Differential gene expression in normal and transformed human mammary epithelial cells in response to oxidative stress.

Oxygen free radicals are generated continuously within mammalian cells as a direct consequence of aerobic metabolism and respiration. Reactive oxygen species (ROS) homeostasis is of critical importance for all aerobic cells, since too much oxygen results in toxicity and low oxygen levels impairs metabolism. The balance between oxidative damage and antioxidant protection controls the levels of ROS through a combination of gene regulatory, biochemical, and physiological mechanisms. An excess of oxygen-free-radical formation causes oxidative stress resulting in oxidative damage to biomolecules, lipid peroxidation, DNA damage, and carcinogenesis [ 1 , 2 ]. In recent years, there have been great advances and developments in our understanding of the molecular mechanisms and factors involved in breast carcinogenesis [ 3 ]. Factors that induce persistent oxidant stress, such as inflammation, are now believed to play an important role in the prognosis and invasiveness of this and other cancers [ 4 ]. The precise mechanisms of oxidative stress generation in breast cancer cells are still not very well understood and documented, although, there are some reports on the oxidant-antioxidant profile in breast cancer patients [ 5 - 8 ]. Some human cancer cell lines produce large amounts of hydrogen peroxide (H 2 O 2 ) while antioxidant enzymes, such as superoxide dismutase and catalase, appear to be down-regulated in cancer cells [ 9 ]. The consequences of enhanced oxidative stress in human breast cancer can also be found in elevated levels of lipid peroxidation products [ 10 , 11 ]. ROS induce several kinds of DNA damage products in malignant cells including strand breaks, base modifications and DNA-protein crosslinks. Work by Frenkel et al [ 12 ] measuring circulating auto-antibodies to the oxidative DNA damage product 5-hydroxymethyl-2’-deoxyuridine (HMdU) in patients with breast or colorectal cancer, suggests that enhanced generation of oxidative DNA damage precedes and stimulates neoplasia. Others studies indicate high levels of DNA oxidation in human cancer tissues compared with corresponding controls [ 1 ]. Strong evidence suggests that carcinoma cells in vitro and in vivo are frequently under persistent oxidative stress [ 13 ]. Similarly, recent reports also indicate that high concentration of free iron in endometriotic cysts promotes carcinogenesis through iron induced persistent oxidative stress, and that malignant cells can survive a high oxidative stress environment [ 14 ]. Microarray analysis of breast cancer cell lines and tumor samples is a powerful tool to understand the global changes in gene expression associated with cancer progression, as well as for the development of profiles that can distinguish, identify and classify discreet subsets of disease and predict disease outcome or response to therapy [ 15 - 18 ]. Recent microarray analyses allowed the comparison of not only gene expression with respect to different phenotypes, but also the evaluation of biological functions, such as oncogenic signalling activity as well as the discovery of new breast cancer genes [ 17 ]. Several studies have reported genes that are differentially expressed in breast cancer cell lines and tumors [ 15 - 20 ]. However, despite the strong link between increased local oxidative stress and breast carcinogenesis, to the best of our knowledge there are no studies on the relationship between oxidative stress responses and breast cancer malignancy progression. In this regard, there are few data addressing whether malignant breast epithelial cells differ from their non-transformed counterparts with regard to their responses to oxidative stress [ 21 ]. The primary purpose of this study was to identify the characteristic gene expression profiles that distinguish the response to oxidative stress in normal and tumorigenic breast cancer cell lines using microarray analyses. Next, by analysing comprehensively the genes differentially expressed, we sought to identify pathways and gene networks significantly regulated in normal and cancer cells in response to oxidative stress. We show that 87% of the genes altered in response to oxidative stress in normal mammary epithelial cells overlap those associated with progression to malignancy. Our findings present strong evidence that persistent oxidative stress is a crucial mechanism in the progression from a normal to malignant state as the genes that are regulated by oxidative stress in normal cells are also the genes that differentiate normal from tumorigenic cell lines.

Although the precise pathways contributing to oxidative stress in cancer cells remain unclear, several fundamental cellular mechanisms and external factors are thought to cause oxidative stress during cancer development and disease progression. In addition to oncogenes activation and lack of functional p53, abnormal metabolism and mitochondrial dysfunction are considered two of the intrinsic factors that cause elevated levels of ROS in cancer cells [ 47 - 49 ]. Genes coding for subunits of the five complexes of the electron transport respiratory chain in the mitochondria were significantly up-regulated by oxidative stress and by progression to cancer. Additionally, a significant number of mitochondrial genes, involved not only in oxidative phosphorylation but also in other mitochondrial functions were up-regulated by oxidative stress and cancer progression. Furthermore, citrate cycle and pyruvate metabolism were also up-regulated, suggesting a profound modification of the energy metabolism in cancer cells and in their non-tumorigenic counterparts when exposed to oxidative stress. Although an increase in glycolysis is a well-described alteration in cancer [ 50 ], our observations that citric acid cycle and oxidative phosphorylation enzymes are also up-regulated in breast cancer cells and normal HMEC under acute oxidative stress suggest that multiple mechanisms up-regulate aerobic energy metabolism during breast cancer progression. Since our experiments were conducted in cells grown in a well-oxygenated tissue culture environment, these results may model changes in smaller, early stage tumors rather than the more hypoxic environment seen in larger tumors. Similar over-expression of energy metabolism proteins has been reported in breast, colon, ovary prostate and renal cancer cell lines and tumors [ 51 ]. Collectively, these results suggest that enhanced oxidative energy metabolism found in normal cells exposed to elevated levels of ROS is a common feature of cancer progression in different tissues, which, according to our findings, may have arisen from an adaptation to persistent high levels of ROS in cancer cells. The experiments reported herein demonstrate that normal and malignant breast epithelial cells differ substantially with regard to acute changes in gene expression in response to oxidative stress induced by elevated GluOx activity. When the gene expression profile response to oxidative stress across the different cell types was compared, there were significant differences in the number of genes regulated in normal HMEC compared to HMLER cells. While normal HMEC cells are significantly affected by oxidative stress, HMLER cells are almost unresponsive to oxidative stress exposure. However, the gene expression profile of normal HMEC cells under oxidative stress conditions shows similarity to that of the malignant HMLER under normal conditions, suggesting that HMLER cells are already in an oxidatively stressed state and are tolerant or resistant to elevated levels of ROS. Redox adaptation mechanisms remain an important concept that, to a large extent, explains the processes by which cancer cells survive under persistent endogenous oxidative stress, becoming resistant to ROS and anticancer agents.

Our model for breast cancer consists of a primary human mammary epithelial cell line (HMEC) obtained from reduction mammoplasty, and two transformed mammary epithelial cell lines derived from these cells, HMLER-1 and HMLER-5, kindly provided by RA Weinberg (M.I.T., Cambridge, MA). HMLER-1 and HMLER-5 were obtained by in vivo transformation of HMEC with a series of oncogenes and cancer-associated genes, including: telomerase catalytic subunit (hTERT), SV40 large-T antigen and H-Ras V12 , an oncogenic allele of H-Ras [ 22 ]. HMLER-1 cells have an intermediate expression level of H-Ras V12 rarely form tumors, while HMLER-5 cells have high-levels of H-Ras V12 and are highly tumorigenic [ 22 ]. To determine optimal conditions for inducing oxidative stress by GluOx treatment, HMEC subconfluent cells in the exponential phase of growth were supplied with 10 mM glucose and varying concentrations of GluOx (control, 0.02 and 0.2U) at varying time intervals (2, 4, 8 and 16 h). Following GluOx treatment cell viability was evaluated as judged by trypan blue exclusion. Cells were counted, washed and cell pellets were collected and frozen at -80 °C for glutathione analysis. Pelleted cells were extracted with 250 μl of methanol:water (50:50, v/v) acidified with 0.01M hydrochloric acid to minimize oxidation of thiols, and spiked with 4ug/ml of the isotopically labeled internal standards L -Cysteine-3,3-d 2 (Cambridge Isotopes Laboratories, Inc. Andover, MA) and DL -Homocysteine-3,3,4,4-d 4 (C/D/N Isotopes Inc. Pointe-Claire, Quebec, Canada). After the extraction buffer was added, cell lysis was induced by 3 cycles of 5 min incubation in dry ice, followed by 1 min sonication in ultrasonic bath. The extract was mixed with 250 μl of 5mM N-ethylmaleimide in water, vortex and incubated for 30 min at room temperature to block the free thiol groups. The cell culture lysates were centrifuged for 15 min at 13,000 rpm to remove the cellular debris. GSSG and GSH were measured using pre-column AccQ·Tag™ Ultra UPLC derivatization kit (Waters Corporation, Milford, MA), following a modified protocol from [ 23 ]. Briefly, reagents for derivatization were prepared and derivatization was performed according to manufacturer’s protocol. LC-MS-PDA analysis was performed on LC-MS system comprised of Waters Acquity UPLC system (Waters Corporation, Milford, MA), equipped with the Acquity photodiode array detector, interfaced with ThermoFisher LTQ mass spectrometer (ThermoFisher Corporation, San Jose CA). UPLC separation was performed on AccQ·Tag™ Ultra column (1.7 μm, 100 mm × 2.1 mm i.d.,) from Waters. Column effluent was ionized by electrospray ionization (ESI), and the mass spectrometer was operated in full scan positive. GSSG and GSH quantitation was performed using calibration curves with isotopically labeled internal standards L -Cysteine-3,3-d 2 (Cambridge Isotopes Laboratories, Inc. Andover, MA) and DL -Homocysteine-3,3,4,4-d 4 (C/D/N Isotopes Inc. Pointe-Claire, Quebec, Canada), that were added to each sample. Commercially available forms of the GSSG and GSH were purchased from Sigma-Aldrich (Sigma-Aldrich, St. Louis, MO), and used to prepare the calibration curves. Cells were cultured in basal medium supplemented with 10 ng/ml of human epidermal growth factor (EGF), 0.5 μg/ml hydrocortisone, 100 U/ml penicillin-streptomycin and 10 μg/ml insulin under controlled conditions in a humidifier incubator maintained at 5% CO 2 and 37°C. Triplicates of HMECs, HMELR-1 and HMLER-5 subconfluent cells in the exponential phase of growth and supplied with 10 mM glucose, were either left untreated (control) or treated with GluOx 0.2 U/ml for 2 hours. This treatment did not affect cell viability as judged by trypan blue exclusion. Cells were removed from the dishes by scraping, counted, washed and cell pellets were collected and frozen at -80°C for microarray analysis. Total RNA was isolated from samples using Trizol Reagent, according to standard protocols (Invitrogen, Carlsbad, CA), followed by a cleanup procedure with RNeasy MinElute Cleanup Kit (Qiagen, Valencia, CA). RNA quality was assessed using UV spectra characteristics (NanoDrop, Wilmington, DE) and microanalysis (Agilent Bioanalyzer) for size and integrity of the total RNA. Affymetrix U133 Plus 2.0 human oligonucleotide microarray containing over 47,000 transcripts and variants, including 38,500 well-characterized human genes was used for gene expression analysis. Preparation of in vitro cRNA, oligonucleotide array hybridization and scanning were performed according to Affymetrix (Santa Clara, CA) protocols. We generated a probe set based gene expression data file from quantified image files with the GeneChip Multi-Array Average (GCRMA) method [ 24 ] associated packages from the BioConductor tool suite ( http://www.bioconductor.org/ ) [ 20 , 25 ], using R version 2.10 ( www.r-project.org ) and annotated with Unigene annotations from the February 2009 mapping version of the human genome. All 18 CEL files were analysed simultaneously, yielding a data matrix of probe sets by cell lines in which each value is the calculated log abundance of each gene probe set for each cell line under oxidative stress or control treatment conditions. Background subtraction, quantile normalization and gene data summarization was done by GCRMA. Differential expression analysis was performed using the linear modelling features of the limma package [ 26 ]. The positive False Discovery Rate [ 27 ] multiple-testing adjustment was applied to correct p - values . All genes in a comparison of interest with an adjusted p-value ≤ 0.05 were considered as statistically significant regardless of the fold difference in expression level. All genes significantly differentially expressed across different experimental conditions were analyzed in Database for Annotation, Visualization and Integrated Discovery (DAVID) version 2 ( http://david.niaid.nih.gov/david/version2/index.htm ) [ 28 ] to discover KEGG pathways that are significantly represented in these groups of genes. Additionally, gene network relationships and pathway analysis was performed using the Ingenuity Pathways Analysis (Ingenuity® Systems, www.ingenuity.com ) a repository database of molecular interaction, regulatory events gene to phenotype associations and chemical knowledge. With this gene expression analysis tools we identified known molecular network genes and pathways significantly represented among differentially expressed genes. Both, DAVID and Ingenuity Pathways Analysis use a Fisher’s Exact p -value to calculate the association between a particular set of differentially expressed genes and known pathways and gene networks that are most significantly enriched. A Fisher’s Exact p -value = 0 represent perfect enrichment. All pathways with a p -value ≤ 0.05 were considered strongly enriched with differentially expressed genes.

We are interested in understanding the mechanism by which oxidative stress is involved in breast cancer development and progression. Accumulated evidence suggests that cancer cells seem to function with higher levels of oxidative stress in vitro and in vivo compared with their normal counterparts, and consequently have increased levels of ROS [ 29 , 30 ]. In spite of evidence that cancer cells are oxidatively stressed, several important questions have not yet been addressed: (i) whether there is any difference in the oxidative stress response mechanisms of carcinogenic cells compared to their normal counterparts or whether the antioxidant system in cancer cells is suppressed; and (ii) what mechanisms, mediated by oxidative stress, in normal cells may contribute to their progression towards malignancy, if any. In an effort to further understand the mechanisms of oxidative stress associated with the progression of breast malignancy, our study concentrated on examining the gene expression profile of normal human primary mammary epithelial cells and fully tumorigenic transformed breast cancer cells in response to oxidative stress induced by GluOx. Cultured normal mammary epithelial cells (HMECs) were treated with varying levels of H 2 O 2 generated in the cultured media by glucose oxidase as described in the methods section. The effects of the GluOx treatment on the levels of reduced (GSH) and oxidized (GSSG) glutathione are shown in Fig. 1 . We used this information to determine the GluOx dose and minimum incubation time to induce oxidative stress. The results show that after 2h treatment with 0.2U/ml of GluOx the pool of free GSH was reduced significantly compared to untreated cells, or cells that were treated with 0.02U/ml of GluOx regardless of the incubation time. Similarly, the levels of GSSG increased after 2h treatment with 0.2U/ml of GluOx, while in untreated cells and cells treated with 0.02U/ml the levels of GSSG were below the limits of detection. We also observed that cell death increases progressively with time, in cultures treated with 0.2U/ml of GluOx. While after 2h treatment cell viability was not affect, based on trypan blue exclusion, 10%, 20% and 90% of the cells die after 4h, 8h, and 16h treatment with 0.2U/ml of GluOx, respectively. As previously reported by others [ 31 - 34 ], glucose oxidase treatment is a standard method of generating the hydrogen peroxide that enables exposure of treated cells to a constant dose of hydrogen peroxide. Since hydrogen peroxide will decay over time, incubation with GluOx in the presence of glucose is preferable to treatment with a single bolus dose of hydrogen peroxide itself [ 34 ]. The ratio of GSSG-to-GSH is an indicator of cellular health, with GSH constituting up to 98% of cellular glutathione under normal conditions. An increased GSSG-to-GSH ratio is considered indicative of oxidative stress [ 35 ]. Our GSSG-to-GSH ratio and cell viability results indicate that a treatment with 0.2U/ml of GluOx for 2h induces adequate acute oxidative stress without affecting cell viability and is suitable to study gene expression response to oxidative stress conditions. Unlike other tissue culture models of breast cancer, the HMEC/HMLER system permits the comparison of oxidative stress-response in normal human breast epithelial cells with intermediate and fully malignant variants produced by completely defined genetic changes. We determined the influence of oxidative stress on the global gene expression profile of three cell types representing stages in the malignant progression of breast cells: normal mammary epithelial cells (HMECs), an immortalized mammary epithelial cell line expressing low levels of H-Ras V12 (HMLER-1), and a fully malignant, tumorigenic cell line expressing high levels of H-Ras V12 (HMLER-5). Triplicate cultures for each cell type were either grown under normal conditions or exposed to oxidative stress by treatment with glucose/GluOx. Initial data analysis of genes statistically significantly differentially expressed in response to GluOx treatment revealed significant transcriptional profile changes in normal HMECs, but markedly fewer in HMELR-1 or HMELR-5 cancer cell lines. In HMEC cells we found 11895 genes (25% of all the transcripts measured) differentially expressed in response to oxidative stress. Among them, 5966 had higher expression levels and 5929 had lower expression levels in oxidatively stressed HMECs compared to the corresponding control. Functional analysis of these gene sets revealed that pathways involved in cell cycle, energy metabolism, amino acids metabolism, protein biosynthesis, transport and degradation, fatty acid metabolism, biosynthesis and metabolism of antioxidant molecules were significantly up-regulated by oxidative stress ( Table 1 ). Pathways associated with cancer progression, apoptosis, cell junction and several cancer signalling pathways including p53, TGF-β and Notch, were down-regulated ( Table 2 ). In HMLER-1 cell line expressing low level of oncogenic H-Ras V12 , 634 genes were differentially expressed in response to oxidative stress, with 346 down-regulated and 288 up-regulated. Functional analysis showed that ubiquitin mediated proteolysis, proteosome, folic acid biosynthesis and pyruvate metabolism were up-regulated ( Table 1 ), while cell cycle, p53 and mTOR signalling pathways, and cancer progression pathways were down-regulated ( Table 2 ). In HMLER-5 cell line, expressing high levels of oncogenic H-Ras V12 , 379 genes were found regulated by oxidative stress treatment, with 265 down-regulated and 114 up-regulated. Functional analysis showed that oxidative phosphorylation was up-regulated ( Table 1 ), and pathways in cancer progression were down-regulated ( Table 2 ). Distribution of common genes differentially expressed in response to oxidative stress in HMEC, HMLER-1 and HMLER-5 cell types is shown in Figure 2 . These findings suggest that normal and tumorigenic cells differ strongly in their response to oxidative stress. In contrast, the gene expression profile of tumorigenic cell lines are comparatively unaffected by oxidative stress, while the gene expression in normal cells undergoes significant changes stimulated by the elevated intracellular levels of ROS. Since persistent oxidative stress is believed to contribute to cancer, the failure of oxidative stress to up-regulate pathways in cancer progression was initially surprising. However, considering the low number of genes in HMLER-1 and HMLER-5 cells that responded to oxidative stress induced by GluOx treatment, we considered that the genetic changes associated with oxidative stress observed in HMEC might have already occurred in cancer cells due to a persistent oxidative stress status and high levels of ROS, making them tolerant or resistant to further oxidative treatment with GluOx. To test this hypothesis, we compared the set of genes differentially expressed in normal HMEC in response to oxidative stress with those genes associated with cancer malignancy progression in HMLERs (HMLERs control vs HMEC GluOx) ( Figure 3 ). A set of genes associated with cancer malignancy progression identified by comparing unstressed tumorigenic HMLER cell lines with normal HMEC included genes in the TNFa/NFkB signalling pathway (i.e. AKT1 NFKBIA and MAP3K3), Wnt signalling pathway (i.e. CDC2, PIN1, TAX1BP3), and tumor suppressors (i.e. BRCA1, CDKN2A, NF1, NF2, PTEN). Further analysis of the 11895 genes differentially expressed in normal HMEC cells in response to oxidative stress, showed that 10545 (89%) are also differentially expressed during the progression to malignancy by HMLER cells (8550 or 72% by both HMLERs, 906 or 8% HMLER-5 only, and 1089 or 9% HMLER-1 only, from Figure 3 ). Among them we found genes with common responses--referring to those genes that were up or down-regulated in both oxidative stress and cancer malignancy progression (7182 out of 8550 – 84%); genes with opposite responses--referring to those that were up-regulated in oxidative stress, but down-regulated in malignancy progression and vice versa (1026 out of 8550 - 12%); and genes with heterogeneous response--referring to those which expression level varied, but were not particularly associated with either oxidative stress or malignancy progression exclusively (342 out of 8550 - 4%) ( Supplementary Tables 1-3 ). Moreover, we found that several of these genes with common and heterogeneous response are part of known gene networks and pathways related to cancer ( Supplementary Fig 1 Pathways in Cancer). Functional analysis of the pathways significantly enriched with genes that were induced or repressed in response to oxidative stress and cancer progression, in breast cells, is summarized in Tables 3 and 4 . As expected, the overlapping distribution of genes regulated by oxidative stress and cancer results in an overlap of biochemical pathways they represent. These observations present strong evidence that adaptation to persistent oxidative stress is a crucial mechanism in the progression from a normal to malignant state as the genes that are regulated by oxidative stress in normal cells are also the genes that differentiate normal from tumorigenic cell lines. We identified 67 genes related to DNA damage signalling pathways that were common among the genes differentially expressed in HMEC in response to oxidative stress and also during cancer malignancy progression in HMELER-1 and HMLER-5 ( Table 5 ). These 67 genes expanded several functional groups including nucleotide excision repair (NER), base excision repair (BER), mismatch excision repair (MER), DNA polymerases catalytic subunits, genes defective in diseases associated with sensitivity to DNA damaging agents, and other identified genes with DNA repair function. Of the 67 genes significantly altered, 47 were up-regulated by oxidative stress and cancer progression (e.g., tumor suppressor BRCA, 8-oxoGTPase NUDT1); ten genes were down-regulated by oxidative stress and cancer progression (i.e. ERCC6, XPA, RAD23B, ERCC4); four genes were up-regulated by oxidative stress and down-regulated by cancer progression (MGMT, RAD52, XPC, ERCC5); and two genes were down-regulated by oxidative stress and up-regulated by cancer progression (FANCL, XRCC2). For the two cancer cell lines, the similar overall trend in gene-expression changes indicated that even low expression of H-Rasv 12 had resulted in the induction of genes involved in DNA damage signalling pathways. The association between DNA damage by oxidative stress, DNA repair genes and the characteristic of breast cancer in patients and cell lines is well documented [ 30 , 36 ]. Oxidative DNA damage was reported to increase in human breast cancer tissues and transformed human breast cell lines compared to their normal counterparts [ 37 ]. Among those DNA repair or DNA damage signalling genes upregulated by oxidative stress and cancer progression, is worth mentioning RAD51C, which encodes a strand transfer protein involved in recombination repair of damage DNA, which over-expression in breast tissues has been associated with tumor progression [ 38 ]. Also, NUDT1 coding a protein that hydrolyzes oxidized purine nucleoside triphosphates, such as 8-oxo-dGTP and 2-hydroxy-dATP, to monophosphates, thereby preventing misincorpotation, was up-regulated. These are strong indicators of DNA damage and the formation of modified DNA bases, which are considered an important event in ROS-induced carcinogenesis [ 13 , 39 ]. In this regard, DNA retrieved from human breast cancer specimens, as well as adjacent non-malignant tissue demonstrate elevated levels of modified purines including 8-oxoguanine, 8-oxoadenine, and FapyGua (2,6-diamino-4-hydroxy-5-formamidopyrimidine), as compared to DNA from normal breast tissues [ 40 ]. Progression from normal to malignant breast tissue is accompanied by a shift from a high ratio of ring-open (Fapy) purines to 8-oxo adducts to one favouring the 8-oxo adducts [ 41 ], reflecting a change from a reducing nuclear environment to an oxidizing one during breast carcinogenesis [ 42 ]. Furthermore, DNA derived from invasive ductal breast cancer specimens of patients with metastatic disease have a higher total content of modified purines than does DNA derived from non-metastatic breast cancers [ 43 ], suggesting a role for oxidative DNA damage in the progression to the metastatic state. Our findings of specific DNA repair genes differentially expressed during both oxidative stress and cancer progression may help explain the specific DNA modifications associated with oxidative DNA damage that occur in breast cancer cells. The similarity in expressions patterns of DNA repair genes in response to oxidative stress and cancer malignancy progression support our hypothesis that breast cancer cells are persistently exposed to high levels of ROS, and have similar oxidative stress response mechanisms compared to those found in normal HMECs under acute oxidative stress conditions. In addition to the increased expression levels of DNA repair genes, the levels of several ROS-scavenging enzymes were significantly altered in malignant cells and in normal HMEC treated with GluOx, suggesting aberrant regulation of redox homeostasis and stress adaptation in cancer cells. Cells control ROS levels by balancing ROS generation with their elimination by ROS-scavenging systems, such as, superoxide dismutase, glutathione peroxidase, glutaredoxin, peroxiredoxin, thioredoxin and catalase. We looked at the expression patterns of the genes coding for the major ROS-scavenging enzymes and found that all except catalase were significantly altered in their expression in response to oxidative stress treatment and cancer progression ( Table 6 ). Of the 41 genes identified, 28 were up-regulated and 4 down-regulated by both oxidative stress treatment and cancer progression. Interestingly, seven genes including glutathione S-transferases (GSTO2, GSTM3, GSTP1, MGST2, GSTK1) and peroxiredoxin (PRDX5) were up-regulated by oxidative stress, but down-regulated during cancer progression. Superoxide dismutase SOD1, responsible for degrading superoxide radicals into oxygen and H 2 O 2 , was up-regulated by oxidative stress and by progression to cancer. Similarly, glutathione peroxidases (GPX3, GPX4, GPX8), enzymes involved in reducing lipid hydroperoxides to their corresponding alcohols and reducing H 2 O 2 to water, were up-regulated by oxidative stress and by progression to cancer. Peroxiredoxin PRDX5 increased in oxidatively stressed HMEC but decreased in HMELR, while PDRX2 and PDRX6 were up-regulated in both oxidatively stressed normal cells and cancer cells. Members of the peroxiredoxins protein family play an antioxidant protective role in mammalian cells and also mediate signal transduction. PRDX2 particularly, may have a proliferative effect and play a role in cancer development or progression [ 44 ]. Previous work indicates that treatment of normal epithelial cells with continuous but sublethal amounts of exogenous oxidants confers resistance to higher levels of oxidative stress treatment [ 45 ]. Persistent elevated levels of ROS in cancer cells may exert selective pressure towards cells capable of adapting to oxidative stress conditions. Similar to our findings, a proteomics approach indicated that oncogenic H-ras-transformed cells expressed higher levels of antioxidant proteins such as PDRX3 and thioredoxin peroxidase (Tsa1p) compared to their non-tumorigenic parental cells [ 46 ]. Related to the ROS-scavenging enzymes system and to the cellular redox homeostasis is the pentose phosphate pathway (PPP). Our results also reveal that the PPP is significantly up-regulated in cancer cell lines and in response to oxidative stress in HMECs. The PPP occurs exclusively in the cytoplasm and is the major source of NADPH in cells, producing approximately 60% of this electron donating cofactor required, which is used to reduce glutathione via glutathione reductase (GSR). Likewise, glutaredoxins (GLRX, GLRX3 and GLRX5) and thioredoxins (TXN and TNX2), which posses an active center disulfide bond and function as electron carries in the glutathione system, depend on the availability of NADPH to go from an oxidized to reduced state. Indeed, GSR, GLRX, GLRX3, GLRX5 TXN and TNX2, were significantly up-regulated by oxidative stress in HMEC and by cancer progression in HMLER cells, which clearly suggest that increasing the pool of NADPH and GSH is a specific adaptation mechanism of breast cancer cells to elevated ROS levels. Moreover, by up-regulation of the PPP, cancer cells increase the production of ribose-5-phosphate (R5P), which is used in synthesis of nucleotides and nucleic acids, and erythrose-4-phosphate (E4P), which is used in the synthesis of aromatic amino acids.