BQ323636.1, a Novel Splice Variant to NCOR2, as a Predictor for Tamoxifen-Resistant Breast Cancer Free

Breast cancer is the most common female malignancy. The estrogen receptor (ER) signaling pathway is a fundamental proproliferative pathway in the context of breast cancer (1). Upon its activation by binding with estrogen, ER activates target gene transcription and cell growth either directly through its genomic pathway or indirectly through nongenomic pathway that involves the PI3K/AKT pathway. Most of the well-characterized ER-target genes are oncogenic that could promote cancer cell proliferation or apoptosis resistance; for example, cyclin D1, antiapoptotic Bcl2, and pS2 (2–4).

Around 70% of patients with breast cancer are ER⁺ and can benefit from antiestrogen therapy. Tamoxifen is a selective estrogen receptor modulator (SERM) that acts as an antagonist of estrogen in the context of breast and has been the most commonly prescribed antiestrogen drug for preventing patients who are ER⁺ from cancer relapse or metastasis for over four decades (5). Binding of tamoxifen to ER triggers the recruitment of nuclear corepressor 2 (NCOR2) as well as other corepressors, such as GPS2, TBLR1, HDAC3, to suppress the proproliferative ER signaling pathway (6–8). In this corepressive machinery, NCOR2 interacts with itself between regions encompassing amino acids 290 to 427 and 1,788 to 1,903 to form dimers in an antiparallel fashion, serving as a central platform on which additional corepressor proteins assemble (9). The NCOR2 protein consists of four known repression domains within the N-terminal portion (denoted RD1 to RD4) and a series of two C-terminal nuclear receptor interaction domains (denoted S1 and S2; refs. 10, 11). Alternative splicing of NCOR2 at the C-terminal is commonly reported, generating multiple corepressor isoforms with conserved repression domains but with different affinities for different nuclear receptors (12, 13).

In the adjuvant setting, a large number of patients who have ER⁺ tumors are treated with tamoxifen for prevention of disease relapse and metastasis. Despite the success of tamoxifen, resistance is an outstanding problem with up to 50% of nonresponding patients with many initial responders experiencing relapse (14). However, there is no clinically in-use biomarker for prediction of patients' response to tamoxifen and the mechanisms for developing tamoxifen resistance are still not well understood.

By SpliceArray profiling, our group has reported a novel splice variant of NCOR2, named BQ323636.1 (BQ in short), with truncation at the C-terminus, retaining only the RD1 (15). In a cohort of 77 patients with breast cancer, at the mRNA level, BQ323636.1 overexpression was significantly associated with tamoxifen resistance, poor overall, and disease-free survival (15). The result led us to postulate that this splice variant could serve as a biomarker for prediction of tamoxifen response. BQ is capable of modulating transcription of ER by counteracting the transcription repressive activity of NCOR2 (15), suggesting that it could confer tamoxifen resistance and mediate cancer cell growth through competing and counteracting the repressive functions of its wild-type.

Our previous study only focused on the association of BQ32636.1 with tamoxifen resistance at the mRNA level because there was no commercially antibody available to distinguish this isoform from its wild-type and the molecular mechanism through which BQ could confer resistance to tamoxifen was unknown. In the present study, we generated a mouse monoclonal antibody specific for BQ323636.1, confirmed the role of BQ323636.1 in conferring tamoxifen resistance at protein level using both in vitro and in vivo models, and characterized the molecular mechanism through which BQ323636.1 confers resistance to tamoxifen. Most importantly, we also confirmed in two independent breast cancer cohorts from Hong Kong and the United Kingdom that BQ, applied in immunohistochemical (IHC) staining on formalin-fixed paraffin-embedded primary breast cancer tissue, could be a reliable biomarker in predicting tamoxifen resistance.

Human breast cancer cell lines MCF-7 and ZR-75-1 were purchased from the ATCC and were reauthenticated by short tandem repeat profiling (15). MCF-7 was cultured in Dulbecco's Modified Eagle's Medium supplemented with 10% FBS and 1% penicillin/streptomycin. ZR-75-1 was cultured in Improved Minimum Essential Medium (IMEM) supplemented with 5% FBS and 1% penicillin/streptomycin. LCC2 and AK-47 are two tamoxifen-resistant cell lines derived from MCF-7 and ZR-75-1, respectively, kindly provided by Dr. Robert Clarke (Georgetown University Medical School, Washington, DC; ref. 16) and used in our previous study (15). LCC2 and AK-47 were both cultured in IMEM supplemented with 5% charcoal-stripped FBS and 1% penicillin/streptomycin. MCF-10A was purchased from ATCC and cultured in MEGM Mammary Epithelial Cell Growth Medium (Lonza) with 100 ng/mL cholera toxin, 0.5 mg/mL hydrocortisone, 20 ng/mL EGF, 10 μg/mL insulin, 5% FBS, and 1% penicillin/streptomycin. All the cell lines used in this study have been passaged and kept fewer than 6 months after the reauthentication or thawing.

TRIzol reagent (Invitrogen) was used for total RNA extraction following the manufacturer's protocol. Up to 2 μg total RNA was reverse transcribed into cDNA by SuperScript III reverse transcriptase (Invitrogen) following the manufacturer's protocol. Real-time PCR reaction was carried out with the ABI 7900HT Fast Real-time PCR system. The primer sequences used were presented in Supplementary Table S1.

Cell lysates were prepared as previously described (17). The proteins were probed with antibody against BQ3236363.1 (Patent Cooperation Treaty filed), NCOR2 (ab24551, Abcam), ERα (HC-20, Santa Cruz), 6xHis-tag (9F2, Wako), β-tubulin (H-235, Santa Cruz), lamin B (C-20, Santa Cruz), and β-actin (AC-74, Sigma). Secondary antibodies were: HRP-conjugated anti-rabbit antibody (P0448, Dako), HRP-conjugated affiniPure anti-mouse antibody (Jackson Immnuno Research Laboratories), and HRP-conjugated anti-goat antibody (SC2922, Santa Cruz). Protein A–HRP (ThermoFisher Scientific) was used to avoid detection of the heavy and light chains. The images were digitalized and, if necessary, processed by adjusting brightness or contrast only.

Cell viability was studied by MTT assay as previously described (15). 4-Hydroxy tamoxifen (4-OH tamoxifen), an active metabolite of tamoxifen was used. 4-OH tamoxifen and 17-β-estradiol** (E2) were purchased from Sigma and were dissolved in ethanol for in vitro assay.

Cells were lysed in IP lysis buffer and precleared with 30 μL of Dynabeads Protein A/G by rotating at 4°C for 4 hours. After preclearing, protein concentration was measured and the lysate split into equal amount of proteins per tube and incubated with specific primary antibody or IgG-negative control at 4°C for O/N with gentle rotating. On the second day, 40 μL of beads were added to the mixture and incubated at 4°C for another 4 hours. After incubation, the beads were washed five times with cold PBS and boiled at 100°C in 1× SDS loading dye for 5 minutes to elute the proteins. Proteins were separated by SDS–PAGE gel electrophoresis, transferred to nitrocellulose membrane, and hybridized with the antibodies same as Western blotting.

Female nude mice, ages 5 to 6 weeks, were used for this study. On the day of inoculation, 1 × 10⁶ ZR-75-1 cells or 5 × 10⁶ MCF-7 cells were mixed with Matrigel (BD Bioscience) at a ratio of 1:1 and the 100-μL cell mixture was injected into the abdominal mammary fat pad of mice. When the tumors were palpable, mice were randomized into treatment and control groups where treatment group received 4-OH tamoxifen (Sigma) dissolved in ethanol and diluted in peanut oil (Sigma), given by subcutaneous injection 500 μg per day (500 μg/mL) for 6 consecutive days per week. The control group received solvent only. The tumor sizes were measured regularly using calipers and the tumor volume calculated as longest diameter × (shortest diameter)²/2. At the endpoint of the experiment, mice were euthanized, and tumors were harvested. All the procedures have been reviewed and approved by HKU Committee on the Use of Live Animals in Teaching and Research (CULATR No.:3259-14).

Before pulse labeling, the cells were incubated in plain DMEM without methionine and cysteine (DMEM –met/cys, Gibco) for 1 hour. While waiting, ³⁵S-methoinine stock (43.3 mCi/mL, Perkin-Elmer) was diluted in DMEM –met/cys supplemented with 10% FBS (working concentration was 0.2 mCi per 100-mm culture dish (hot medium). After cell starvation, hot medium was added and the cells labeled with ³⁵S for 1 hour. After the pulse period, the hot medium was removed, cells were washed and harvested at indicated time points (chase period). Cell pellets were collected and lysed for immunoprecipitation (IP) using anti-NCOR2 antibody (Abcam) or anti-BQ323636.1 antibody (PCT filed) and the samples analyzed by SDS-PAGE.

Two independent cohorts totaling 2,095 patients with breast cancer with pathological and clinical follow-up data of over 20 years were used for this study. The cohort from Hong Kong comprised 234 cases of breast cancer diagnosed between the years 1992 to 2008 retrieved from the records of the Department of Pathology, Queen Mary Hospital of Hong Kong, with approval by the Institutional Review Board of The University of Hong Kong (UW 08-147). The remaining 1,861 cases were obtained from Nottingham University Hospital, which consisted of a large cohort of patients comprising a well-characterized consecutive series of early-stage sporadic primary operable invasive breast cancers from patients enrolled into the Nottingham Tenovus Primary Breast Carcinoma Series that presented at Nottingham City Hospital between 1989 and 1998. The study was approved by the Nottingham Research Ethics Committee 2 under the title “Development of a molecular genetic classification of breast cancer.”

All patients had early operable primary breast cancer undergoing surgery as their primary treatment. TMA sections were obtained from the surgical resection samples. Histological sections of all cases were reviewed by the pathologist, the representative paraffin tumor blocks were chosen as donor block for each case, and the selected areas were marked for construction of TMA blocks. For the Hong Kong cohort, each case was constructed as duplicate in the TMA, and the average score of the duplicate was taken as the score.

There were from both cohorts a total of 1,271 cases that could be assessed and scored for BQ323636.1 staining. Of these 1,271 primary breast cancer samples, 358 cases were ER⁺ and had been given adjuvant tamoxifen treatment with available follow-up clinical data. Tamoxifen resistance is defined as patients who had been treated with tamoxifen in the adjuvant setting but subsequently developed disease relapse or distant metastases. Only cases with a clear history of tamoxifen response were used for analysis.

The IHC was performed as previously described (17). BQ323636.1 antibody (clone D-12) was at diluted 1:50. An Aperio ScanScope system (Aperio Technology) was used to assess BQ323636.1 expression. The intensities and percentages of the nuclear staining were scored by two independent individuals using H-scoring system (18). H-score = (1 × % of cells stained at intensity category 1) + (2 × % of cells stained at intensity category 2) + (3 × % of cells stained at intensity category 3). The cutoff was set as the median of H-score, which was 130.

Results were compared by two-tailed Student t test in Excel unless otherwise stated. Data from TMA were analyzed in SPSS (IBM, version 20). Where appropriate, the data were dichotomized into two groups including high or low expression using median expression level as cutoff. The correlations were analyzed by χ² tests. The expression levels of BQ were compared between different groups using Mann–Whitney U rank test. Survival analyses were done by Kaplan–Meier estimates followed by log-rank test, and Cox regression model. P values of less than 0.05 were considered statistically significant. In the figures, data were expressed as mean ± SD; *, P < 0.05; **, P < 0.01; ***, indicates P < 0.001.

Methods for immunofluorescent staining, transient transfection, luciferase reporter assay, lentiviral transfection for generation of stable cell line, and subcellular fractionation are described in Supplementary Materials and Methods.

The monoclonal anti-BQ323636.1 antibody was generated by InVivo Biotech, and the protocol for antibody generation and quality control is described in Supplementary Materials and Methods. The epitope of this antibody maps to the last 11 amino acid residues on BQ323636.1 protein that distinguish the variant from its wild-type.

To confirm the correlation between BQ and tamoxifen resistance, we examined the expression level of BQ in two pairs of cell lines: MCF-7 (ER⁺, tamoxifen sensitive) versus LCC2 (ER⁺, tamoxifen resistant) and ZR75-1 (ER⁺, tamoxifen sensitive) versus AK47 (ER⁻, tamoxifen resistant). The expression level of BQ323636.1 was examined by Western blot using our generated anti-BQ323636.1 specific antibody. Results showed that both mRNA and protein expression of BQ was higher in tamoxifen-resistant cells (Fig. 1A and B) compared with tamoxifen-sensitive cells.

To confirm the effect of BQ on tamoxifen resistance, BQ was stably overexpressed using a lentiviral transduction system in tamoxifen-sensitive cell lines MCF-7 and ZR-75-1, which express low levels of BQ323636.1. Overexpression of BQ323636.1 was confirmed by RT-qPCR and Western blot (Supplementary Fig. S1). The responses of ZR75-1-BQ/ZR75-1-vec and MCF-7-BQ/MCF-7-vec to tamoxifen treatment were studied by MTT assay. Overexpression of BQ could significantly induce resistance to tamoxifen in both ZR-75-1 and MCF-7 cell lines (Fig. 1C and D). This result was consistent with our previous finding using transient transfection (15) and strengthens the evidence supporting a role of BQ323636.1 in tamoxifen resistance.

The involvement of BQ323636.1 in tamoxifen resistance was further confirmed using an in vivo mouse model. As shown in Fig. 1E, ZR75-1-vec tumors were sensitive to tamoxifen treatment as marked by significant reduction of tumor growth when compared with control. In comparison, BQ323636.1-overexpressed tumors did not respond to tamoxifen treatment (Fig. 1F). Similar results were also observed using MCF-7 (Supplementary Fig. S2), showing that the effects were not cell line specific. Taken together, these results provide solid proof to support the claim that BQ323636.1 overexpression was one of the factors contributing to tamoxifen resistance.

Having established that BQ323636.1 overexpression could confer tamoxifen resistance, the next step was to understand the mechanism. It has been observed that this splice variant could modulate transcription of ERE-containing luciferase reporter in a way opposite to its wild-type (15). Because NCOR2 in response to tamoxifen functions as a corepressor to suppress ER-target genes; therefore, it was hypothesized that BQ323636.1 could compete with NCOR2 to restrict the suppressive functions of NCOR2, which in turn relieved the transcriptional suppression imposed by tamoxifen.

To test this, the effects of BQ323636.1 and/or NCOR2 overexpression on transcription of ERE-containing luciferase reporter (ERE-E1b-luc) were studied, and the results were expressed as relative luciferase activity normalized against the vector control. Consistent to its corepressor role in breast cancer, ectopic overexpression of NCOR2 was found to suppress luciferase activities in both ZR-75-1 and MCF-7 in all three conditions including nontreated (EtOH), treated with E2, or treated with tamoxifen (Fig. 2A and B). However, when BQ323636.1 was overexpressed together with NCOR2, the luciferase activities were significantly higher when compared with NCOR2 overexpression alone (Fig. 2A and B), indicating that the suppression on luciferase activity imposed by NCOR2 was, at least partially, rescued.

It was of great importance to understand how BQ could interfere with the functions of NCOR2. From these and previous results (15), BQ did not appear to exert a dominant-negative effect on NCOR2, but instead, competed with NCOR2. It has been reported that NCOR2 functions as a dimer in the corepressor complex by binding to itself in an antiparallel fashion (aa290–427 binds to aa1788–1903; ref. 9), serving as a central dock for the further recruitment of other corepressor proteins. As BQ protein retains only the RD1 portion of NCOR2, theoretically, it would have lost the ability to interact with ER and with other corepressor proteins such as HDACs. Therefore, we proposed that BQ323636.1 could bind to NCOR2, forming a defective platform to which other corepressor proteins cannot be fully recruited.

To test this conjecture, coimmunoprecipitation was first performed to study the interactions among ER, NCOR2, and BQ. As shown in Fig. 2C, the interaction between NCOR2 and ER was confirmed in ZR-75-1 and MCF-7. More importantly, it was found that BQ323636.1 interacted with NCOR2 but not with ER, consistent with the prediction and supporting the notion that BQ323636.1 bound to NCOR2. Furthermore, tamoxifen-resistant cells express more BQ, which compete with NCOR2; hence, it is expected that less NCOR2 be available for interaction with ER in LCC2 compared with tamoxifen-sensitive cell MCF-7 as shown in Fig. 2C(iii).

Next, the effects of BQ323636.1 overexpression on the interactions between NCOR2 and the corepressor proteins HDAC3, TBLR1, and GPS2 were studied by co-IP. To this end, the effects of BQ323636.1 on the expressions of NCOR2, HDAC3, TBLR1, and GPS2 were first examined by Western blot. It was observed that BQ323636.1 overexpression did not alter the expression levels of these proteins (Supplementary Fig. S3). Co-IP further revealed that BQ323636.1 overexpression suppressed the interaction between HDAC3 and NCOR2, GPS2 and TLBR1, respectively, for the tamoxifen-sensitive cell lines MCF-7 and ZR-75 (Fig. 2Di and ii). Likewise, tamoxifen-resistant cells LCC2, which express more BQ that competes with NCOR2, suppress the interaction between HDAC3 and NCOR2, GPS2 and TLBR1, respectively, when compared with MCF-7 (Fig. 2Diii).

Under normal conditions, ERα binds to estrogen response element (ERE) of different genes such as GREB, LY6E, NRIP and pS2 to initiate the transcription (19). NCOR2 can bind to ERα on ERE to form a gene repressor complex which can thus repress the effect on ERα on transcription (20). ChIP assays performed showed that overexpression of BQ could reduce the amount of NCOR2 interacting with the ERE genes pS2, NRIP, LY6E and GREB (Supplementary Fig. S4). Collectively, these data suggest that BQ323636.1 binds to NCOR2, to functionally interfere with the formation of NCOR2 gene co-repressor complex of ERα, thus confirming that the recruitment of the co-repressor complex to the endogenous gene promoters is indeed affected by BQ overexpression.

It had been previously observed that whilst protein levels of BQ323636.1 were high, there were very low expression levels of BQ323636.1 mRNA compared with the NCOR2 wild-type, suggesting there may be some posttranscriptional mechanism regulating BQ323636.1 expression (15). The phosphorylation site Ser2410 on NCOR2 reported to be responsible for proteasome-mediated degradation is deleted in BQ323636.1. Thus, compared with NCOR2, BQ323636.1 may be less susceptible to degradation. To study the proteasome-mediated degradation of BQ323636.1, MCF-7 and ZR-75-1 were treated with MG132, a potent proteasome inhibitor, for a time course as indicated in Fig. 3A and B. It was found that MG132 treatment could cause the accumulation of NCOR2 but not BQ323636.1, suggesting that NCOR2 protein was actively degraded through the proteasome-mediated degradation pathway whilst BQ323636.1 was not. To gain definite knowledge of the turnover of these two proteins in the cells, pulse-chase analyses were performed to determine the half-life. It was found that NCOR2 degraded continuously along with time, while BQ323636.1 protein stayed relatively constant after an initial decrease from 0 to 2 hours (Fig. 3C). The half-life of NCOR2 was around 4 hours while that of BQ323636.1 was undetermined using the same time course (Fig. 3C). Overall, our finding indicated that BQ323636.1 protein would have a higher stability than NCOR2 protein, probably because of the impaired proteasome-degradation pathway.

Previously, in a small cohort of 77 patients with breast cancer, we have shown that BQ323636.1 overexpression at the mRNA level is significantly associated with tamoxifen resistance (15). As we have demonstrated, the BQ323636.1 protein binds to NCOR2 to suppress the formation of corepressor complexes. Hence, more informative results can be obtained by assessing BQ323636.1 protein expression in relation to tamoxifen resistance. With the use of the specific monoclonal anti-BQ323636.1 antibody, protein expression of BQ323636.1 was directly examined in breast cancer tumor tissue by immunohistochemistry as in vivo confirmation of the usefulness of BQ323636.1 as a predictive marker for tamoxifen resistance.

Representative images of IHC staining are shown in Fig. 4A. It was observed that, whilst BQ323636.1 was expressed in the cytoplasm at a fairly similar level for all patients, the intensity and distribution of nuclear BQ323636.1 staining was heterogeneous and varied from patient to patient. Hence, semiquantitative assessment of nuclear BQ323636.1 expression levels was performed using the H-score system, which takes into account both the intensity and percentage of cells stained at the intensity category, which was named BQ nuclear score.

Three hundred fifty-eight ER⁺ primary breast cancer samples recruited from two independent Hong Kong and United Kingdom cohorts were used for analysis. All patients had received adjuvant tamoxifen treatment with available follow-up clinical data and clear history of tamoxifen response. Tamoxifen resistance is defined as tamoxifen-treated patients who subsequently developed disease relapse or distant metastases. The clinicopathologic data of these 358 cases are shown in Supplementary Table S1, for which 78 cases were from Hong Kong and 280 cases from the United Kingdom. Both cohorts of patients when analyzed separately, each gave statistically significant results (Supplementary Figs. S5 and S6), and the pooled results are presented.

Statistical analysis revealed the BQ nuclear score was significantly associated with tamoxifen resistance by χ² test (P = 1.79 × 10⁻⁶), and significantly higher in patients found to be tamoxifen-resistant (Mann–Whitney U rank test, P = 0.001, Fig. 4B). In tamoxifen-treated patients, BQ nuclear score was significantly correlated with cancer metastasis (χ² test, P = 4.27 × 10⁻⁷) and significantly higher in patients who developed metastasis (Mann–Whitney U rank test, P = 0.00009, Fig. 4C). BQ nuclear score was also significantly associated with disease relapse (χ² test, P = 1.87 × 10⁻⁴) and significantly higher in patients who developed disease relapse (Mann–Whitney U rank test, P = 0.018, Fig. 4D).

Consistent with its role in predicting tamoxifen resistance, BQ nuclear score was significantly associated with poorer survival by Kaplan–Meier estimate (log-rank test, P = 1.13 × 10⁻⁴ and P = 4.02 × 10⁻⁵ for overall survival and disease-specific survival, respectively, Fig. 4E and F). By Cox-regression analysis (Table 1), it was also observed that BQ nuclear score was significantly associated with poorer overall survival (RR = 1.765, 95% CI, 1.317–2.365; P = 1.42 × 10⁻⁴) as well as with poorer disease-specific survival (RR = 2.093, 95% CI, 1.459–3.002; P = 6.03 × 10⁻⁵). Multivariate cox-regression analyses showed that after adjustment for age, tumor stage, tumor grade, and progesterone receptor (PR) status, BQ nuclear score remained to be significantly associated with higher risk of death; overall survival (RR = 2.117, 95% CI, 1.520–2.948; P = 9.18 × 10⁻⁶) and for disease-specific survival (RR = 2.604, 95% CI, 1.746–3.885; P = 2.74 × 10⁻⁶).

With the use of nuclear BQ score, this scoring method could predict tamoxifen resistance of the patients with breast cancer with a sensitivity of 52.9% and specificity of 72.0%. Positive predictive value is 59.0% while the negative predictive value is 66.7%. The results indicate that BQ323636.1 overexpression is an independent prognostic marker for patients who have received tamoxifen treatment, consistent with its role in predicting tamoxifen resistance.

The estrogen (ER) signaling pathway is an important proproliferative pathway in breast cancer, which upon binding with estrogen, activates target gene transcription and cell growth. More than two-thirds of patients with breast cancer are ER⁺ and can be treated with endocrine therapy. Tamoxifen is most commonly prescribed to patients who are ER⁺ for the prevention of breast cancer relapse or metastasis, but despite significant successes in improving patients' survival, tamoxifen resistance remains an outstanding issue (14). The binding of tamoxifen to ER triggers the recruitment of NCOR2 as well as other corepressors, such as GPS2, TBLR1, HDAC3, to suppress the ER-mediated cell proliferative signaling pathway (6–8). The mechanisms underlying tamoxifen resistance have been extensively studying but not yet well understood.

The pharmacologic activity of tamoxifen is dependent upon its conversion by the metabolic enzyme CYP2D6 to its active metabolite, endoxifen. Patients with reduced CYP2D6 activity were considered to benefit less from tamoxifen treatment and CYP2D6 became a patented marker for prediction of tamoxifen response (21, 22). Subsequent clinical studies however have reported conflicting results, some supportive that patients with reduced CYP2D6 activity derive inferior therapeutic benefit from tamoxifen (23), whilst others finding the contrary (24). Therefore, it remains questionable whether CYP2D6 can serve as a robust biomarker for predicting tamoxifen resistance.

Therefore, there is as yet no available robust biomarker to discriminate patients who are ER⁺ who will be sensitive to tamoxifen from those who will be resistant. Patients with ER⁺ disease will thus be prescribed with tamoxifen but clinicians can only document drug failure when cancer reoccurs and/or metastasizes.

BQ323636.1 is a splice variant of NCOR2 identified by our group with exon 11 skipping during mRNA splicing, resulting in an early translation stop codon and a truncated protein product retaining only the N-terminal fragment compared with its wild-type (15). Although the BQ mRNA level is 2% to 5% of NCOR2, BQ323636.1 protein, on the other hand, was found to be expressed at similar levels to NCOR2 (15). This could be due in part to BQ323636.1 being not actively degraded through the proteasome-mediated pathway with a longer half-life in the cell compared with that of NCOR2, or that the translation efficiency of BQ323636.1 was higher than NCOR2. These mechanisms could help maintain protein expression of BQ323636.1 at a functionally significant level in the cell. The fact that NCOR2 functions as a dimer and that BQ323636.1 binds to NCOR2 could amplify the functional importance of the splice variant in the cells.

In this study, in vitro and in vivo evidence was presented to support that BQ323636.1 overexpression could confer tamoxifen resistance. Moreover, the functional roles of BQ323636.1 were studied. BQ323636.1 overexpression rescued the suppressive effect of NCOR2 overexpression exerted on the transcription of an ERE-containing luciferase reporter and compromised the inhibitory effects of tamoxifen on ER-target gene expression. Such effects can be explained by the fact that BQ323636.1 binds to NCOR2 but not to ER in the cells and that BQ323636.1 overexpression inhibits the interaction between HDAC3 and NCOR2, GPS2 and TBLR1, respectively, indicating the suppressed formation of corepressor complex. Based on our findings, the mechanism by which BQ323636.1 overexpression contributes to tamoxifen resistance is proposed as follows (Fig. 5). In the absence of BQ323636.1, in response to tamoxifen treatment, NCOR2 homodimerizes in an antiparallel fashion, serving as a dock for further recruitment of other corepressor proteins, such as TBLR1, HDACs, and GPS2, to form a complete and functional corepressor complex that will bind to ER to suppress gene transcription and halt cell growth. In contrast, when BQ323636.1 is present, it binds to NCOR2, forming a faulty dock to which other corepressor proteins cannot be fully recruited, resulting in a defective corepressor complex which cannot bind to ER or is unable to fully suppress gene transcription (Fig. 5). Hence, the cells become inert to tamoxifen treatment.

Statistical analyses using TMA constructed from 358 primary breast cancer samples obtained from two independent cohorts of patients, recruited from Hong Kong and from the United Kingdom showed that, for patients who have been treated with tamoxifen, BQ323636.1 nuclear overexpression is significantly associated with tamoxifen resistance (χ² test, P < 0.001), which is defined as disease relapse or metastasis after being treated with tamoxifen. The correlation between nuclear BQ323636.1 enrichment and tamoxifen resistance is also observed in cell lines. It is also noted that, not only BQ323636.1 is expressed at higher levels in tamoxifen-resistant cell lines (LCC2 and AK47) when compared with their parental tamoxifen-sensitive cell lines (MCF-7 and ZR-75), but nuclear localization is observed more distinctly in tamoxifen-resistant cell lines LCC2 and AK-47. Based on the existing knowledge, when present in the nucleus, BQ323636.1 is likely to be functionally active to mediate tamoxifen resistance. It is not yet clear why BQ323636.1 is expressed in both the cytoplasm and nucleus. Moreover, it would be very interesting to further study what regulates the shuttling of BQ323636.1 between nucleus and cytoplasm.

Nuclear BQ323636.1 overexpression was significantly associated with tamoxifen resistance by χ² test, P = 1.79 × 10⁻⁶ with a sensitivity of 52.9% and specificity of 72.0%. The performance of BQ as a biomarker of tamoxifen response can be further improved using an expanded cohort. On the other hand, the cytoplasmic staining of BQ in IHC may have interfered with scorer's interpretation of nuclear BQ expression. Unlike NCOR2, which is localized only in the nucleus, BQ was found to be expressed in both cytoplasm and nucleus. This raises a very interesting question as to what are the differences between cytoplasmic and nuclear BQ and what mechanisms, such as posttranslational modifications, controls the shuttling of this protein between the cytoplasm and the nucleus. By investigating such questions, it may be possible to generate an antibody that specifically recognizes nuclear BQ, thus improving the accuracy of tamoxifen response prediction. Phosphorylation is a key regulatory step redirecting cellular localization between the cytoplasm and nucleus of many proteins. For example, nuclear iASSP, which is enriched in melanoma metastasis and associated with poor patient survival, is the phosphorylated form of iASSP and phosphorylation by cyclin B1/CDK1 on Ser-84 and Ser-113 triggers iASSP to enter the nucleus (25). Phosphorylation of FOXO3 on Ser-7 by p38 MAPK triggers the transition of FOXO3 from cytoplasm to nucleus upon doxorubicin treatment (26). Therfore, it is postulated that phosphorylation profile may distinguish nuclear from cytoplasmic BQ, and phosphorylation could signal BQ to enter the nucleus from the cytoplasm.

Aromatase inhibitors (AI) act by inhibiting the synthesis of estrogen from androgen by blocking the activity of the aromatase CYP19A1, resulting in reduced estrogen for induction of the ER-mediated signaling pathway (27). Our studies have confirmed that BQ activates ER-mediated signaling pathway in a ligand independent manner. Overexpression of BQ significantly enhanced the activity of ERE even in the absence of estrogen or tamoxifen (Fig. 2A and B); therefore, it is likely that BQ may also be a predictive marker for AI resistance. This needs confirmation with a cohort of patients with available information of AI treatment and response.

To conclude, the current study has provided in vitro and in vivo evidence to support the notion that BQ overexpression could confer tamoxifen resistance and that BQ could be used as a predictive marker for patients' response to tamoxifen treatment. The molecular mechanism through which BQ could induce tamoxifen resistance has also been studied and proposed. More studies are required using patient samples from different ethnic backgrounds to further verify the reliability of BQ in predicting tamoxifen resistance. Furthermore, it would be interesting to investigate the significance of BQ overexpression in relation to treatment with other novel antiestrogens.

K. Cheung is a consultant/advisory board member for Genomic Health and reports receiving commercial research grants from AstraZeneca. No potential conflicts of interest were disclosed by the other authors.

Conception and design: E.A. Rakha, U.-S. Khoo

Development of methodology: C. Gong, E.P.S. Man, H. Tsoi, T.K.W. Lee, S.-T. Ma, I.O. Ellis, U.-S. Khoo

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Gong, E.P.S. Man, H. Tsoi, P. Lee, S.-T. Ma, L.-S. Wong, M.-Y. Luk, E.A. Rakha, A.R. Green, I.O. Ellis, E.W.-F. Lam, K.-L. Cheung, U.-S. Khoo

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Gong, E.P.S. Man, H. Tsoi, P. Lee, A.R. Green, K.-L. Cheung, U.-S. Khoo

Writing, review, and/or revision of the manuscript: C. Gong, H. Tsoi, E.A. Rakha, A.R. Green, I.O. Ellis, E.W.-F. Lam, K.-L. Cheung, U.-S. Khoo

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Tsoi, E.A. Rakha, U.-S. Khoo

Study supervision: E.A. Rakha, U.-S. Khoo

This project was kindly supported by the Innovative Technology Fund, HKSAR (ITF Ref. ITS/015/13), the Committee on Research and Conference Grants from the University of Hong Kong Project number 201309176148, the Medical Research Council, UK (MR/N012097/1), and Cancer Research UK (A12011). We would like to thank Dr. Robert Clarke (Georgetown University Medical School, Washington, DC) for kindly providing us LCC2 and AK-47 cell lines and Dr. Carolyn Smith (Baylor College of Medicine, Houston, TX) for providing the luciferase reporter ERE-TK-Luc. We also thank Faculty Core Facility of LKS Faculty of Medicine, HKU, for providing technical support in confocal microscopy. We thank the Nottingham Health Science Biobank and Breast Cancer Now Tissue Bank for the provision of tissue samples. U.-S. Khoo is Ada Chan Professor in Oncological Pathology.

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