A Splice Variant of NCOR2, BQ323636.1, Confers Chemoresistance in Breast Cancer by Altering the Activity of NRF2

Breast cancer is the most common type of female cancer. Reactive oxygen species (ROS) are vital in regulating signaling pathways that control cell survival and cell proliferation. Chemotherapeutic drugs such as anthracyclines induce cell death via ROS induction. Chemoresistance development is associated with adaptive response to oxidative stress. NRF2 is the main regulator of cytoprotective response to oxidative stress. NRF2 can enhance cell growth, antioxidant expression, and chemoresistance by providing growth advantage for malignant cells. Previously, we identified BQ323636.1 (BQ), a novel splice variant of nuclear co-repressor NCOR2, which can robustly predict tamoxifen resistance in primary breast cancer. In this study, we found that BQ was overexpressed in epirubicin-resistant cells and demonstrated that BQ overexpression could reduce the levels of epirubicin-induced ROS and confer epirubicin resistance. In vivo analysis using tissue microarray of primary breast cancer showed direct correlation between BQ expression and chemoresistance. In vitro experiments showed BQ could modulate NRF2 transcriptional activity and upregulate antioxidants. Luciferase reporter assays showed that although NCOR2 repressed the transcriptional activity of NRF2, the presence of BQ reduced this repressive activity. Co-immunoprecipitation confirmed that NCOR2 could bind to NRF2 and that this interaction was compromised by BQ overexpression, leading to increased transcriptional activity in NRF2. Our findings suggest BQ can regulate the NRF2 signaling pathway via interference with NCOR2 suppressive activity and reveals a novel role for BQ as a modulator of chemoresistance in breast cancer.


Introduction
Breast cancer is the most common malignancy in women. Surgery, radiation therapy, hormonal therapy, chemotherapy, and targeted therapy have greatly improved breast cancer patients' survival. However, a high incidence of recurrence is common in aggressive and advanced diseases [1]. Drug resistance is a major cause of treatment failure in cancer chemotherapy. A better understanding of the mechanisms of chemotherapy failure is needed to improve current therapeutic methods and prediction of patients' outcome.

Epirubicin Resistant Cells Are More Refractory to Oxidative Stress
Chemotherapeutic drugs, such as anthracyclines and platinum compounds, induce cellular damage and cell death in cancer cells partly via the production of reactive oxygen species (ROS) [31]. However, the cytotoxic effect of a chemotherapeutic drug is often compromised in cells with hyperactive antioxidant signaling. Therefore, factors that enhance antioxidant signaling might be able to compromise the effect of chemotherapeutic drugs and promote chemoresistance.
To test this hypothesis, we first compared the ROS levels in the parental MCF-7 epirubicinresistant (MCF-7 EpiR) cells [32] with the parental MCF-7 breast cancer cells. To this end, the ROS levels were determined by CellROX Deep Red Reagent, which is a fluorogenic probe for measuring cellular oxidative stress (Supplementary Figure S1). The results showed that MCF-7 EpiR cells had lower steady state ROS levels ( Figure 1A). On treatment with oxidant tert-Butyl hydroperoxide (tBHP), although the parental MCF-7 cells were sensitive to oxidative stress, MCF-7 EpiR cells were comparably more refractive ( Figure 1A). The results suggested that resistant cells have higher antioxidative stress capacities. Through SRB and clonogenic assays, we confirmed that the MCF-7 EpiR cells had enhanced cell viability under cytotoxic oxidative conditions ( Figure 1B,C) (Supplementary Figure S2).
Likewise, on treatment with epirubicin, MCF-7 EpiR cells were more refractive to epirubicininduced ROS than MCF-7 cells (Figure 2A). The transcription factor NRF2 is a master regulator of oxidative stress and reported to be associated with chemoresistance [33,34]. As antioxidant response element (ARE) is the main transcription binding site recognized by NRF2, an ARE luciferase reporter assay was used to study NRF2 transcription activity (Supplementary Figure S3). We found that MCF-7 EpiR cells displayed higher ARE-associated luciferase activity than MCF-7 cells. This indicated that MCF-7 EpiR cells have higher NRF2 transcriptional activity. Knockdown of NRF2 by specific siRNAs (Supplementary Figure S4  Data was presented as means ± SD and analyzed by two-way ANOVA and found to be significantly different between MCF-7 and MCF-7 EpiR cells, * p < 0.05. (N = 3) (C) MCF-7 and MCF-7 EpiR cells were treated with increasing concentrations of tBHP, and their sensitivity to tBHP was assessed by clonogenic assay. Their clonogenicity in response to tBHP was analyzed by two-way ANOVA and found to be significantly different (** p < 0.01) from one another (N = 3). and Epirubicin-resistant MCF-7 EpiR cells were stained with CellROX Deep Red reagent and analyzed for ROS levels using flow cytometry. Data were analyzed by FlowJo software. The mean fluorescence values were presented as relative ROS level compared to untreated cells (0 µM). Data presented as means ± SD and Student's t-test was used to compare the means: * p < 0.05; ** p < 0.01; *** p < 0.001; n.s. nonsignificant. (N = 3) (B) SRB assay was used to analyze the cell viability of MCF-7 EpiR cells treated with tBHP as compared to MCF-7 cells. Relative cell viability under different concentrations of tBHP treated for 72 h as compared with untreated cells. Data was presented as means ± SD and analyzed by two-way ANOVA and found to be significantly different between MCF-7 and MCF-7 EpiR cells, * p < 0.05. (N = 3) (C) MCF-7 and MCF-7 EpiR cells were treated with increasing concentrations of tBHP, and their sensitivity to tBHP was assessed by clonogenic assay. Their clonogenicity in response to tBHP was analyzed by two-way ANOVA and found to be significantly different (** p < 0.01) from one another (N = 3).

BQ Is Associated with Chemoresistance of Breast Cancer
Given that NCOR2 can inhibit NRF2 activity [27], and that BQ can compete with NCOR2 to regulate ERα transcriptional activity [28,29], we hypothesized that BQ may modulate oxidative stress by enhancing NRF2 activity, and therefore contribute to the development of chemoresistance.
To determine whether BQ may contribute to epirubicin resistance in breast cancer cells, we first compared the expression of BQ in drug-sensitive cells and resistant cells. Western blot was used to determine the expression levels of BQ and NCOR2 in the MCF-7 EpiR, MCF-7 TaxR and parental MCF-7 breast cancer cells. As expected, although BQ levels were higher in the resistant cells, NCOR2 expression was comparatively lower ( Figure 3A), suggesting that high BQ and low NCOR2 expression levels is associated with chemoresistance. To test this conjecture, we next performed immunohistochemical (IHC) staining of BQ on a tissue microarray (TMA) constructed from a cohort of 62 patients with records of chemotherapeutic treatment. Representative images of BQ IHC staining are shown in Figure 3B. Of these the cases with analyzable data, 16 had subsequently suffered cancer relapse, whereas 43 had not. Although the number of cases is relatively small, both Mann-Whitney and Chi square test nevertheless showed statistically significant higher nuclear BQ expression in patients who suffered a relapse (* p = 0.018 and ** p = 0.001, respectively) ( Figure 3C). In addition, and as expected, nuclear BQ expression was enhanced in epirubicin resistant cells (Supplementary Figure S7). These in vivo and in vitro data support our hypothesis that BQ contributes to chemoresistance in breast cancer and its high nuclear expression might be associated recurrence after chemotherapy.
in 6-well plates and treated with increasing doses of epirubicin for 14 days. Their sensitivity to epirubicin was assessed by clonogenic assay. Their clonogenicity in response to epirubicin was analyzed by two-way ANOVA and found to be significantly different (** p < 0.01) from one another. (N = 3) (C) Expression of NRF2 in MCF-7 cells and MCF-7 EpiR cells was detected by Western blot. β-Tubulin served as the loading control. (N = 3).

BQ Is Associated with Chemoresistance of Breast Cancer
Given that NCOR2 can inhibit NRF2 activity [27], and that BQ can compete with NCOR2 to regulate ERα transcriptional activity [28,29], we hypothesized that BQ may modulate oxidative stress by enhancing NRF2 activity, and therefore contribute to the development of chemoresistance.
To determine whether BQ may contribute to epirubicin resistance in breast cancer cells, we first compared the expression of BQ in drug-sensitive cells and resistant cells. Western blot was used to determine the expression levels of BQ and NCOR2 in the MCF-7 EpiR, MCF-7 TaxR and parental MCF-7 breast cancer cells. As expected, although BQ levels were higher in the resistant cells, NCOR2 expression was comparatively lower ( Figure 3A), suggesting that high BQ and low NCOR2 expression levels is associated with chemoresistance. To test this conjecture, we next performed immunohistochemical (IHC) staining of BQ on a tissue microarray (TMA) constructed from a cohort of 62 patients with records of chemotherapeutic treatment. Representative images of BQ IHC staining are shown in Figure 3B. Of these the cases with analyzable data, 16 had subsequently suffered cancer relapse, whereas 43 had not. Although the number of cases is relatively small, both Mann-Whitney and Chi square test nevertheless showed statistically significant higher nuclear BQ expression in patients who suffered a relapse (* p = 0.018 and ** p = 0.001, respectively) ( Figure 3C). In addition, and as expected, nuclear BQ expression was enhanced in epirubicin resistant cells (Supplementary Figure  S7). These in vivo and in vitro data support our hypothesis that BQ contributes to chemoresistance in breast cancer and its high nuclear expression might be associated recurrence after chemotherapy.  . The intensities of NCOR2 and BQ bands were quantified by Image J analysis software which represents their relative protein expressions. Ratios of relative BQ to NCOR2 expression were calculated (right panel). Data presented as means ± SD, ** p < 0.01. (N = 3) (B) Immunohistochemical (IHC) staining of BQ in primary breast cancer samples in tissue microarray. Representative images of low or high BQ nuclear expression were shown. (C) Nuclear BQ expression score was associated with relapse status (Yes for with relapse; No for without relapse) as demonstrated with Mann-Whitney test in primary breast cancer samples; p = 0.018. Nuclear BQ expression score were dichotomized at the median value and Chi square test performed for association with relapse status (Chi-square test p = 0.001).

BQ Overexpression Contributes to Epirubicin Resistance
We next employed clonogenic assay to determine whether BQ overexpression would alter the sensitivity of breast cancer cells to epirubicin. We established two stable BQ-overexpressing cell models from MCF-7 and ZR-75 cell lines (MCF-7 BQ and ZR-75 BQ), respectively. In these cells, BQ overexpression was confirmed by both Qpcr and Western blot (Supplementary Figure S8). The parental and BQ-overexpressing cells were treated with various doses of epirubicin for 14 days. We found that BQ overexpression conferred epirubicin resistance to both MCF-7 and ZR-75 cells ( Figure  4A). Similarly, BQ overexpression could protect cells against cytotoxic oxidative conditions from the results of SRB assays and clonogenic assay with increased cell viability seen in BQ overexpressing cells treated with tBHP (Supplementary Figure S9). To further illustrate that BQ overexpression could protect cells from oxidative stress, ROS levels were measured in BQ overexpressing and parental cells upon epirubicin treatment ( Figure 4B). The results showed that epirubicin treatment induced ROS levels significantly in the empty expression vector harboring cells. The addition of antioxidant NAC reduced the levels ROS induced by epirubicin, suggesting that ROS induced by epirubicin can be depleted by antioxidant. In comparison, in BQ overexpressing cells, the level of ROS induced by epirubicin was significantly reduced. These results suggest that ectopic expression of BQ could limit ROS accumulation induced by chemotherapy.
Consistently, treatment of breast cancer cells with epirubicin induced cellular apoptosis, as revealed by Annexin V Alexa Fluor 647 and SYTOX Blue Dead cell stain double-staining (Supplementary Figure S10). Results showed that BQ overexpression could significantly reduce epirubicin-induced apoptosis. ( Figure 4C). Similarly, combined treatment of NAC and epirubicin in BQ overexpressing cells did not show significant reduction of apoptotic cells compared to control cells, indicating that epirubicin-induced ROS production is involved in epirubicin-induced apoptosis in these cells ( Figure 4C). Although epirubicin treatment induced cleavage of apoptotic markers PARP and Caspase 3, we also found reduced cleavage of these apoptotic markers in BQoverexpressing cells (Supplementary Figure S11). These results suggested that overexpression of BQ . The intensities of NCOR2 and BQ bands were quantified by Image J analysis software which represents their relative protein expressions. Ratios of relative BQ to NCOR2 expression were calculated (right panel). Data presented as means ± SD, * p < 0.05; ** p < 0.01. (N = 3) (B) Immunohistochemical (IHC) staining of BQ in primary breast cancer samples in tissue microarray. Representative images of low or high BQ nuclear expression were shown. (C) Nuclear BQ expression score was associated with relapse status (Yes for with relapse; No for without relapse) as demonstrated with Mann-Whitney test in primary breast cancer samples; * p = 0.018. Nuclear BQ expression score were dichotomized at the median value and Chi square test performed for association with relapse status (Chi-square test p = 0.001).

BQ Overexpression Contributes to Epirubicin Resistance
We next employed clonogenic assay to determine whether BQ overexpression would alter the sensitivity of breast cancer cells to epirubicin. We established two stable BQ-overexpressing cell models from MCF-7 and ZR-75 cell lines (MCF-7 BQ and ZR-75 BQ), respectively. In these cells, BQ overexpression was confirmed by both Qpcr and Western blot (Supplementary Figure S8). The parental and BQ-overexpressing cells were treated with various doses of epirubicin for 14 days. We found that BQ overexpression conferred epirubicin resistance to both MCF-7 and ZR-75 cells ( Figure 4A). Similarly, BQ overexpression could protect cells against cytotoxic oxidative conditions from the results of SRB assays and clonogenic assay with increased cell viability seen in BQ overexpressing cells treated with tBHP (Supplementary Figure S9A-C). To further illustrate that BQ overexpression could protect cells from oxidative stress, ROS levels were measured in BQ overexpressing and parental cells upon epirubicin treatment ( Figure 4B). The results showed that epirubicin treatment induced ROS levels significantly in the empty expression vector harboring cells. The addition of antioxidant NAC reduced the levels ROS induced by epirubicin, suggesting that ROS induced by epirubicin can be depleted by antioxidant. In comparison, in BQ overexpressing cells, the level of ROS induced by epirubicin was significantly reduced. These results suggest that ectopic expression of BQ could limit ROS accumulation induced by chemotherapy.
Consistently, treatment of breast cancer cells with epirubicin induced cellular apoptosis, as revealed by Annexin V Alexa Fluor 647 and SYTOX Blue Dead cell stain double-staining (Supplementary Figure S10). Results showed that BQ overexpression could significantly reduce epirubicin-induced apoptosis. ( Figure 4C). Similarly, combined treatment of NAC and epirubicin in BQ overexpressing cells did not show significant reduction of apoptotic cells compared to control cells, indicating that epirubicin-induced ROS production is involved in epirubicin-induced apoptosis in these cells ( Figure 4C). Although epirubicin treatment induced cleavage of apoptotic markers PARP and Caspase 3, we also found reduced cleavage of these apoptotic markers in BQ-overexpressing cells (Supplementary Figure S11). These results suggested that overexpression of BQ could counteract cellular apoptosis induced by chemotherapy via induced ROS generation, thus reducing the cytotoxic effects of the drugs. Our in vitro results suggested that overexpression of BQ contributes to epirubicin resistance. The involvement of BQ in epirubicin resistance was further confirmed using an in vivo mouse xenograft model ( Figure 4D). We examined the dosage effect of epirubicin on the tumors derived from control and BQ-overexpressing cells. ZR75-empty vector (EV) tumors displayed significant reductions in tumor growth when treated with 2 mg/kg and 5 mg/kg of epirubicin ( Figure 4D(i,ii)), indicating that ZR75 EV tumors were sensitive to epirubicin compared with untreated control. In contrast, tumors with BQ overexpression did not respond to the same range of epirubicin treatment ( Figure 4D(iii,iv)). Taken together, we confirmed that BQ overexpression can confer resistance to epirubicin treatment in breast cancer. could counteract cellular apoptosis induced by chemotherapy via induced ROS generation, thus reducing the cytotoxic effects of the drugs. Our in vitro results suggested that overexpression of BQ contributes to epirubicin resistance. The involvement of BQ in epirubicin resistance was further confirmed using an in vivo mouse xenograft model ( Figure 4D). We examined the dosage effect of epirubicin on the tumors derived from control and BQ-overexpressing cells. ZR75-empty vector (EV) tumors displayed significant reductions in tumor growth when treated with 2 mg/kg and 5 mg/kg of epirubicin ( Figure 4D(i,ii)), indicating that ZR75 EV tumors were sensitive to epirubicin compared with untreated control. In contrast, tumors with BQ overexpression did not respond to the same range of epirubicin treatment ( Figure 4D(iii,iv)). Taken together, we confirmed that BQ overexpression can confer resistance to epirubicin treatment in breast cancer.    4). Representative images of the tumors formed in nude mice treated with or without epirubicin). (ii,iv) Fold changes of tumor volume were illustrated in growth curves. Data presented as mean ± SD and analyzed by two-way ANOVA (* p < 0.05; *** p < 0.001; n.s. nonsignificant).

BQ Overexpression Upregulates the Expression of Antioxidant Enzymes
Given that epirubicin resistant cells are refractory to oxidative stress and that BQ overexpression might contribute to epirubicin resistance, we hypothesized that BQ overexpression could promote the anti-oxidative stress activity of breast cancer cells and thus epirubicin resistance. NRF2 is a master regulator of oxidative stress, controlling the expression of anti-oxidative genes [33]. We have shown previously that BQ can bind to NCOR2 to form a defective co-repressor complex to compromise the suppressive role of NCOR2 [29]. As NCOR2 is a transcriptional repressor of NRF2, it is very likely that BQ overexpression protect cells from oxidative stress through promoting the activity of NRF2.
To probe the link between BQ and NRF2, we evaluated the expression levels of the well-characterized NRF2 downstream antioxidant targets NQO1, HMOX1, GCLC, and GSTP1 by RT-qPCR. The results showed that BQ overexpression led to the upregulation of NQO1, HMOX1 and GCLC at the transcriptional level ( Figure 5A) and NQO1 and HO-1 (gene name: HMOX1) at protein level in both MCF-7 and ZR75 ( Figure 5B). Similarly, epirubicin resistant cells also displayed enhanced expression of NQO1 and HO-1 (Supplementary Figure S12). To further validate our findings in vivo, we performed immunohistochemical (IHC) staining of BQ and NQO1 on TMA constructed from 124 cases of primary breast cancers. Representative images of NQO1 IHC staining were shown in Figure 5C. Nuclear expression of BQ and NQO1 in these clinical samples was scored semiquantitatively as previously described [29]. We found that NQO1 protein expression was positively and significantly correlated with BQ expression (R 2 = 0.4555; p = 0.0173) ( Figure 5D). These results provide in vivo evidence to support our in vitro findings that BQ overexpression upregulates NQO1 expression. Moreover, consistent with our hypothesis, knockdown of NRF2 or NQO1 (Supplementary Figure S13) in BQ overexpressing cells resensitized the cells to epirubicin treatment ( Figure 5E). This further suggests that the effect of BQ on chemoresistance is mediated via NRF2 and NQO1.

BQ Overexpression Does Not Affect NRF2 Expression or Post-Translational Modification
As the transcriptional activity of NRF2 may be regulated by different mechanisms, including its expression level [35], subcellular localization [36], and post-translational modifications, such as Ser-40 phosphorylation, which facilitates the release of NRF2 from its repressor [37] and co-repressors [27], we examined the effect of BQ overexpression on the expression and activity of NRF2. Results showed that BQ overexpression did not alter expression of NRF2, at both mRNA ( Figure 6A) and protein levels ( Figure 6B), nor alter the phosphorylation status of NRF2 Ser-40 in both cell lines ( Figure 6B). Nucleocytoplasmic fractionation showed that BQ overexpression could not enrich NRF2 in the nuclear fraction (Supplementary Figure S14). We also examined the expression of co-repressor NCOR2, a negative regulator that can inhibit the activity of NRF2 [27]. Both qPCR ( Figure 6A) and Western blot results ( Figure 6B) confirmed that BQ overexpression could not alter the expression of NCOR2.

BQ Overexpression Does Not Affect NRF2 Expression or Post-Translational Modification
As the transcriptional activity of NRF2 may be regulated by different mechanisms, including its expression level [35], subcellular localization [36], and post-translational modifications, such as Ser-40 phosphorylation, which facilitates the release of NRF2 from its repressor [37] and co-repressors [27], we examined the effect of BQ overexpression on the expression and activity of NRF2. Results showed that BQ overexpression did not alter expression of NRF2, at both mRNA ( Figure 6A) and protein levels ( Figure 6B), nor alter the phosphorylation status of NRF2 Ser-40 in both cell lines ( Figure 6B). Nucleocytoplasmic fractionation showed that BQ overexpression could not enrich NRF2 in the nuclear fraction (Supplementary Figure S14). We also examined the expression of co-repressor NCOR2, a negative regulator that can inhibit the activity of NRF2 [27]. Both qPCR ( Figure 6A) and Western blot results ( Figure 6B) confirmed that BQ overexpression could not alter the expression of NCOR2.

BQ Modulates the Ability of NCOR2 to Suppress Transcription on Antioxidant Response Element
NCOR2 forms a homodimer in an antiparallel fashion (i.e., N-terminal binds to C-terminals) to serve as a dock for further recruitment of other co-repressor proteins in the co-repressor complex [38]. As a truncated NCOR2 splice variant, BQ retains the only N-terminus of wild type NCOR2 for dimerization with NCOR2. In agreement with this, we have previously demonstrated that BQ can interact with NCOR2 resulting in defective co-repressor activity on the estrogen receptor [29]. As it has been reported that NCOR2 is a co-repressor for transcriptional activity of NRF2 [27], we thus speculated that overexpression of BQ can also result in defective co-repressor activity of NCOR2 on NRF2 ARE element.
To test this hypothesis, we transiently overexpressed NCOR2 in BQ overexpressing cells (Supplementary Figure S15) and measured the luciferase activity driven by antioxidant response element (ARE), an NRF2 response element on the promoter ( Figure 7A). We found that overexpression of NCOR2 alone could suppress ARE driven-luciferase activity, whereas co-expression of NCOR2 with BQ partially released such repression. Therefore, BQ overexpression can compromise the suppressive effect of NCOR2 on the transcriptional activity of NRF2. Through co-immunoprecipitation ( Figure 7B), we confirmed that NCOR2 could indeed interact with NRF2 in both MCF-7 and ZR-75 cells, but that such an interaction was significantly reduced in BQ overexpressing cells. Furthermore, we found BQ was enriched in the NCOR2-immunoprecipitates. This suggested that BQ could interact with NCOR2 to sequester NCOR2 from its interaction with NRF2. Thus, our results confirm that BQ overexpression enhances NRF2-mediated gene expression via inhibiting the interaction between NCOR2 and NRF2.

BQ Modulates the Ability of NCOR2 to Suppress Transcription on Antioxidant Response Element
NCOR2 forms a homodimer in an antiparallel fashion (i.e., N-terminal binds to C-terminals) to serve as a dock for further recruitment of other co-repressor proteins in the co-repressor complex [38]. As a truncated NCOR2 splice variant, BQ retains the only N-terminus of wild type NCOR2 for dimerization with NCOR2. In agreement with this, we have previously demonstrated that BQ can interact with NCOR2 resulting in defective co-repressor activity on the estrogen receptor [29]. As it has been reported that NCOR2 is a co-repressor for transcriptional activity of NRF2 [27], we thus speculated that overexpression of BQ can also result in defective co-repressor activity of NCOR2 on NRF2 ARE element.
To test this hypothesis, we transiently overexpressed NCOR2 in BQ overexpressing cells (Supplementary Figure S15) and measured the luciferase activity driven by antioxidant response element (ARE), an NRF2 response element on the promoter ( Figure 7A). We found that overexpression of NCOR2 alone could suppress ARE driven-luciferase activity, whereas coexpression of NCOR2 with BQ partially released such repression. Therefore, BQ overexpression can compromise the suppressive effect of NCOR2 on the transcriptional activity of NRF2. Through coimmunoprecipitation ( Figure 7B), we confirmed that NCOR2 could indeed interact with NRF2 in both MCF-7 and ZR-75 cells, but that such an interaction was significantly reduced in BQ overexpressing cells. Furthermore, we found BQ was enriched in the NCOR2-immunoprecipitates. This suggested that BQ could interact with NCOR2 to sequester NCOR2 from its interaction with NRF2. Thus, our results confirm that BQ overexpression enhances NRF2-mediated gene expression via inhibiting the interaction between NCOR2 and NRF2.  The effect of NCOR2 and BQ on the activity of ARE was examined by luciferase assay. Luminescence readings were normalized to renilla luciferase activity (as transfection efficiency control). The data were presented as % change of luciferase activity with reference to empty vectors control cells (i.e., BQ and NCOR2 negative). Bars showed as mean ± SD. (Student's t-test, * p < 0.05; ** p < 0.01) (N = 3) (B) The effect of BQ overexpression on the protein interaction between NRF2 and NCOR2. Co-immunoprecipitation was employed to determine the protein-protein interaction. BQ overexpressing cells and control cells were harvested for immunoprecipitation by anti-NCOR2 antibody (IP: NCOR2). Whole cell lysates were used as input. Normal rabbit Immunoglobulin (IgG) was used as pull-down control. (i) Immunoblotting by anti-NCOR2, anti-NRF2 and anti-His antibodies were shown. Corresponding bands were indicated by arrows. The intensities of the NRF2 band were quantified by ImageJ software. (ii) Bars shown relative bindings of NCOR2 to NRF2 at protein level, presented as mean ± SD. (Student's t-test, * p < 0.05; *** p < 0.001). (N = 3).

Discussion
Anthracyclines is one of the most active and widely used chemotherapeutic agents for breast cancer [39]. The development of chemoresistance poses difficulties to cancer therapy as chemoresistant cancer cells become more difficult to eliminate using the same chemotherapeutics. Even though patients might have high initial response rates, they might eventually relapse due to acquired resistance. Therefore, chemoresistance reduces the number of treatment options for advanced diseases. Most anticancer drugs can induce the formation of ROS, which triggers cell apoptosis via the mitochondrial pathway initiated by cytochrome c release or activation of death receptor cascade [31,40]. Altered redox signaling and oxidative stress response could contribute to chemoresistance. Several cellular defense mechanisms against oxidative stress have been shown to be upregulated in cancer cells by NRF2. They include the thioredoxin/thioredoxin reductase (Trx/TrxR) system, peroxiredoxins (Prxs), and several glutathione S-transferases (GSTs) [41]. Consistently, constitutive NRF2 activation can protect cancer cells from chemotherapy [21]. Aberrant NRF2 activation has been reported [42] to be mediated by a number of mechanisms, including somatic mutation of NRF2 and KEAP1 [43], DNA methylation of KEAP1 gene promoter [44], p62 accumulation which compete with NRF2 for KEAP1 [45] and transcriptional activation of NRF2 gene by other oncogenic factors (such as K-Ras, B-Raf, Myc, and mutant p53) [46][47][48][49]. However, the The effect of NCOR2 and BQ on the activity of ARE was examined by luciferase assay. Luminescence readings were normalized to renilla luciferase activity (as transfection efficiency control). The data were presented as % change of luciferase activity with reference to empty vectors control cells (i.e., BQ and NCOR2 negative). Bars showed as mean ± SD. (Student's t-test, * p < 0.05; ** p < 0.01) (N = 3) (B) The effect of BQ overexpression on the protein interaction between NRF2 and NCOR2. Co-immunoprecipitation was employed to determine the protein-protein interaction. BQ overexpressing cells and control cells were harvested for immunoprecipitation by anti-NCOR2 antibody (IP: NCOR2). Whole cell lysates were used as input. Normal rabbit Immunoglobulin (IgG) was used as pull-down control. (i) Immunoblotting by anti-NCOR2, anti-NRF2 and anti-His antibodies were shown. Corresponding bands were indicated by arrows. The intensities of the NRF2 band were quantified by ImageJ software. (ii) Bars shown relative bindings of NCOR2 to NRF2 at protein level, presented as mean ± SD. (Student's t-test, * p < 0.05; *** p < 0.001). (N = 3).

Discussion
Anthracyclines is one of the most active and widely used chemotherapeutic agents for breast cancer [39]. The development of chemoresistance poses difficulties to cancer therapy as chemoresistant cancer cells become more difficult to eliminate using the same chemotherapeutics. Even though patients might have high initial response rates, they might eventually relapse due to acquired resistance. Therefore, chemoresistance reduces the number of treatment options for advanced diseases. Most anticancer drugs can induce the formation of ROS, which triggers cell apoptosis via the mitochondrial pathway initiated by cytochrome c release or activation of death receptor cascade [31,40]. Altered redox signaling and oxidative stress response could contribute to chemoresistance. Several cellular defense mechanisms against oxidative stress have been shown to be upregulated in cancer cells by NRF2. They include the thioredoxin/thioredoxin reductase (Trx/TrxR) system, peroxiredoxins (Prxs), and several glutathione S-transferases (GSTs) [41]. Consistently, constitutive NRF2 activation can protect cancer cells from chemotherapy [21]. Aberrant NRF2 activation has been reported [42] to be mediated by a number of mechanisms, including somatic mutation of NRF2 and KEAP1 [43], DNA methylation of KEAP1 gene promoter [44], p62 accumulation which compete with NRF2 for KEAP1 [45] and transcriptional activation of NRF2 gene by other oncogenic factors (such as K-Ras, B-Raf, Myc, and mutant p53) [46][47][48][49]. However, the underlying molecular mechanism by which NRF2 transcriptional activity is enhanced to confer drug resistance in breast cancer cells is not well understood.
Our study provides evidence to support the idea that BQ can induce chemoresistance in breast cancer by altering the transcriptional suppressive function of NCOR2 on NRF2. Based on our current findings and previous findings [29], we propose a mechanism whereby BQ can contribute to chemoresistance via modulation of oxidative stress (Figure 8). In this model, in the absence of BQ, NCOR2 self-dimerizes to form a central platform to recruit other corepressor components to form a functional NCOR2 corepressor complex. This repressive complex binds to NRF2 and suppresses its transactivation of target antioxidant genes via the ARE ( Figure 8A). Conversely, in the presence of BQ, BQ complexes with NCOR2 and sequesters it from forming transcriptional suppressive complex with NRF2. This leads to defective co-repressor activity of NCOR2 on NRF2 ARE element, and increased induction of ARE-mediated transcription of antioxidant enzymes ( Figure 8B). The enhanced expression of antioxidant enzymes can eliminate the excessive cytotoxic ROS induced by chemotherapeutic drugs. Thus, BQ overexpression can provide survival advantages to breast cancer cells under chemotherapy. Based on the role of BQ in modulating oxidative stress, a novel oncogenic role for BQ in contributing to epirubicin resistance is proposed. This current study strengthens the potential value of BQ not only as a biomarker for tamoxifen resistance in breast cancer but also could serve as a potential predictor of chemoresistance.
Cancers 2020 16 of 24 underlying molecular mechanism by which NRF2 transcriptional activity is enhanced to confer drug resistance in breast cancer cells is not well understood.
Our study provides evidence to support the idea that BQ can induce chemoresistance in breast cancer by altering the transcriptional suppressive function of NCOR2 on NRF2. Based on our current findings and previous findings [29], we propose a mechanism whereby BQ can contribute to chemoresistance via modulation of oxidative stress (Figure 8). In this model, in the absence of BQ, NCOR2 self-dimerizes to form a central platform to recruit other corepressor components to form a functional NCOR2 corepressor complex. This repressive complex binds to NRF2 and suppresses its transactivation of target antioxidant genes via the ARE ( Figure 8A). Conversely, in the presence of BQ, BQ complexes with NCOR2 and sequesters it from forming transcriptional suppressive complex with NRF2. This leads to defective co-repressor activity of NCOR2 on NRF2 ARE element, and increased induction of ARE-mediated transcription of antioxidant enzymes ( Figure 8B). The enhanced expression of antioxidant enzymes can eliminate the excessive cytotoxic ROS induced by chemotherapeutic drugs. Thus, BQ overexpression can provide survival advantages to breast cancer cells under chemotherapy. Based on the role of BQ in modulating oxidative stress, a novel oncogenic role for BQ in contributing to epirubicin resistance is proposed. This current study strengthens the potential value of BQ not only as a biomarker for tamoxifen resistance in breast cancer but also could serve as a potential predictor of chemoresistance. Figure 8. Proposed mechanism of how BQ may contribute to chemoresistance via the regulation of NRF2 in breast cancer. (A) Without BQ expression, NCOR2 self-dimerizes in antiparallel fashion, forming a central platform to recruit other corepressor components, such as GPS2, TBL1, and HDACs, to form a functional NCOR2 corepressor complex. The NCOR2 corepressor complex interacts with NRF2 to suppress the transcriptional activity of ARE which is used to regulate the expression of antioxidant genes. (B) In the presence of BQ nuclear overexpression, BQ binds with NCOR2This interaction compromises the ability of NCOR2 to fully recruit other corepressor components such as HDACs. This gives rise to a non-fully functional NCOR2 corepressor complex, thus relieving the suppressive effect on NRF2. Upregulation of ARE mediated transcription of antioxidant genes is resulted.
Alternative splicing has recently been viewed as a new hall mark of cancer [50]. Research has now uncovered various alternative spliced isoforms implicated in breast cancer, including BRCA1, DMTF1, FGFR, HER2, KLF6, Survivin, and TP53 [51]. Abnormal expression of splice factors in cancer changes the alternative splicing of pre-mRNAs. Deregulation of these splicing factors has been linked to breast cancer development [52]. More than 30 human splicing factors have now been defined, and some of them are characterized as either proto-oncogenes, such as SRSF1, SRFS3, DAM1, and PELP1, or tumor suppressors, such as RBM and PRMT6, in breast cancer [52]. From our result in Figure 3A, Figure 8. Proposed mechanism of how BQ may contribute to chemoresistance via the regulation of NRF2 in breast cancer. (A) Without BQ expression, NCOR2 self-dimerizes in antiparallel fashion, forming a central platform to recruit other corepressor components, such as GPS2, TBL1, and HDACs, to form a functional NCOR2 corepressor complex. The NCOR2 corepressor complex interacts with NRF2 to suppress the transcriptional activity of ARE which is used to regulate the expression of antioxidant genes. (B) In the presence of BQ nuclear overexpression, BQ binds with NCOR2This interaction compromises the ability of NCOR2 to fully recruit other corepressor components such as HDACs. This gives rise to a non-fully functional NCOR2 corepressor complex, thus relieving the suppressive effect on NRF2. Upregulation of ARE mediated transcription of antioxidant genes is resulted.
Alternative splicing has recently been viewed as a new hall mark of cancer [50]. Research has now uncovered various alternative spliced isoforms implicated in breast cancer, including BRCA1, DMTF1, FGFR, HER2, KLF6, Survivin, and TP53 [51]. Abnormal expression of splice factors in cancer changes the alternative splicing of pre-mRNAs. Deregulation of these splicing factors has been linked to breast cancer development [52]. More than 30 human splicing factors have now been defined, and some of them are characterized as either proto-oncogenes, such as SRSF1, SRFS3, DAM1, and PELP1, or tumor suppressors, such as RBM and PRMT6, in breast cancer [52]. From our result in Figure 3A, the ratio of BQ to NCOR2 was altered in epirubicin resistant breast cancer cells. As BQ is an alternative splicing product of NCOR2, our findings also underscore the fact that deregulation of splicing factors might induce drug resistance. However, the factors for regulation of BQ production are as unclear. Once these specific splicing factor(s) responsible for BQ production is identified, targeting such factors could be a new therapeutic approach for breast cancer.
In summary, this study has characterized a new molecular mechanism involved in the regulation of NRF2 mediated by BQ, and provides a better understanding of the molecular basis for the regulation of oxidative stress and the development of chemoresistance in breast cancer. Our present work also suggests that BQ could be a novel therapeutic target as well as a predictive marker for chemoresistance.

Luciferase Reporter Assay
Cells were co-transfected with pGL3-ARE and Renilla luciferase control plasmid pRL-SV40 using Lipofectamine 3000 (Invitrogen, Carlsbad, CA, USA). For promoter analysis, cells were collected 24 h post-transfection, washed twice in PBS, and harvested for Firefly/Renilla luciferase assays using the Dual-Luciferase reporter assay system (Promega, Madision, WI, USA) according to manufacturer's instruction. Luminescence was then read using Infinite 200 Pro microplate plate reader (TECAN, Mannedorf, Switzerland).

Real-Time Quantitative PCR
Total RNA was extracted with the RNeasy Mini Kit (Qiagen, Hilden, Germany) following the manufacturer's protocol. Up to 1000 ng total RNA was reverse transcribed into cDNA by PrimeScript RT Reagent Kit (Takara, Kusatsu, Japan) following the manufacturer's protocol. SYBR green-based real-time PCR reaction was carried out using ABI 7900HT Real-Time PCR machine (Applied Biosystems, Foster City, CA, USA). Relative expression levels were determined using the 2 −∆∆Ct method. Primer sequences used in this study were listed in Table S1.

Western Blot Analysis
Cell lysates were prepared for Western blot analysis as previously described [54]. The proteins were probed with primary antibody against BQ323636.1 [29],

Sulforhodamine B (SRB) Assay
A total of 4000 cells were seeded in each well of a 96-well plate. Cells were treated with tert-butyl hydroperoxide (tBHP) (Sigma-Aldrich, St Louis, MO, USA) the next day. After culture for 1 to 5 days, cells were fixed with 100 µL of trichloroacetic acid Sigma-Aldrich, St Louis, MO, USA) for 1h at 4 • C. Plates were then washed five times with running water and 100 µL of SRB solution (0.4% SRB in 0.1% acetic acid) was added. The plates were stained for 1 h at room temperature and rinsed three times with 1% acetic acid and air-dried. One-hundred microliters of 10 mM Tris base solution was then added to each well to dissolve the protein-bound dye, and optical density at 492 nm was measured in a microplate reader (TECAN, Mannedorf, Switzerland).

Clonogenic Assay
A total of 1000 cells were seeded into each well of the six-well plates. Cells were then treated with tBHP or epirubicin (EBEWE, Unterach, Austria) and grown for 14 days until visible colonies formed. Colonies were rinsed three times with PBS and fixed with 4% PFA for 20 min at room temperature. The colonies were stained with 0.5% crystal violet (Sigma-Aldrich, St Louis, MO, USA) and air-dried. Two milliliters of 33% acetic acid was used to dissolve bound crystal violet. Optical density was then measured at 492 nm using a microplate reader (TECAN, Mannedorf, Switzerland).

ROS Measurement by Flow Cytometry
Detection of ROS in cells was done by CellROX Deep Red Flow Cytometry Assay Kit (Molecular Probes, Eugene, OR, USA) according to manufacturer's instructions. The CellROX Deep Red reagent is a cell permeable dye that does not fluoresce in a reduced state while it becomes fluorescent upon oxidation by ROS. The emission maxima of CellROX Deep Red is approximately 665 nm and is measurable by flow cytometry. Briefly, cells were treated with epirubicin (EBEWE, Unterach, Austria). After treatment, cells were trypsinized and counted by hemocytometer. 1 × 10 6 cells/mL were resuspended in cell culture medium (without phenol red). Cells were stained with CellROX Deep Red reagent in the 37 • C incubator for 1 h. The samples were immediately analyzed by flow cytometry by LSR Fortessa analyser (BD Biosciences, NJ, USA). The data were analyzed by the FlowJo analysis software (FlowJo LLC, Ashland, OR, USA) and the values for mean fluorescence intensity, which represent the relative ROS levels, were obtained.

Flow cytometric Determination of Apoptosis
Cells were analyzed by Annexin V Alexa Fluor 647 stain (Invitrogen, Carlsbad, CA, USA) plus SYTOX Blue Dead cell stain (Invitrogen, Carlsbad, CA, USA) double-staining method according to manufacturer's protocol. The cells after designed treatments were collected and subjected to the analysis. A minimum of 10,000 events were then analyzed by LSR Fortessa analyzer (BD Biosciences, NJ, USA) with FlowJo analysis software (FlowJo LLC, Ashland, OR, USA) for acquisition and analysis.

Immunoprecipitation
Cell were lysed in IP buffer (1% NP-40, 150 mM NaCl, 50 mM Tris-HCl pH 7.4, 10 mM NaF, 1 mM Na 3 VO 4 , 10 mM N-ethyl-amide and protease inhibitors [Complete protease inhibitor cocktail; Roche, Mannheim, Germany]) and precleared with 30 µL of Dynabeads Protein A (Invitrogen, Carlsbad, CA, USA) for 4 h at 4 • C. Protein concentration of the precleared lysates were determined and the lysate was immunoprecipitated with the indicated antibodies or IgG negative control (Dako, Glostrup, Denmark) for overnight at 4 • C. Forty microliters of Dynabeads Protein A was added to the lysates and incubated for a further 4 h at 4 • C. The beads were washed five times with ice-cold PBS, followed by boiling at 100 • C in 2× SDS loading dye for 10 min to elute the proteins. Eluted proteins were analyzed by Western blotting. The intensities of the protein bands were quantified by ImageJ software. Ratio of (IP: NCOR2) to (INPUT) was calculated. This value represented the relative bindings of NCOR2 to NRF2 at protein level.

Subcellular Fractionation
Subcellular fractionation was performed using the NE-PER TM Nuclear Cytpoplasmic Extraction reagents (Thermo Scientific, Rockford, IL, USA) following manufacturers' protocol. Protein concentrations of the cytoplasmic and nuclear extracts were determined. The extracts were then mix and boiled with 5X SDS loading dye for subsequent Western blotting analysis.

Orthotopic Mouse Model
Female nude mice of 5-6 weeks old were used for this study. On the day of inoculation, approximately 1 × 10 7 ZR-75 EV or ZR-75 BQ cells were injected to the abdominal mammary fat pad of the mouse. When the tumors were palpable, these mice were randomized into epirubicin treatment and untreated control groups. The treatment groups received epirubicin (EBEWE, Unterach, Austria) diluted in PBS by intraperitoneal injection, whereas the control group received PBS only. Tumor sizes were measured regularly using calipers and the tumor volumes were calculated as longest diameter × (shortest diameter) 2 /2. At the end of experiment, mice were euthanized, and tumors harvested. All the procedures were reviewed and approved by HKU Committee on the Use of Live Animals in Teaching and Research (CULATR No: 5103-19).

Tissue Microarray
One-hundred-and-forty-one cases of breast cancer diagnosed between the years 1993 to 2003 with clinical follow up data were retrieved from the records of the Department of Pathology, Queen Mary Hospital of Hong Kong, with approval (UW 08-147) by the Institutional Review Board of The University of Hong Kong. Histological sections of these cases were reviewed by the pathologist, the representative paraffin tumor blocks were chosen as donor block for each case. Selected areas were marked for construction of tissue microarray (TMA) blocks. A total of 124 could be assessed and scored for both BQ and NQO1 expression (Table S2). Similarly, 62 cases of breast cancer diagnosed between the years 1993 to 2003 with clinical follow up data and with record of having received chemotherapeutic treatment were retrieved and analyzed for BQ expression and relapse status (Table S3).

Immunohistochemistry
The immunohistochemistry procedures were described previously [29]. TMA sections were deparaffinized and rehydrated with repeated incubation of ethanol and xylene; 0.01 M citrate buffer (pH 6) was used for antigen retrieval and the slides were then immersed in 3% H 2 O 2 (in methanol) at room temperature for quenching endogenous peroxidase. The slides were then rinsed in 1× PBS with 0.05% Tween-20 (PBST) twice, followed by incubation with primary antibodies at 4 • C overnight. Antibodies used were anti-BQ323636.1 (1:50) [29] and anti-NQO1 (1:200) (#sc-32793, Santa Cruz, CA, USA). The slides were washed with 1 x PBST and incubated with EnVision+ System-HRP labelled Polymer Anti-Rabbit (Dako, Glostrup, Denmark) at room temperature for 30 min in dark. The excess reagent was washed off with 1× PBST and Chromogen DAB/substrate reagent (Dako Glostrup, Denmark) was added to the slides and incubated for 6 min. The slides were dehydrated and mounted. To visualize and assess protein expressions, Aperio ScanScope system (Aperio technology, USA) was used. TMA slides were scanned, and individual TMA spots were assessed using Aperio's image viewer, ImageScope. The intensities and percentages of BQ and NQO1 nuclear staining were scored as previously described [29] by two independent individuals.

Statistical Analysis
Statistical analysis was conducted GraphPad Prism version 6 (GraphPad Software Inc, San Diego, CA, USA) and SPSS version24 (IBM, Armonk, NY, USA). Results were expressed as mean ± SD from at least three independent experiments. Statistical differences between two groups' means were evaluated by two-tailed Student's t-test. Two-way analysis of variance (ANOVA) was used between groups of 2 variables. Significant difference was considered when p-value: * p < 0.05, ** p < 0.01 and *** p < 0.001. The correlation between BQ323636.1 with relapses was analyzed by Mann-Whiney U Rank test analysis. p < 0.05 was considered statistically significant. The data was also dichotomized into two groups for protein expression, as high or low expressers with median expression level as the cutoff. The correlations were analyzed by chi square test. The correlation between BQ323636.1 and NQO1 expression was analyzed by linear regression analysis. p < 0.05 was considered statistically significant.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6694/12/3/533/s1, Figure S1: ROS levels of MCF-7 and MCF-7 EpiR under tBHP treatment, Figure S2: The effect of tetr-Butyl-Butyl hydorperoxide (tBHP) on cell growth of MCF-7 and MCF-7 EpiR, Figure S3: ARE-transcriptional activity of MCF-7 and MCF-7 EpiR cells, Figure S4: Knockdown of NRF2 in MCF-7 EpiR cells, Figure S5: Treatment of Epirubicin on MCF-7 under normal culture condition, Figure S6: Subcellular localization of NRF2 in MCF-7 and MCF-7 EpiR cells, Figure S7: Subcellular localization of BQ in MCF-7 and MCF-7 EpiR cells, Figure S8: Overexpression of BQ323636.1 in MCF-7 and ZR-75 cells, Figure S9: The effect of BQ overexpression on protecting breast cancer cells from oxidative stress, Figure S10: The effect of BQ overexpression on epirubicin-induced apoptosis, Figure S11: The effect of BQ overexpression on apoptotic markers PARP and Caspase 3 under epirubicin treatment, Figure S12: Protein Expression of antioxidant enzymes in MCF-7 and MCF-7 EpiR cells, Figure S13: Knockdown of NRF2 or NQO1 in BQ overexpressing cells, Figure S14: The effect of BQ overexpression on the expression of NRF2 and NCOR2 in different subcellular fractions, Figure S15: Overexpression of NCOR2 in BQ overexpressing cells, Figure S16: Original blots for the cropped/composite Western blot images, Figure S17: Protein quantification using ImageJ from the cropped/composite western blot images, Table S1: RT-PCR Primer sequences used in this study, Table S2: Clinical characteristics of 124 cases of primary breast cancers analyzed in this study, Table S3: Clinical characteristics of 62 breast cancer patients with record of having received chemotherapeutic treatment analyzed in this study.