Peroxiredoxin-1 protects estrogen receptor α from oxidative stress-induced suppression and is a protein biomarker of favorable prognosis in breast cancer

ZR-75-1, T47D, MCF7 and SKBR3 cell lines were purchased from the European Collection of Cell Cultures (Wiltshire, UK). All cell lines were maintained through continuous passaging, and were confirmed to be free of contamination by Mycoplasma spp. Cells were maintained in DMEM (ZR-75-1 cell line) or RPMI-1640 (other cell lines) media (Sigma-Aldrich, St. Louis, MO, USA), with 10% FBS (Sigma-Aldrich), 100 U/mL penicillin, 100 μg/mL streptomycin (Gibco, Rockville, MD, USA) and 1.46 mg/mL L-glutamine (Gibco). Additionally, the growth medium for ZR-75-1 was supplemented with 1 nM β-estradiol (Sigma-Aldrich), and did not contain sodium pyruvate, which mediates elimination of H₂O₂ from the culture medium [21]. Tissue culture experimental techniques are further described in the Supplementary Methods (Additional file 1).

Adenanthin has been generated as previously described [20]. H₂O₂ and peroxynitrite were obtained from Sigma-Aldrich and Millipore (Billerica, MA, USA), respectively.

Antibody generation against PRDX1 was as described previously [22]. The Human Protein Atlas (HPA) Consortium carried out the initial quality control of the polyclonal anti-PRDX1 antibody (HPA007730) via optimization and testing on a variety of tissues (48 normal tissues and 16 cancer tissues) [23]. Immunoblotting and immunohistochemistry procedures are described in Supplementary Methods (Additional file 1).

Clinical materials originating from three independent cohorts of breast cancer patients were utilized in this study. Demographic and clinical characteristics for all cohorts are presented in this study are described in Supplementary Table S1 (Additional file 2).

Cohort 1 was represented on a reverse phase protein array (RPPA) containing protein extracts from 58 breast cancer cell lines and 998 human breast tumors, with available clinicopathological data for 712 cases. The clinical samples were collected at M. D. Anderson Cancer Center (MDACC) Houston, TX, USA, Hospital Clinico Universitario de Valencia, Spain, University of British Columbia, Vancouver, BC, Canada and Baylor College of Medicine, Houston, TX, USA [24]. Complete clinical information was available for 574 patients. All tumors in this training set were collected by excision during their primary surgery, followed by freezing of the tissue. Tumor content was verified by histopathology. All tissues were collected under Institutional Review Board (IRB)-approved laboratory protocols, awarded by the M. D. Anderson Cancer Center IRB, University of British Columbia IRB, and the Hospital Clinico Universitario de Valencia IRB. Patients consented to bank their specimens. As this was a retrospective study, each IRB provided a waiver of informed consent.

Cohort 2 consisted of a tissue microarray (TMA) with 498 consecutive invasive breast cancer cases (442 in final analysis) diagnosed at the Department of Pathology, Malmö University Hospital, between 1987 and 1992 [25]. The median age at diagnosis was 65 years (range, 27 to 96) and the median follow-up time was 11 years (range, 0 to 17). Two hundred and sixty-three patients were dead at the last follow-up (December 2004), 90 of whom were considered to be as a direct result of breast cancer (breast cancer-specific death). Complete endocrine treatment data were available for 379 patients, 160 of whom received adjuvant tamoxifen. Information on adjuvant chemotherapy was available for 382 patients, of whom 23 patients received treatment. The study has been approved by the Ethics Committee at Lund University. Informed consent was obtained for all included patients and opting out was an option.

Cohort 3 was represented on a RPPA consisting of 410 primary breast tumors collected by The Cancer Genome Atlas (TCGA) Consortium. These frozen primary tumor specimens collected from newly diagnosed patients with invasive breast adenocarcinoma undergoing surgical resection and who had received no prior treatment for their disease (chemotherapy or radiotherapy). Owing to the short median overall follow-up (20 months) and the small number of overall survival events (44 out of 410), survival analysis was not carried out on this cohort. This cohort was used solely to correlate quantitative levels of PRDX1 protein expression with expression of 180+ other proteins assessed on the same array. The TCGA project collects high-quality breast tumor samples and makes available the clinical information, molecular/genomic profiling data, and histopathology slide images on the TCGA data portal [26]. The TCGA data is organized into two categories: one that is openly accessible to the public and one that has controlled access, available only to qualified researchers obligated to secure the data. The open access data set contains only information that is not individually unique and does not pose a risk of patient re-identification. All the data used within this manuscript was obtained from the open access data set and has passed the criteria for unrestricted publication with the following statement ‘No restrictions; all data available without limitations’ listed at the publication guidelines section of the TCGA data portal [26].

Protein was extracted from tumor tissue and cell lines and probed for protein expression by reverse phase protein array (RPPA) analysis as previously described [24, 27‐30]. RPPA analysis was completed independently for cohorts 1 and 3. The technique is further described in the Supplementary Methods (Additional file 1).

Slides were scanned at 20X magnification using a ScanScope XT slide scanner (Aperio Technologies, Vista, CA, USA). The Spectrum Analysis algorithm package, ImageScope analysis software and Color Deconvolution algorithm (version 9; Aperio Technologies.) were applied to quantify immunohistochemical (IHC) staining. These algorithms were used to calculate the average positive intensity (API), as well as the area of positive staining, and the percentage of weak (1+), medium (2+), and strong (3+) positive staining. The final API was subtracted from 255, as these intensity ranges on an 8-bit scale of 0 to 255 (black to white, respectively). The maximum value from both cores for each patient was used for statistical analysis.

In light of conflicting results regarding the prognostic relevance of PRDX1 expression in breast cancer, a fundamental step for our study was comprehensive validation of the anti-PRDX1 antibody. Antibody specificity was confirmed in several breast cancer cell lines (T47D, ZR-75-1 and SKBR3), whereby PRDX1 expression was modified using short hairpin loop RNA (shRNA)-mediated gene knockdown and overexpression of cDNA encoding V5-tagged PRDX1, before antibody validation by immunoblotting (Figure 1A and C; complete gel displayed in Figure S1A in Additional file 3), RPPA analysis (Figure 1B) and IHC (Figure 1D and E). IHC performed on formalin-fixed paraffin-embedded (FFPE) SKBR3 cells revealed a decrease in staining intensity in cells expressing either of two shRNA molecules against PRDX1 (Figure 1D, lower panel) compared to non-targeting or parental cell controls. A significant decrease was seen in percentage 3,3-diaminobenzidine (DAB) positivity of these knockdown cells (Figure 1E). This observation confirmed the specificity of the antibody in the IHC setting prior to staining of clinical specimens. Figure 1F shows representative examples of different intensities of DAB staining, that is, expression of PRDX1 protein on TMA cores. PRDX1 protein expression was found to be predominantly cytoplasmic throughout the cores (Figure 1G). An automated algorithm was used to develop a quantitative scoring model of PRDX1 protein expression, with the respective mark-up image shown. To rule out the possibility of the antibody binding to PRDX2, a protein with high homology for PRDX1, cell lines overexpressing a pLenti6-PRDX2-V5 plasmid were generated. Modulation of PRDX1 expression levels did not affect PRDX2 protein expression, and vice versa (Figure S1D in Additional file 3). An additional PRDX1-targeting antibody was tested [18]; however, this antibody did not satisfactorily detect differential protein expression compared to mRNA expression measured in the shRNA-expressing cell lines (Figure S1B-C in Additional file 3). These extensive validation steps allow us a high level of certainty that the antibody used is specific to PRDX1 in all techniques used throughout the study.

RPPA technology is a particularly useful tool to aid in the identification and validation of protein and phosphoprotein biomarkers using limited amounts of protein from clinical samples. PRDX1 protein expression was assessed on a RPPA cohort with clinical data available for 712 primary human breast tumors. Protein expression data was dichotomized based on the median PRDX1 expression values. High PRDX1 expression was associated with low tumor grade (P <0.001), older age at diagnosis (P <0.001) and human epidermal growth factor receptor 2 (Her2) negativity (P = 0.001), while it displayed a borderline association with positive ER status (P = 0.05) (Table S2 in Additional file 2).

Kaplan-Meier analysis demonstrated that increased levels of PRDX1 protein expression were associated with improved RFS (P = 0.004), OS (P = 0.036) and BCSS (P = 0.005). Interestingly, when various subcohorts were analyzed, high PRDX1 was associated with improved RFS (P = 0.010) and BCSS (P = 0.013) only in the subgroup of ER-positive tumors (Figure 2). Univariate Cox regression analysis demonstrated that, in these ER-positive cases, PRDX1 was associated with improved RFS (hazard ratio (HR) = 0.62, 95% confidence interval (CI) = 0.43 to 0.90, P = 0.011) and BCSS (HR = 0.59, 95% CI = 0.39 to 0.90, P = 0.014) (Table 1). When evaluated by a multivariate Cox proportional hazards model, PRDX1, as assessed by RPPA, was not an independent predictor of RFS or OS (Table 1). However, high PRDX1 protein levels trended toward a significant association with improved BCSS (HR = 0.56, 95% CI = 0.31 to 1.01, P = 0.055).

Although RPPA technology is an ideal platform for initial biomarker discovery, it was necessary to validate these findings in an independent cohort using a more clinically applicable platform, namely IHC (cohort 2). Due to core loss during sectioning, tumors from 442 (88.8%) patients were suitable for analysis. IHC staining was scored on a continuous scale based on staining intensity (Figure 1F), and the median expression level was used to stratify the cohort into high and low PRDX1 protein staining. On the basis of this classification, any possible associations between protein expression and a variety of well-defined clinicopathological variables in the TMA cohort were investigated (Table S3 in Additional file 2). PRDX1 protein expression correlated with smaller tumor size (P = 0.011), low tumor grade (P = 0.002), negative Ki67 status (P = 0.004) and lobular subtype (P = 0.003), whilst there was a borderline significant correlation with positive ER status (P = 0.05).

The relationship between differential expression of PRDX1 and patient survival was subsequently examined. In the overall cohort, increased levels of PRDX1 protein trended toward an improved RFS (P = 0.061), OS (P = 0.100) and BCSS (P = 0.068), but these values were not significant. The positive correlation between high PRDX1 expression and favorable prognosis was again limited to the ER-positive subgroup of cases. Kaplan-Meier analysis demonstrated that, in the ER-positive subgroup, increased levels of PRDX1 protein were associated with an improved RFS (P = 0.011), OS (P = 0.010) and BCSS (P = 0.002) (Figure 3). To compare the prognostic impact of PRDX1 with established factors, Cox regression analysis was performed (Table 2). Univariate Cox regression analysis confirmed that, in ER-positive tumors, high PRDX1 expression associated with improved RFS (HR = 0.60, 95% CI = 0.40 to 0.89, P = 0.012), BCSS (HR = 0.69, 95% CI = 0.52 to 0.92, P = 0.010) and OS (HR = 0.44, 95% CI = 0.24 to 0.75, P = 0.003). Importantly, multivariate analysis within the ER-positive subset showed that PRDX1 was a significant independent predictor of improved RFS (HR = 0.62, 95% CI = 0.40 to 0.96, P = 0.032), BCSS (HR = 0.44, 95% CI = 0.24 to 0.79, P = 0.006) and OS (HR = 0.61, 95% CI = 0.44 to 0.85, P = 0.004), when adjusted for well-established variables such as tumor size, grade, age, nodal, progesterone receptor (PR), Her2 and Ki67 status (Table 2).

RPPA technology can be used to explore cellular signaling pathways activated during cancer progression, such as proteins involved in cellular functions such as growth, proliferation and apoptosis [27]. Utilizing an independent cohort of 410 patients (cohort 3), quantitative protein and phosphoprotein expression levels (165 proteins assessed) were screened to identify alterations in response to differential PRDX1 protein expression. This cohort was stratified into ER-positive and ER-negative cohorts based on the quantitative level of ERα protein expression (Figure 4A). This analysis showed a large number of the proteins as significantly correlated with PRDX1 protein expression in the ER-positive cohort (Table S4 in Additional file 2). Thus, a more stringent approach was necessary to refine the proteins altered in response to differential PRDX1 expression levels, whereby proteins altered between the highest quartile (75th percentile and above) versus the lowest quartile (up to 25th percentile) of PRDX1 expression were selected (Figure 4B). A statistically significant positive correlation was seen between PRDX1 protein and ERα (fold change (FC) = 1.54, P = 0.017). Other proteins that were altered include Claudin-7 (FC = 1.60, P <0.001), HSP70 (FC = 0.61, P <0.001), Collagen VI (FC = 0.61, P = 0.009) and pThr202/pTyr204-MAPK (FC = 0.65, P = 0.008) (Table 3) (represented graphically in Figure 4C). Interestingly, several of the proteins which correlated with PRDX1 expression are known to be regulated by oxidative stress and/or play a role in signaling in the context of ER-positive breast cancer.

Since the hypoxic environment of tumors is likely to contain high levels of ROS [31], induction of oxidative stress in cell lines may be a close mimic of intra-tumoral conditions. In addition, oxidative stress is a very well-known modulator of several oncogenic signaling pathways, including the Akt-mediated pathway, where effects of PRDX1 have been already described in mouse mammary tissue [6]. Therefore, we examined whether breast cancer cell lines exposed to oxidative stress had altered levels of these estrogen and breast cancer signaling proteins, and if modulation of PRDX1 expression could alter any oxidative stress-induced signaling changes. Increased oxidative stress was accomplished via treatment with exogenous H₂O₂ or peroxynitrite. The ER-positive ductal carcinoma cell line, ZR-75-1, was primarily utilized, being a standard model of ER-positive and estrogen-dependent tumors, with low-level oncogenic signaling [32, 33]. In the previous RPPA screen, we observed that ERα and pThr202/pTyr204-MAPK were associated with PRDX1 protein expression in ER-positive tumors (cohort 3). Thus, we set about validating these findings in our model cell line. We observed that H₂O₂ treatment resulted in decreased ERα protein expression in parental ZR-75-1 cells with suppression of the ER activity surrogate, PR, as well as enhanced phosphorylation of serine 473 on Akt, another oncogenic signaling molecule (Figure S2A in Additional file 4).

Subsequent studies demonstrated that oxidative stress-mediated suppression of ERα can be regulated by inducing changes in PRDX1 protein expression. In two independently transduced ZR-75-1 cell lines stably expressing different PRDX1-targeting shRNAs, PRDX1 knockdown-enhanced H₂O₂-mediated suppression of ERα protein (suppressed at 12.5 μM H₂O₂), relative to parental and non-targeting control (NTC) expressing cells (50 μM H₂O₂). Conversely, PRDX1 overexpression rendered the cells resistant to suppression of ERα by oxidative stress (Figure 5A). PRDX1 knockdown also enhanced oxidative stress-induced Akt phosphorylation (Ser473) with phosphorylation occurring at a lower H₂O₂ concentration (Figure 5B). This effect was also seen when cells were treated with the reactive nitrogen species, peroxynitrite (Figure 5C), which was in accordance with previous observations in other ER-positive breast cancer cell lines [34]. Although H₂O₂ induced a decrease of E-cadherin protein expression in breast cancer cells [35] (Figure S2A in Additional file 4), these changes were only minimally affected by changes in PRDX1 expression. Also, H₂O₂ induced only marginal or no changes in pThr202/pTyr204-MAPK (ERK1/2) and ERβ expression regardless of PRDX1 expression levels [36, 37] (Figure 5). This is consistent with the selectivity of the association between PRDX1 and ER expression. ER activity was also assessed using an estrogen transcriptional response element (ERE)-luciferase reporter assay, whereby PRDX1 knockdown could potentiate oxidative stress-mediated suppression of ERα activity (25 μM H₂O₂: P = 0.027; 50 μM: P = 0.037; 100 μM: P = 0.002) (Figure S2B in Additional file 4). In order to determine if PRDX1 expression was driven by ERα promoter activity, we assessed PRDX1 protein expression in ZR-75-1 cells with or without stimulation with the ER ligand, 17β-estradiol. No change in PRDX1 protein content was seen following 17β-estradiol stimulation, and as expected, ERα protein is suppressed upon estradiol stimulation (Figure S2C in Additional file 4). Although ESR1 mRNA expression is decreased by induction of oxidative stress (Figure S3A in Additional file 5), PRDX1 silencing does not enhance this suppression. Alternatively in PRDX1-silenced ZR-75-1 cells, an ESR1 induction is seen at lower H₂O₂ concentrations, which returns to baseline levels above 50 μM H₂O₂. Importantly, Supplementary Figure S3B (Additional file 5) demonstrates that ERα protein levels diminished in these PRDX1-silenced cells across all H₂O₂ concentrations used, which suggests that PRDX1 differentially regulates ERα mRNA and protein expression. ZR-75-1 cells were treated with H₂O₂, and a transcription inhibitor (actinomycin D) (Figure S3C in Additional file 5) or proteasome inhibitor (MG132) in order to determine if oxidative stress affects ERα protein stability rather than mRNA levels. These results showed that inhibition of proteasomal degradation using MG132 prevents oxidative stress-induced ERα suppression and this inhibition is enhanced in the absence of PRDX1 expression (Figure S3D in Additional file 5), which suggests that the proteasomal degradation of ERα via is more active under low-PRDX1 conditions. This implies that PRDX1 may protect ERα from oxidative stress-induced suppression of the protein itself, rather than transcriptional/mRNA stability regulation. In addition, silencing of PRDX1 in ZR-75-1 and T47D cells did not alter the cell proliferation and apoptosis (Figure S4A-D in Additional file 6).

As expected, PRDX1 shRNA-expressing cells showed a reduced potential to metabolize extracellular H₂O₂, similar to the effects of adenanthin, a chemical inhibitor of PRDX1 and PRDX2 antioxidant activity [20] (Figure S2C in Additional file 4). Similar to the effect of exogenous H₂O₂ treatment, inhibition of PRDX1 antioxidant activity using adenanthin suppressed ERα levels (Figure 5D; reducing gel). Interestingly, treatment of T47D and ZR-75-1 cell lines with adenanthin resulted in a decrease of PRDX1 dimerization with a dose-dependent increase in PRDX1 monomer levels (Figure 5D; non-reducing gel), which has not been previously reported. Silencing of PRDX1 or PRDX2 diminishes the adenanthin-induced suppression of ERα protein (Figure S5 in Additional file 7). This suggests that PRDX1 and PRDX2 might both have similar and/or partially overlapping effects in protecting ERα from degradation. Importantly for this study, however, adenanthin was initially described as a more potent inhibitor of PRDX1 than PRDX2, as adenanthin had a significantly lower half-maximum inhibitory (IC₅₀) values for PRDX1 (1.5 μM) compared to PRDX2 (15 μM) [20]. Indeed, MCF7 cells, which express levels of PRDX1 comparable to ZR-75-1 and T47D, but significantly more PRDX2 protein, require higher concentrations of adenanthin in order to suppress ERα (Figure S5D-E in Additional file 7). This observation suggests that PRDX1 is a primary target for adenanthin, and that PRDX2 can partially substitute for PRDX1-mediated protection of ERα. This potentially opens a new avenue for further investigation of PRDX2 in ER-positive breast cancer.