0.10) identified in the first step. This table shows the results of the second step.

Abbreviations: Chemo., chemotherapy; AA, major homozygous; Aa, heterozygous; aa, minor homozygous.

aHere, we present only 3 items: genotype (AA vs. Aa+aa), chemotherapy, and the interaction between them. Other parameters (tumor size, lymph node status, ER, and HER2 status) are not shown.

Relationship between host genotypes and ER phenotype

It is well established that ER-positive tumors tend to be resistant to chemotherapy (29). It is possible that the observed interaction between chemotherapy resistance and genetic variants is caused by a linkage of genetic variants to tumor ER phenotype. Therefore, we studied the relationship between genotype and tumor phenotype (Supplementary Table S1). However, no significant genotype–phenotype association was found for the 4 significant polymorphisms.

Functional polymorphisms alter expression of NQO2

We have shown that genetic variations in NQO2 and GSTM1 affect the resistance to chemotherapy. This is biologically plausible for GSTM1-Null/Present because the GSTM1-null genotype results in GSTM1 enzyme deficiency. However, the functional basis of multiple variations in NQO2 requires further inspection. Among the 4 crude significant polymorphisms, either in the overall population or in subgroups (Table 2), two (NQO2-I-29/D, NQO2-rs2071002) located in the promoter can affect gene expression, and one located in an exon results in an amino acid change (NQO2-rs1143684). The remaining SNP NQO2-rs9501910, which is located in an intron, has no potential function. However, it has high linkage disequilibrium (LD) to NQO2-rs1143684 with a D' of 0.88 and r2 of 0.77 (Fig. 2A), indicating that it is a linked marker rather than a causal SNP. We previously reported that the I-allele of NQO2-I-29/D introduces binding sites for the transcriptional repressor Sp3, and the C-allele of NQO2-rs2071002 creates a new Sp1 binding site (ref. 6; Fig. 2B). Here, we further show that the D-allele of NQO2-I-29/D and the C-allele of NQO2-rs2071002 are associated with higher activity of the NQO2 gene promoter in human normal breast cells (Fig. 2C) and lead to higher expression of NQO2 in normal breast and peritumoral tissues (Fig. 2D) compared with their wild-type counterparts. However, NQO2 genotype and NQO2 expression are not correlated in cancer tissue (Fig. 2D).

Figure 2.

Genetic variants in NQO2 and functional investigation. A, pairwise linkage disequilibrium (LD) among selected variants in the NQO2 gene in breast cancer patients. The value within each diamond represents the pairwise correlation between polymorphisms [measured as D' (left) and r2 (right)] defined by the top left and the top right sides of the diamond. The red-to-white gradient reflects higher to lower D' values; the black-to-white gradient reflects higher to lower r2 values. Haploview 4.1 software is used to draw the LD plot. B, schematic graphs of luciferase reporter plasmids. NQO2 promoter regions (−537 to +529 bp, the transcriptional start site is designated as +1) containing both I-29/D and rs2071002 (+237A>C) polymorphisms are cloned into a pGL3-Basic reporter gene. The I-29 allele of I-29/D introduces transcriptional–repressor Sp3 binding sites, whereas the A allele of rs2071002 abolishes the binding site of transcriptional–activator Sp1. C, promoter activity in normal HBL-100 breast cells in vitro. Arbitrary units denote fold increase (pGL3-Basic vector is set as one). Significantly low promoter activity is observed in the I-A haplotype compared with the other 3 haplotype constructs (all P < 0.05). After transfection of Sp1 expression vector, the I-A haplotype shows a lower activity compared with I-C (P < 0.01), D-A (P < 0.05), and D-C (P < 0.0001). Data represent mean values, with error bars showing SE. Comparisons are done by ANOVA analysis and adjusted by the Games–Howell method if equal variances are not assumed. D, box plots of NQO2 mRNA expression according to genotypes and tissue types. We collected 40 pairs of tissue samples including normal breast (>3 cm away from tumor), peritumor (1–2 cm away from tumor), and cancer tissue from the same patient. The 40 pairs specimens are divided into 3 groups according to germline genotypes: group I, both major homozygous (II and AA, n = 15); group III, at least one minor homozygous (DD or CC, n = 7); group II, the remaining genotypes (n = 18). In normal tissue and peritumor tissue, the 3 genotype groups have differential expression of NQO2 (P = 0.0001 and P = 0.0003, respectively, by Kruskal–Wallis test); in cancer tissue, genotypes have no effect on gene differential expression (P = 0.752 by Kruskal–Wallis test). GAPDH is used for normalization.

Figure 2.

Genetic variants in NQO2 and functional investigation. A, pairwise linkage disequilibrium (LD) among selected variants in the NQO2 gene in breast cancer patients. The value within each diamond represents the pairwise correlation between polymorphisms [measured as D' (left) and r2 (right)] defined by the top left and the top right sides of the diamond. The red-to-white gradient reflects higher to lower D' values; the black-to-white gradient reflects higher to lower r2 values. Haploview 4.1 software is used to draw the LD plot. B, schematic graphs of luciferase reporter plasmids. NQO2 promoter regions (−537 to +529 bp, the transcriptional start site is designated as +1) containing both I-29/D and rs2071002 (+237A>C) polymorphisms are cloned into a pGL3-Basic reporter gene. The I-29 allele of I-29/D introduces transcriptional–repressor Sp3 binding sites, whereas the A allele of rs2071002 abolishes the binding site of transcriptional–activator Sp1. C, promoter activity in normal HBL-100 breast cells in vitro. Arbitrary units denote fold increase (pGL3-Basic vector is set as one). Significantly low promoter activity is observed in the I-A haplotype compared with the other 3 haplotype constructs (all P < 0.05). After transfection of Sp1 expression vector, the I-A haplotype shows a lower activity compared with I-C (P < 0.01), D-A (P < 0.05), and D-C (P < 0.0001). Data represent mean values, with error bars showing SE. Comparisons are done by ANOVA analysis and adjusted by the Games–Howell method if equal variances are not assumed. D, box plots of NQO2 mRNA expression according to genotypes and tissue types. We collected 40 pairs of tissue samples including normal breast (>3 cm away from tumor), peritumor (1–2 cm away from tumor), and cancer tissue from the same patient. The 40 pairs specimens are divided into 3 groups according to germline genotypes: group I, both major homozygous (II and AA, n = 15); group III, at least one minor homozygous (DD or CC, n = 7); group II, the remaining genotypes (n = 18). In normal tissue and peritumor tissue, the 3 genotype groups have differential expression of NQO2 (P = 0.0001 and P = 0.0003, respectively, by Kruskal–Wallis test); in cancer tissue, genotypes have no effect on gene differential expression (P = 0.752 by Kruskal–Wallis test). GAPDH is used for normalization.

Close modal

Validation of interaction between genotypes and resistance to chemotherapy

It is critical to validate our findings that NQO2-I-29/D, NQO2-rs2071002, and GSTM1-Null/Present are associated with disease progression in the nonchemotherapy group. To validate the integral effect of multiple genetic variations, we developed a combined “genetic score” by assigning “0” to risk genotypes and “1” to protective genotypes of the 3 variants. Thus, each patient had a genetic score ranging from 0 to 3. For convenience, we further divided patients into 2 groups (score 0–1, and score 2–3). In the test set, the original genetic score (with 4 categories) had a significant discrimination capability in DFS in the nonchemotherapy group (P = 0.0027, Supplementary Fig. S2A) but not in the chemotherapy group (P = 0.088, Supplementary Fig. S2B). The modified genetic score (with 2 categories) displayed similar discrimination capability (Supplementary Fig. S2C and S2D). An interaction of genetic score and chemotherapy was also observed (P = 0.004, Table 3). In the validation set, which consisted of 339 patients mainly with familial/early-onset breast cancer, the genetic score tended to become a predicator of DFS in the nonchemotherapy group (P = 0.050, Fig. 3A) but not in the chemotherapy group (P = 0.304, Fig. 3B). Similarly, an interaction between genetic score and chemotherapy was observed with a borderline significance (HR = 2.07, P = 0.045).

Figure 3.

Combined genetic factors improve the prediction of disease progression in patients not undergoing chemotherapy. Effect of genetic score (0–1 vs. 2–3) on DFS in primary breast cancer patients treated without (A) or with (B) adjuvant chemotherapy in a second population (n = 339) are shown, with log-rank P values of 0.050 and 0.304, respectively. C, ROC curves assessing the discriminatory performance of the combined model and traditional model for prediction of disease progression in patients not undergoing chemotherapy. Variables for regression of the traditional model include age, lymph nodes status, tumor size, ER, PR, HER2, and endocrine therapy. Genetic score is added in the combined model. The probability of disease events is estimated as eL/(1 + eL), in which L is derived from logistic regression analysis. P = 0.047 for AUC comparison.

Figure 3.

Combined genetic factors improve the prediction of disease progression in patients not undergoing chemotherapy. Effect of genetic score (0–1 vs. 2–3) on DFS in primary breast cancer patients treated without (A) or with (B) adjuvant chemotherapy in a second population (n = 339) are shown, with log-rank P values of 0.050 and 0.304, respectively. C, ROC curves assessing the discriminatory performance of the combined model and traditional model for prediction of disease progression in patients not undergoing chemotherapy. Variables for regression of the traditional model include age, lymph nodes status, tumor size, ER, PR, HER2, and endocrine therapy. Genetic score is added in the combined model. The probability of disease events is estimated as eL/(1 + eL), in which L is derived from logistic regression analysis. P = 0.047 for AUC comparison.

Close modal

Predictive value of polymorphisms in disease progression

The genetic factors that are capable of predicting disease progression could still have no clinical utility unless they can offer additional information beyond what classic predictors can already tell us. We evaluated the predictive value of genetic score in disease progression in patients receiving no chemotherapy. Genetic score (0–1 vs. 2–3) was added in a traditional model. ROC analysis showed an AUC of 0.70 (95% CI: 0.60–0.80) for the traditional model and 0.78 (95% CI: 0.69–0.86) for the combined model (Fig. 3C), suggesting that adding genetic factors to classic factors markedly improved the prediction capability (P = 0.047 for AUC comparison).

In this prospective observational study, we noted that associations between DFS and germline polymorphisms in the estrogen–quinone metabolizing genes involved in OS were modified by adjuvant chemotherapy. The observed interaction was successfully validated. Although the genetic variations associated with an enhanced ROS-metabolizing capability would reduce the hazard of disease progression among women receiving no chemotherapy, they would compromise the efficiency of adjuvant chemotherapy.

Several prior studies have examined the association between breast cancer disease outcome and genotypes of GSTM1 and NQO2 (30–36). Our study is the only study that has investigated the interaction between chemotherapy and genotypes. Our results consistently indicate that genetic variants in estrogen–quinone metabolizing genes play protective roles in disease progression if no chemotherapy is administered. However, this effect disappears after chemotherapy is administered; in other words, patients harboring genotypes related to low ROS levels receive limited, if any, benefit from chemotherapy.

The conflicting outcome between the chemotherapy and nonchemotherapy groups is explainable. In general, the protective genotypes that are associated with higher expression or elevated activity of estrogen–quinone metabolizing enzymes would reduce ROS levels and subsequently inhibit OS-induced cancer cell proliferation, angiogenesis, and blood supply to residual cells or tumors in breast cancer patients after surgery (11). Some may argue that breast tumors have significantly higher OS levels than normal tissues and that a moderate decrease of ROS caused by germline genetic variants might have limited effects. This is true for advanced large tumors. However, after surgery, patients have only disseminated cancer cells or a tiny residual tumor rather than a large mass. The small quantity of cancer cells is exposed to the normal microenvironment and affected by its surrounding, which consists of thousands of normal cells. Variability in ROS levels in the microenvironment caused by germline variations may influence the outcome of dormancy and extinction or proliferation and dissemination of residual cancer cells. Moreover, decreased ROS can protect normal breast cells from OS-induced DNA damage and reduce genetic instability, resulting in prevention of second primary breast cancers. However, when chemotherapy is administered, the situation is changed. As we know, most chemotherapy regimens exert their cytotoxic effects by elevating the OS levels within breast carcinomas, pushing malignant cells “over the edge” and increasing OS damage to a level that the cancer cells cannot cope with (14). Although anthracyclines act primarily by interfering with topoisomerase II activity, some studies indicated that many active chemotherapeutic drugs in breast cancer treatment (including anthracyclines and cyclophosphamide) are also ROS-generating agents (14, 37). Therefore, we might speculate that when chemotherapy is administered to a patient with decreased ROS levels, the originally protective effect of ROS might cause resistance to chemotherapy. The precise mechanism of interaction between ROS and chemoresistance in vivo, however, needs further investigation. Another explanation is that, because both anthracyclines and cyclophosphamide are metabolized through reactions mediated by GSTMs (13), the presence of GSTM1 could accelerate the inactivation and metabolism mechanisms of these therapeutic agents.

The interaction between chemotherapy and genotypes of ROS-reducing genes can account for the conflicting results seen in previous studies with regard to the relationship between GSTM1-null genotype and disease progression in an adjuvant setting. For patients receiving no chemotherapy, the GSTM1-present genotype tends to be protective (GSTM1 is one of the 16 cancer-related genes within the 21-gene Oncotype-DX assay for predicting recurrence of tamoxifen-treated, node-negative breast cancer; ref. 38). For patients undergoing chemotherapy, GSTM1-present probably has a similar effect to GSTM1-null (30–33). For mixed patients, the results tend to be nonsignificant (34, 35). We also noted some SNPs in GSTM4 that were associated to different extents with DFS regardless of whether chemotherapy was used or not. This issue needs further study. Moreover, our data imply that NQO2 has similar features to GSTM1. It is probable that NQO2 has an intrinsic metabolizing capability for many chemotherapeutic agents. We also found no change in expression of NQO2 gene in cancer cells with different germline genotypes, which indicates that many other factors (e.g., methylation, aberrant regulation) rather than only the regulative effect of polymorphic alleles influence NQO2 expression in cancer cells. This observation may suggest that NQO2 expression levels in healthy tissues affect disease progression, perhaps, by influencing tumor microenvironment.

Our study has several limitations. First, as a prospective observational study, but not a clinical trial, the chemotherapy regimens in our study are not uniform. It is difficult to determine whether the chemoresistance is specific to a particular agent or to several agents. Second, genetic variants in other OS-related genes as well as combinations of these genotypes might provide a more accurate prediction of chemotherapy resistance. Third, because recurrence events are time dependent, and the time-dependent model is complicated to establish and to utilize, we employed a feasible modeling procedure in our analysis, which however might compromise the results.

In summary, our results suggest that although reduced OS levels might be important in preventing breast cancer progression, they probably compromise the effectiveness of adjuvant chemotherapy. Thus far, there are few rigorous data in the literature to support individualized regimens of chemotherapy according to host genotypes. The new understanding of interactions between chemotherapy resistance and host genetic factors could impact basic research as well as clinical management, potentially leading to individualized strategies for adjuvant chemotherapy. For instance, for patients with aggressive disease and also harboring genotypes associated with higher expression or enhanced activity of ROS-reducing enzymes, physicians may consider choosing chemotherapeutic agents that exert their efforts by non/low OS-mediated mechanisms such as taxanes and vinca alkaloids, rather than anthracyclines and cyclophosphamide (14, 37). Finally, we propose that combined germline genotypes of multiple genes involved in ROS pathways might achieve a more precise prediction of disease progression on the basis of classic somatic factors.

No potential conflicts of interest were disclosed.

This research is supported by grants from the National Natural Science Foundation of China (30971143, 30972936, 81001169), the Shanghai United Developing Technology Project of Municipal Hospitals (SHDC12010116), the Key Clinical Program of the Ministry of Health (2010-2012), the 2009 Youth Fund of Shanghai Public Health Bureau, the 2009 Youth Fund of Shanghai Medical College, and the Shanghai Committee of Science and Technology Fund for 2011 Qimingxing Project (for K-D. Yu, 11QA1401400). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Yager
JD
,
Davidson
NE
. 
Estrogen carcinogenesis in breast cancer
.
N Engl J Med
2006
;
354
:
270
82
.
2.
Vasiliou
V
,
Ross
D
,
Nebert
DW
. 
Update of the NAD(P)H:quinone oxidoreductase (NQO) gene family
.
Hum Genomics
2006
;
2
:
329
35
.
3.
Fagerholm
R
,
Hofstetter
B
,
Tommiska
J
,
Aaltonen
K
,
Vrtel
R
,
Syrjakoski
K
, et al
NAD(P)H:quinone oxidoreductase 1 NQO1*2 genotype (P187S) is a strong prognostic and predictive factor in breast cancer
.
Nat Genet
2008
;
40
:
844
53
.
4.
Thompson
PA
,
Ambrosone
C
. 
Molecular epidemiology of genetic polymorphisms in estrogen metabolizing enzymes in human breast cancer
.
J Natl Cancer Inst Monogr
2000
:
125
34
.
5.
Pearson
WR
,
Vorachek
WR
,
Xu
SJ
,
Berger
R
,
Hart
I
,
Vannais
D
, et al
Identification of class-mu glutathione transferase genes GSTM1-GSTM5 on human chromosome 1p13
.
Am J Hum Genet
1993
;
53
:
220
33
.
6.
Yu
KD
,
Di
GH
,
Yuan
WT
,
Fan
L
,
Wu
J
,
Hu
Z
, et al
Functional polymorphisms, altered gene expression and genetic association link NRH:quinone oxidoreductase 2 to breast cancer with wild-type p53
.
Hum Mol Genet
2009
;
18
:
2502
17
.
7.
Yu
KD
,
Fan
L
,
Di
GH
,
Yuan
WT
,
Zheng
Y
,
Huang
W
, et al
Genetic variants in GSTM3 gene within GSTM4-GSTM2-GSTM1-GSTM5-GSTM3 cluster influence breast cancer susceptibility depending on GSTM1
.
Breast Cancer Res Treat
2010
;
121
:
485
96
.
8.
Olivier
M
,
Langerod
A
,
Carrieri
P
,
Bergh
J
,
Klaar
S
,
Eyfjord
J
, et al
The clinical value of somatic TP53 gene mutations in 1,794 patients with breast cancer
.
Clin Cancer Res
2006
;
12
:
1157
67
.
9.
Kalinsky
K
,
Jacks
LM
,
Heguy
A
,
Patil
S
,
Drobnjak
M
,
Bhanot
UK
, et al
PIK3CA mutation associates with improved outcome in breast cancer
.
Clin Cancer Res
2009
;
15
:
5049
59
.
10.
Udler
M
,
Maia
AT
,
Cebrian
A
,
Brown
C
,
Greenberg
D
,
Shah
M
, et al
Common germline genetic variation in antioxidant defense genes and survival after diagnosis of breast cancer
.
J Clin Oncol
2007
;
25
:
3015
23
.
11.
Brown
NS
,
Bicknell
R
. 
Hypoxia and oxidative stress in breast cancer. Oxidative stress: its effects on the growth, metastatic potential and response to therapy of breast cancer
.
Breast Cancer Res
2001
;
3
:
323
7
.
12.
Toullec
A
,
Gerald
D
,
Despouy
G
,
Bourachot
B
,
Cardon
M
,
Lefort
S
, et al
Oxidative stress promotes myofibroblast differentiation and tumour spreading
.
EMBO Mol Med
2010
;
2
:
211
30
.
13.
Tan
SH
,
Lee
SC
,
Goh
BC
,
Wong
J
. 
Pharmacogenetics in breast cancer therapy
.
Clin Cancer Res
2008
;
14
:
8027
41
.
14.
Halliwell
B
. 
Oxidative stress and cancer: have we moved forward?
Biochem J
2007
;
401
:
1
11
.
15.
Goldhirsch
A
,
Ingle
JN
,
Gelber
RD
,
Coates
AS
,
Thurlimann
B
,
Senn
HJ
. 
Thresholds for therapies: highlights of the St Gallen International Expert Consensus on the primary therapy of early breast cancer 2009
.
Ann Oncol
2009
;
20
:
1319
29
.
16.
Yu
KD
,
Di
GH
,
Fan
L
,
Wu
J
,
Hu
Z
,
Shen
ZZ
, et al
A functional polymorphism in the promoter region of GSTM1 implies a complex role for GSTM1 in breast cancer
.
FASEB J
2009
;
23
:
2274
87
.
17.
Yin
W
,
Di
G
,
Zhou
L
,
Lu
J
,
Liu
G
,
Wu
J
, et al
Time-varying pattern of recurrence risk for Chinese breast cancer patients
.
Breast Cancer Res Treat
2009
;
114
:
527
35
.
18.
Harvey
JM
,
Clark
GM
,
Osborne
CK
,
Allred
DC
. 
Estrogen receptor status by immunohistochemistry is superior to the ligand-binding assay for predicting response to adjuvant endocrine therapy in breast cancer
.
J Clin Oncol
1999
;
17
:
1474
81
.
19.
Jacobs
TW
,
Gown
AM
,
Yaziji
H
,
Barnes
MJ
,
Schnitt
SJ
. 
Specificity of HercepTest in determining HER-2/neu status of breast cancers using the United States Food and Drug Administration-approved scoring system
.
J Clin Oncol
1999
;
17
:
1983
7
.
20.
Goldhirsch
A
,
Glick
JH
,
Gelber
RD
,
Coates
AS
,
Thurlimann
B
,
Senn
HJ
. 
Meeting highlights: international expert consensus on the primary therapy of early breast cancer 2005
.
Ann Oncol
2005
;
16
:
1569
83
.
21.
Goldhirsch
A
,
Glick
JH
,
Gelber
RD
,
Senn
HJ
. 
Meeting highlights: International Consensus Panel on the Treatment of Primary Breast Cancer
.
J Natl Cancer Inst
1998
;
90
:
1601
8
.
22.
Li
WF
,
Hu
Z
,
Rao
NY
,
Song
CG
,
Zhang
B
,
Cao
MZ
, et al
The prevalence of BRCA1 and BRCA2 germline mutations in high-risk breast cancer patients of Chinese Han nationality: two recurrent mutations were identified
.
Breast Cancer Res Treat
2008
;
110
:
99
109
.
23.
Thalib
L
,
Wedren
S
,
Granath
F
,
Adami
HO
,
Rydh
B
,
Magnusson
C
, et al
Breast cancer prognosis in relation to family history of breast and ovarian cancer
.
Br J Cancer
2004
;
90
:
1378
81
.
24.
Arun
B
,
Bayraktar
S
,
Liu
DD
,
Gutierrez
Barrera AM
,
Atchley
D
,
Pusztai
L
, et al
Response to neoadjuvant systemic therapy for breast cancer in BRCA mutation carriers and noncarriers: a single-institution experience
.
J Clin Oncol
2011
;
29
:
3739
46
.
25.
Fourquet
A
,
Stoppa-Lyonnet
D
,
Kirova
YM
,
Sigal-Zafrani
B
,
Asselain
B
. 
Familial breast cancer: clinical response to induction chemotherapy or radiotherapy related to BRCA1/2 mutations status
.
Am J Clin Oncol
2009
;
32
:
127
31
.
26.
Bell
PA
,
Chaturvedi
S
,
Gelfand
CA
,
Huang
CY
,
Kochersperger
M
,
Kopla
R
, et al
SNPstream UHT: ultra-high throughput SNP genotyping for pharmacogenomics and drug discovery
.
Biotechniques
2002
;
Suppl
:
70
2
,
4, 6–7
.
27.
Wang
W
,
Jaiswal
AK
. 
Sp3 repression of polymorphic human NRH:quinone oxidoreductase 2 gene promoter
.
Free Radic Biol Med
2004
;
37
:
1231
43
.
28.
Aldea
C
,
Alvarez
CP
,
Folgueira
L
,
Delgado
R
,
Otero
JR
. 
Rapid detection of herpes simplex virus DNA in genital ulcers by real-time PCR using SYBR green I dye as the detection signal
.
J Clin Microbiol
2002
;
40
:
1060
2
.
29.
Colleoni
M
,
Bagnardi
V
,
Rotmensz
N
,
Gelber
RD
,
Viale
G
,
Pruneri
G
, et al
Increasing steroid hormone receptors expression defines breast cancer subtypes non responsive to preoperative chemotherapy
.
Breast Cancer Res Treat
2009
;
116
:
359
69
.
30.
Gor
PP
,
Su
HI
,
Gray
RJ
,
Gimotty
PA
,
Horn
M
,
Aplenc
R
, et al
Cyclophosphamide-metabolizing enzyme polymorphisms and survival outcomes after adjuvant chemotherapy for node-positive breast cancer: a retrospective cohort study
.
Breast Cancer Res
2010
;
12
:
R26
.
31.
Yang
G
,
Shu
XO
,
Ruan
ZX
,
Cai
QY
,
Jin
F
,
Gao
YT
, et al
Genetic polymorphisms in glutathione-S-transferase genes (GSTM1, GSTT1, GSTP1) and survival after chemotherapy for invasive breast carcinoma
.
Cancer
2005
;
103
:
52
8
.
32.
Ambrosone
CB
,
Sweeney
C
,
Coles
BF
,
Thompson
PA
,
McClure
GY
,
Korourian
S
, et al
Polymorphisms in glutathione S-transferases (GSTM1 and GSTT1) and survival after treatment for breast cancer
.
Cancer Res
2001
;
61
:
7130
5
.
33.
DeMichele
A
,
Aplenc
R
,
Botbyl
J
,
Colligan
T
,
Wray
L
,
Klein-Cabral
M
, et al
Drug-metabolizing enzyme polymorphisms predict clinical outcome in a node-positive breast cancer cohort
.
J Clin Oncol
2005
;
23
:
5552
9
.
34.
Khedhaier
A
,
Remadi
S
,
Corbex
M
,
Ahmed
SB
,
Bouaouina
N
,
Mestiri
S
, et al
Glutathione S-transferases (GSTT1 and GSTM1) gene deletions in Tunisians: susceptibility and prognostic implications in breast carcinoma
.
Br J Cancer
2003
;
89
:
1502
7
.
35.
Syamala
VS
,
Sreeja
L
,
Syamala
V
,
Raveendran
PB
,
Balakrishnan
R
,
Kuttan
R
, et al
Influence of germline polymorphisms of GSTT1, GSTM1, and GSTP1 in familial versus sporadic breast cancer susceptibility and survival
.
Fam Cancer
2008
;
7
:
213
20
.
36.
Choi
JY
,
Barlow
WE
,
Albain
KS
,
Hong
CC
,
Blanco
JG
,
Livingston
RB
, et al
Nitric oxide synthase variants and disease-free survival among treated and untreated breast cancer patients in a Southwest Oncology Group clinical trial
.
Clin Cancer Res
2009
;
15
:
5258
66
.
37.
Conklin
KA
. 
Chemotherapy-associated oxidative stress: impact on chemotherapeutic effectiveness
.
Integr Cancer Ther
2004
;
3
:
294
300
.
38.
Paik
S
,
Shak
S
,
Tang
G
,
Kim
C
,
Baker
J
,
Cronin
M
, et al
A multigene assay to predict recurrence of tamoxifen-treated, node-negative breast cancer
.
N Engl J Med
2004
;
351
:
2817
26
.