SpliceArray Profiling of Breast Cancer Reveals a Novel Variant of NCOR2/SMRT That Is Associated with Tamoxifen Resistance and Control of ERα Transcriptional Activity

Breast cancer is the most prevalent female cancer worldwide. About 70% of breast cancers express estrogen receptor (ER), a member of the nuclear hormone family of intracellular receptors. Estrogen binds to the ligand-binding domain of the ER and induces conformational change in protein structure for dimerization that is favorable for subsequent interaction with coactivator molecules (1). The activated ER then transactivates estrogen-responsive genes and induces subsequent cell response.

Tamoxifen, a selective ER modulator, is commonly used as first-line adjuvant treatment to prevent cancer recurrence for ER-positive (ER+) breast cancers. It reduces the annual breast cancer mortality by one-third and disease recurrence by almost half (2). In breast tissues, tamoxifen acts as an antagonist of ER, competes with estrogen for binding to ER, blocks the transcription activation activity of ER, and arrests cancer cells in the cell cycle. However, almost half of the patients with prolonged tamoxifen treatment will fail to respond to tamoxifen over time and develop disease recurrence eventually, that is, acquired tamoxifen resistance (3) for reasons as yet unclear. Postulated mechanisms are loss of ER expression, aberrant changes in coregulators, dysregulation of receptor tyrosine kinase signaling or cell survival signaling, and apoptosis (4).

Numerous gene expression profiling studies (5–10) have been done by different research groups with the aim of improving prognostication and classification of breast cancers, including tamoxifen responsiveness, to assist the selection of patients for the most appropriate adjuvant systemic therapies. However, each of the studies provided gene lists showing very little concordance in gene identity.

Alternative splicing is a key posttranscriptional mechanism for generating multiple protein products from a single gene, occurring in perhaps 40% to 60% of human genes. Alternatively spliced isoforms of a given protein can display different and even antagonistic biologic functions. Microarray chips detect a small fragment of each gene transcript and unless specifically designed, are unlikely to discriminate the expression of all possible isoforms for most of the genes studied. Over or reduced expression of some gene isoforms that have different functions from their wild-type might account for differences in gene identity between the different gene expression profile models. In recent years, various studies have revealed a number of alternatively spliced variants that are associated with cancers (11, 12). Splicing microarray profiling in prostate cancer found that 30% splice variants of 222 prostate cancer–related genes had significant changes in expression between tumors and normal samples (13). A number of alternative spliced variants that distinguished tumors from normal tissues have also been identified in ovarian and breast cancers (14, 15).

We postulated that alternatively spliced variants may play a role in tamoxifen resistance in breast cancer and conducted custom designed splice variant profiling on a panel of breast cancer cell lines with differential response to tamoxifen treatment. A novel splice variant of the gene NCOR2 was identified that was differentially expressed between parental tamoxifen-sensitive and -derived resistant breast cancer cell lines. The variant was further validated to significantly correlate with tamoxifen resistance and poor survival in breast cancers. Subsequent functional studies were carried out to elucidate the underlying role of this splice variant in tamoxifen resistance.

Human breast cancer cell lines MCF7 and ZR-75-1 were purchased from American Type Culture Collection and were authenticated by short tandem repeat profiling. MCF7 was cultured in Dulbecco's Modified Eagle's Medium supplemented with 10% FBS and 10 ng/mL human recombinant insulin and ZR-75-1 was cultured in Improved Minimum Essential Medium (IMEM) supplemented with 5% FBS. LCC2 and AK-47 were kindly provided by Dr. Robert Clarke (Georgetown University Medical School, Washington, D.C.; ref. 16). LCC2, a tamoxifen-resistant cell line derived from MCF7/LCC1 by stepwise in vitro selection of prolonged tamoxifen treatment was cultured in IMEM supplemented with 5% charcoal-stripped FBS and AK47, derived from ZR75-1, was cultured in IMEM supplemented with 5% charcoal-stripped FBS.

Primary tumors were collected from 77 patients diagnosed with breast cancer between the years 1993 and 2001 at Queen Mary Hospital (Hong Kong, China). Five- to 10-year clinical follow-up data and clinicopathologic information were available. The protocol was approved by the Institutional Review Board (UW 06-379 T/1404). Histologic sections of all cases were reviewed by a pathologist. Sections containing at least 70% of invasive carcinoma were used for total RNA extraction.

TRIzol reagent (Invitrogen) was used for total RNA extraction. Breast cancer frozen tissues were cut in 20-μm thick cryostat sections for total RNA extraction.

SpliceArray profiling was conducted by ExonHit Therapeutics, Inc. Custom designed 4 × 44K microarray slides from Agilent Technologies were used. The probe sets were designed to detect the wild-type forms as recommended in National Center for Biotechnology Information RefSeq and all possible alternatively spliced forms. Each alternative splice event can be modeled with a short and a long form. The reference sequence of this gene can be either the short form or long form, depending on the alternative splice event type, and the alternatively spliced variant will be the other form.

A panel of breast cancer cell lines with differential response to tamoxifen was used in the SpliceArray profiling, as shown in Supplementary Table S1A. The cell lines were treated at tamoxifen concentrations 0 μmol/L, 0.1 μmol/L, and 1 μmol/L for 5 days before RNA extraction. Every sample was cultured and sampled in triplicate. All RNA samples went through quality control before cDNA synthesis and Cy3 dye labeling.

SpliceArray profiling analysis was conducted using Partek Genomics Suite. Analysis was conducted on each cell line cDNA sample individually. The probe intensity data were Log₂ transformed then quartile normalized to make the distribution of probe set intensities for each array similar within the set of arrays. Low or nonexpressed probe sets were filtered out if all samples' intensity values fell below the predetermined Log₂-based value.

For data obtained from the array profiling, a two-way ANOVA model was used to conduct statistical tests on the filtered probe set level intensities between each cell line sample group to generate P values and fold change values. “Top hit lists” were generated with P ≤ 0.05 and fold change ≥ 1.8 as the cutoff, representing significant differences between the samples being compared. Clustering of the profiling data was conducted using Cluster 3.0 (17) and visualized using Java TreeView (18).

In vivo correlation analysis was based on the clinicopathologic characteristics of the 77 Chinese breast cancer samples collected from Queen Mary Hospital as described above. The expression levels of splice variant BQ323636.1 were analyzed against the tamoxifen responsiveness of the patients with breast cancer and other clinicopathologic features by independent sample t test. Survival analyses were done by Kaplan–Meier estimates and Cox regression model. P values of less than 0.05 were considered statistically significant.

Up to 5 μg of total RNAs were reverse transcribed into cDNA by SuperScript III reverse transcriptase (Invitrogen). Real-time PCR reaction was carried out with the ABI 7900HT Fast Real-time PCR system. For TaqMan real-time quantitative PCR (qPCR), a standard curve was set up for quantification. The reaction volume was 15 μL in 96-well plate and 10 μL in 384-well plate. The reaction consisted of 1× TaqMan Universal PCR Master Mix, No AmpErase UNG (Applied Biosystems), 0.1 μmol/L 6-carboxyfluorescein-labeled TaqMan MGB probe, 0.4 μmol/L forward and reverse primers, and 2 μL cDNA. The 2-step thermal cycling method consisted of 95°C for 10 minutes, followed by 40 cycles of 95°C for 15 seconds, and 60°C for 60 seconds. The primer and TaqMan probe sequences used were presented in Supplementary Table S2.

Cells were lysed and SDS-PAGE was conducted as previously described (19). Harvested cells were lysed in protein lysis buffer (Cell Signaling) containing 1 × Complete Protease Inhibitor Cocktail Tablet and 1 mmol/L phenylmethylsulfonylfluoride. Twenty micrograms of each protein sample were separated by SDS-PAGE using 8% gel. The proteins then were transferred to polyvinylidene fluoride membrane (Millipore) and probed with monoclonal antibody against NCOR2 (A01, Abnova), ERα (HC-20, Santa Cruz Biotech), 6xHis-tag (9F2, Wako), and β-actin (AC-74, Sigma). Secondary antibodies were: horseradish peroxidase–conjugated anti-rabbit antibody (P0448, Dako) and horseradish peroxidase-conjugated affiniPure anti-mouse antibody (Jackson ImmnunoResearch Laboratories).

Cells were seeded at a density of 1.5 × 10⁴ in 96-well plates. On the day of conducting MTT assay, cells were incubated with culture medium containing 10% MTT (USB, Affymetrix) stock solution (5 mg/mL) at 37°C for 4 hours. After incubation, the medium was discarded and replaced with 100 μL isopropanol with 4 mmol/L HCl and 0.1% NP-40 to dissolve the purple precipitates. Extinction of the samples was measured by the Infinite 200 microtiter plate reader (Tecan) at 570 nm with a reference wavelength at 750 nm. Background MTT readings were subtracted from the samples.

pcDNA3.1-BQ323636.1 was constructed by inserting the human variant BQ323636.1 cDNA into pcDNA3.1(+) (Invitrogen). pcDNA3.1-6xHis-BQ323636.1 was constructed by Dr. Tori Chan, The University of Hong Kong, Hong Kong. pcDNA3.1-ERα was constructed by inserting full-length human ERα cDNA into pcDNA3.1(+). The luciferase reporter ERE-E1b-Luc was a gift from Dr. Carolyn Smith (Baylor College of Medicine, Houston, TX; ref. 20) and ERE-TK-Luc was a gift from Dr. Hung-Ying Kao (Case Western Reserve University, Cleveland, OH; ref. 21).

The ERE firefly reporter and pRL-SV40 were cotransfected with pcDNA3.1-ERα and pcDNA3.1-BQ323636.1 or its empty vector into HeLa cells with Lipofectamine 2000 transfection reagent (Invitrogen). Cells grown to 90% confluence were grown in Opti-MEM I-reduced serum medium without phenol red before transfection. The cells were transfected with desired amount of DNA or Lipofectamine 2000 reagent following the manufacturer's instruction. Six hours after transfection, the medium was changed to complete medium and incubated at 37°C with 5% CO₂.

After ligand treatment for 24 hours, the activities of firefly and renilla luciferases were measured sequentially using dual-luciferase reporter assay system (Promega) according to manufacturer's protocol. The intensities of firefly and renilla luminescence were recorded by Infinite F200 (Tecan). Background readings were subtracted from the samples.

To identify the alternatively spliced variants that correlate with tamoxifen resistance status in breast cancer, custom designed SpliceArray profiling was conducted. Breast cancer–related genes involved in cell signaling, cell proliferation, cell cycle, apoptosis and invasion, 417 in total, were chosen for profiling study. These 417 genes had a total of 4,088 evidence events investigated that were represented by 21,249 evidence probes. The evidence event refers to all the known and predicted alternative splice events identified in publicly available sequence databases of the gene.

A panel of breast cancer cell lines with differential response to tamoxifen was used in the SpliceArray profiling (Supplementary Table S1A). AK 47 (ER-negative) can still respond to tamoxifen treatment; however, it is more resistant to tamoxifen compared with its parental cell line ZR-75-1 (ER+) as its IC₅₀ was much higher (Supplementary Fig. S1). Cell line RNA samples were prepared in independent triplicates. The expression level of each alternatively spliced variant is expressed as the probe intensities of long form versus short form ratio. A large number of splice variants that had expression levels correlating with ER status or tamoxifen responsiveness status in the panel of breast cancer cell lines screened were observed (data not shown). To identify the significant and meaningful splice variants that are truly involved in tamoxifen resistance, splice variants that fulfilled at least one of the criteria below were selected:

• At basal level, they were differentially expressed (fold change >1.8, P < 0.05) between tamoxifen-sensitive (TamS) and tamoxifen-resistant (TamR) cell lines in both 2 pairs, that is, MCF7 versus LCC2; ZR75-1 versus AK47;

• Under tamoxifen treatment, they had opposite trend of change (fold change >1.8, P < 0.05) between TamS and TamR cell lines in both pairs;

• At basal level, they were differentially expressed (fold change >1.8, P < 0.05) between ER+ and ER− cell lines.

The significant fold change cutoff was set at 1.8, as this is the minimum difference that may be detected by TaqMan real-time quantitative PCR. Using these criteria, 11 splice variants of 5 genes were selected (Supplementary Fig. S2) and subjected to validation by qPCR. The validation results were summarized in Supplementary Table S1B and the fold changes of their long versus short form ratio within the cell line pairs are illustrated in Fig. 1. At basal level (without tamoxifen treatment), 4 splice variants were validated in MCF7/LCC2 cell line pair and 5 splice variants were validated in ZR75-1/AK47 cell line pair (Fig. 1).

One of the splice variants of TGF-β1, BF756452.1, with a novel alternative splice acceptor site of 41 bps at the 5′ end upstream of exon 10 and a novel stop codon that leads to early termination of the protein before exon 10 was downregulated the most in both cell line pairs. The other splice variant BU149200.1, with a novel intron of 139 bps within the last exon, was the most upregulated variant in both cell line pairs. The splice variants of TGF-β1 validated from our study are being further investigated, but are outside the context of this article.

Splice variant of ERα, AX066402.1, with a deletion of the first 151 amino acids at the N-terminal of the protein due to exon 1 skipping had the relative large fold change in AK47/ZR75-1 cell line pair. However, because both the variant and the ERα wild-type had very low or no expression in ER− cell lines, its functional significance in ER− cells was limited, thus it was not chosen for functional studies.

Besides the 3 variants introduced above, fold change of BQ323636.1, a novel variant of nuclear receptor corepressor 2 (NCOR2), was significantly upregulated in AK47 compared with ZR75-1 (2.5-fold in log₂ scale). Exon 11 skipping in this splice variant leads to early termination of the protein product. Considering the known biologic functions of the genes validated in this study, NCOR2 was the most likely to bear functional relationship with tamoxifen resistance. It interacts with ERα when ERα is complexed with tamoxifen and represses the transcriptional activity of ERα. It has been shown to be directly involved in the antagonist activity of tamoxifen on ERα. Therefore, its splice variant, BQ323636.1, may play an important role in tamoxifen resistance and was selected for further analysis.

By SpliceArray profiling, we had identified a novel alternatively spliced variant BQ323636.1 that was significantly differentially expressed in both MCF7/LCC2 and ZR75-1/AK47 cell line pairs both at basal level and under tamoxifen treatment (Supplementary Table S1B). BQ323636.1 has exon 11 skipped in mature mRNA, which results in early termination of its protein product, retaining only the N-terminal of the NCOR2 wild-type protein (Fig. 2A). To validate the differential expression of splice variant BQ323636.1 between TamS and TamR cell lines at the mRNA level, real-time TaqMan qPCR was conducted on a new batch of cell line mRNA samples in triplicates. Primers were designed to specifically target either this splice variant or the NCOR2 wild-type form (NM_006312; Fig. 2B).

In the ZR75-1/AK47 cell line pair, qPCR results expressed as long versus short form (Fig. 2C, i), showed consistent results as in the SpliceArray profiling. To represent the expression level of BQ323636.1 in relation to that of its wild-type BQ323636.1 versus wild-type ratio (BQ/WT ratio), that is, the short versus long form, is shown in Fig. 2C, ii. Consistent with SpliceArray profiling data, the BQ/WT ratio was significantly higher in AK47 cell line (derived TamR) compared with its ZR75-1 cell line (parental TamS) by about 2-fold, both at basal level and under tamoxifen treatment. However, qPCR validation was unable to show significant differential expression of BQ323636.1 variant between the MCF7 and LCC2 cell line pair (data not shown). This may imply that the biologic role of BQ323636.1 in tamoxifen resistance involves loss of ER expression.

In addition, we also examined the expression level of the variant alone (Fig. 2C, iii) or wild-type alone (Fig. 2C, iv). Differential expression of BQ323636.1 was only observed under tamoxifen treatment, but not at basal level. This difference was however not observed for the NCOR2 wild-type. The differential expression of BQ323636.1 only under tamoxifen treatment for an ER− TamR cell line suggests its potential functional role of this variant in presence of tamoxifen.

The protein product of the variant BQ323636.1 was predicted to be a truncated protein product of 362 amino acids with molecular mass of 42.6 kDa. Its amino acid sequence corresponds to the N-terminus of NCOR2 wild-type protein, with only the last 10 amino acids before the stop codon translated from exon 12 out of frame (Fig. 2A).

If the differential expression of BQ323636.1 observed between ZR75-1 and AK47 has a functional role in biologic processes, such differential expression is also expected at the protein level. A NCOR2 polyclonal antibody (A01, Abnova) targets the first 1–111 amino acids of wild-type NCOR2 and was used to detect the protein product of BQ323636.1 in Western blot analysis. This antibody can detect the wild-type NCOR2 at size 250 kDa and the variant at size 42.6 kDa, respectively. Between protein size 40 and 50 kDa, a protein band was observed in both ZR75-1 and AK47, identified as the variant BQ323636.1 (Fig. 3A). Size of the variant was confirmed by overexpression of BQ323636.1 in both HeLa and 293T cell line with plasmid pcDNA3.1-His-BQ323636.1 (Supplementary Fig. S3). Consistent with the qPCR results, under tamoxifen treatment, the protein expression level of BQ323636.1 in AK47 was significantly higher than in ZR75-1 compared with that of wild-type. At basal level, such difference was also observed, but not as significant as under tamoxifen treatment.

To investigate whether the upregulation of BQ323636.1 was the cause of tamoxifen resistance, BQ323636.1 was transiently transfected into ER+, tamoxifen-sensitive cell line ZR-75-1, and MTT assays were conducted to assess the proliferation of cells in response to tamoxifen treatment. The decrease in cell proliferation observed under tamoxifen treatment was slower in BQ323636.1-overexpressing cells (Fig. 3B) than that of the vector control over a range of drug concentrations. This difference was statistically significant (P < 0.05) under 15 μmol/L tamoxifen treatment (Fig. 3C), indicating that the variant can mediate tamoxifen resistance in ER+ cells as well.

To reveal the clinical significance of this variant in vivo, we investigated for possible correlation of BQ323636.1 expression with clinicopathologic parameters in 77 breast cancer samples. BQ323636.1 mRNA expression was examined by TaqMan real-time qPCR and analyzed between clinical parameter groups using independent sample t test. Tamoxifen resistance was defined as those patients treated with tamoxifen only in the adjuvant setting who subsequently developed relapse or distant metastasis. Consistent with the observation in cell lines, the BQ323636.1 expression level was significantly higher in tamoxifen resistant (TamR) patients than those who were sensitive (TamS) to tamoxifen treatment (P = 0.006; Fig. 4A). The BQ/WT ratio showed similar trend, though not statistically significant (Supplementary Table S3).

Consistently, in 37 patients who were ER+ and had received tamoxifen treatment, higher BQ323636.1 expression level was significantly associated with development of distant metastasis (P = 0.048; Fig. 4B). However, BQ323636.1 expression level was not significantly correlated with chest wall/breast relapse (P = 0.453; Supplementary Fig. S4). This might be due to limited sample size, as among the 37 patients, there were only 27 patients whose relapse statuses were available.

Survival analysis was conducted to investigate the correlation between the variant's expression level and the clinical outcomes of the patients with breast cancer. Among the 77 patients, 37 patients, who had received tamoxifen treatment, were dichotomized by the BQ323636.1 expression level (cutoff value is the median). It was found that higher BQ323636.1 expression level was associated with both poorer overall survival (P = 0.041; Fig. 4C) and poorer disease-free survival (P = 0.008; Fig. 4D). This association remained significant after being adjusted by tumor size, tumor grade, HER2 and PR status of the patients (Table 1), and analyzed by multivariate Cox regression model (HR = 11.8989; P = 0.030 for overall survival; HR = 10.764; P = 0.038 for disease-free survival; Table 2). Though there were only 37 patients included, these significant findings would support the relevance of this variant in acquiring resistance among patients who have received tamoxifen treatment.

Moreover, it was found that both BQ323636.1 expression (P = 0.005) and the BQ/WT ratio (P = 0.008; Supplementary Fig. 5A and Supplementary Table S3) correlated to the de novo ER status of the patients at diagnosis, being significantly higher in ER− patients. Progesterone receptor (PR) is the downstream target gene of ERα. Following the pattern for ER status, the BQ323636.1 expression level was significantly higher in PR− patients (P = 0.035; Supplementary Fig. 5B). The BQ/WT ratio also showed a similar trend but not statistically significant. While there was no correlation observed for HER-2 status, the BQ/WT ratio significantly correlated with triple-negative (negative for ER, PR, and HER2 overexpression) cancers (P = 0.018; Supplementary Fig. 5C). There was no significant correlation between BQ323636.1 expression level and other clinicopathologic parameters including tumor size, tumor grade, and tumor type (Supplementary Table S3).

As tamoxifen directly targets ERα in breast cancer cells, we next investigated whether BQ323636.1 modulates ERα transcription activation activity when ERα is bound by estrogen (activated ERα) or tamoxifen (inactivated ERα).

Studies elucidating the role of NCOR2 on ERα transcription activity have been conflicting depending on the luciferase reporter used. Using the ERE-TK-Luc reporter, NCOR2 acted as a corepressor for ERα-mediated transcription (21, 22), whereas on the ERE-E1b-Luc reporter, it was shown that NCOR2 was required for full ERα transcriptional activity (20). We therefore used both luciferase reporters in turn to test the modulation of BQ323636.1 on ERα-mediated transcription activation (Fig. 5).

When ERα was unliganded (treated with vehicle EtOH), BQ323636.1 overexpression significantly suppressed the transcription on both ERE-E1b and ERE-TK promoters. On E2 treatment, BQ323636.1 overexpression exhibited opposing effects on luciferase activity between ERE-E1b and ERE-TK promoters. On ERE-E1b promoter, overexpression of BQ323636.1 induced significant decrease in transcription activity (Fig. 5A); whereas on ERE-TK promoter, significant increase in the transcription activity was seen (Fig. 5B). On the basis of previous reports of NCOR2 functions on ERE transcription, that is, as a coactivator on ERE-E1b promoter and a corepressor ERE-TK promoter, BQ323636.1 seems to exert the opposite effect in modulating ERα transcriptional activation activity compared with wild-type NCOR2.

In the presence of tamoxifen, regardless of whether E2 was present, BQ323636.1 consistently exhibited suppression effect on ERE transcription for both EREs. Overexpression of BQ323636.1 decreased luciferase activity by 70% for ERE-E1b promoter and by 50% for ERE-TK promoter. In conclusion, BQ323636.1 was shown to be able to modulate the transcriptional activation activity of ERα, indicating its functional importance in breast cancer.

Our SpliceArray profiling identified a novel splice variant BQ323636.1 of NCOR2 that had significantly higher expression in derived tamoxifen-resistant cell line AK47 than its parental tamoxifen-sensitive cell line ZR75-1, both at mRNA level and protein level. In vitro, overexpression of the variant in tamoxifen-sensitive cell line ZR-75-1 induced its resistance to tamoxifen. In vivo, it consistently correlated with tamoxifen resistance in breast cancer samples. Higher expression of BQ323636.1 was associated with tamoxifen resistance and, consequently, poorer overall and disease-free survival. Higher expression of BQ323636.1 was associated with negative ER and PR status and metastasis; higher BQ323636.1 to wild-type ratio was associated with de novo negative ER status, and triple-negative status — all features representing poor response to tamoxifen, poor prognosis, and more aggressive phenotypes in breast cancer.

NCOR2, also commonly known as SMRT (the silencing mediator of retinoic acid and thyroid hormone receptor), was initially discovered as the suppressor for retinoic acid and thyroid hormone receptors (23). As ERα is a critical ligand-activated transcription factor in breast cells controlling multiple important cell signaling pathways, its transcription activation on its target genes is strictly regulated by coactivators and corepressors. There are more than 23 corepressors of ERα identified that bind to different functional domains of ERα and suppress the activity of ERα through various mechanisms (24). NCOR2, together with its homologous gene NCOR, are the best characterized corepressors for ERα and interact with ERα in the presence of antagonist, that is, tamoxifen (25–27). When tamoxifen acts as an antagonist in the cell context, it binds to ERα and induces ERα to undergo conformational change that is more favorable for the binding of corepressors rather than coactivators (28, 29), so as to suppress the transcriptional activation activity of ERα.

We therefore investigated whether BQ323636.1 also regulated ERα transcription activation activity by luciferase assay. On E2 treatment, BQ323636.1 was observed to exhibit an opposite effect in modulating ERα transcriptional activation activity compared with wild-type NCOR2. For ERE-E1b promoter, on which NCOR2 was reported to act as a coactivator when E2 was present (25, 30), overexpression of BQ323636.1 significantly suppressed ERE-E1b transcription. For ERE-TK promoter, on which NCOR2 was reported to be a corepressor (22), overexpression of BQ323636.1 significantly promoted ERE-TK transcription. However, under tamoxifen treatment, regardless of whether E2 was present, BQ323636.1 consistently suppressed the transcription on both ERE-E1b and ERE-TK promoters. Therefore, when ERα is liganded with E2, the effect of NCOR2 and BQ323636.1 on its transcriptional activation activity is promoter-specific. Regardless of whether NCOR2 wild-type acts as a coactivator or corepressor on the ERE promoter, BQ323636.1 exerts an effect opposite to that of its wild-type. However, once tamoxifen is applied, BQ323636.1 uniformly exhibits suppressive effect on the transcription of both ERE promoters.

Splice variant BQ323636.1 skips exon 11 but retains the N-terminal 362 amino acids (42.6 kDa) of the NCOR2 wild-type protein, corresponding to the first repression domain RD-1 (Supplementary Fig. 6A). NCOR2 is a large protein (250 kDa) with its N-terminus interacting with other transcriptional corepressors to regulate transcription and its C-terminus binding to nuclear receptors, that is, ERα (Supplementary Fig. 6B). At the N-terminus of NCOR2, amino acids 1–316 interact with other transcription corepressors such as GPS2 (G protein pathway suppressor 2; ref. 21) and TBLR1 (transducin β-like 1 X-linked receptor 1; ref. 31; Supplementary Fig. 6B). The 2 SANT motifs, corresponding to amino acids 432–489 and amino acids 611–669, are critical components of a deacetylase activation domain (DAD) that bind and activate HDAC3. The repression domain RD-3 interacts with HDAC 4 and HDAC5, and the repression domain RD-4 interacts with HDAC3, GPS2, and TBLR1. As the SANT motifs important for NCOR2-HDAC3 interaction are missing in BQ323636.1 protein product, the modulation of BQ323636.1 on ERα transcriptional activity may not be through histone deacetylases (HDAC) but through the interaction with other corepressors, possibly GPS2 or TBLR1. The variant may compete with NCOR2 wild-type for binding to these protein partners and thus inhibit the function of its wild-type to facilitate the development of tamoxifen resistance (Supplementary Fig. S7). More comprehensive functional studies will be needed to elucidate the role of this variant in tamoxifen resistance.

If BQ323636.1 is functionally important, its corresponding amino acid sequence should be highly conserved among different species. Indeed, the protein product of BQ323636.1 (highlighted by the frame in Supplementary Fig. 6C), was the most conserved region in NCOR2 with above 90% similarity across all 7 species. This implies that BQ323636.1 contains the most important functional domain in NCOR2 protein.

Very low mRNA expression level of BQ323636.1 compared with the NCOR2 wild-type was observed in previous qPCR results; however, in Western blot analysis, the protein level of BQ323636.1 was rather high, similar, or even higher than its wild-type form. Therefore, there may be some posttranscriptional mechanisms promoting the translation of BQ323636.1 or protecting it from degradation. A phosphorylation site at Ser2410 on NCOR2 has been reported to be the recognition motif to export NCOR2 out of the nucleus for proteasome-mediated degradation (32). This phosphorylation site is deleted in BQ323636.1, which may protect it from protein degradation, whereas the NCOR2 wild-type that contains the phosphorylation site is recognized to undergo protein degradation. This may be a possible explanation for the relatively high protein expression of BQ323636.1 observed in breast cancer cell lines.

By revealing the clinical significance and the underlying function of BQ323636.1 in tamoxifen resistance and breast tumor progression, splice variant BQ323636.1 may be developed as a potential prognostic biomarker for tamoxifen treatment response and furthermore as a novel therapeutic target to overcome tamoxifen resistance in clinical application. Tamoxifen resistance is a complex outcome involving several mechanisms and pathways. It is unlikely that any single gene or mechanism confers tamoxifen resistance. In this study, we showed that alternatively spliced variants may also play a critical role in tamoxifen resistance, which is an important perspective that was neglected often in previous tamoxifen resistance studies. By integrating splice variant that is significant in causing tamoxifen resistance with other previous well-characterized targets, a combination set of more effective therapeutic targets may be developed as a novel therapy to overcome this clinical obstacle in breast cancer.

No potential conflicts of interest were disclosed.

Conception and design: K.Y.K. Chan, U.-S. Khoo

Development of methodology: L. Zhang, C. Gong, S.L.Y. Lau, K.Y.K. Chan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Zhang, C. Gong, S.L.Y. Lau, J.W.H. Tsang, U.-S. Khoo

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Zhang, C. Gong, K.Y.K. Chan

Writing, review, and/or revision of the manuscript: L. Zhang, C. Gong, N. Yang, J.W.H. Tsang, K.Y.K. Chan, U.-S. Khoo

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Zhang, A.N.Y. Cheung, J.W.H. Tsang, K.Y.K. Chan

Study supervision: K.Y.K. Chan, U.-S. Khoo

Verified the cell lines used in the study: O.G.W. Wong

The authors thank Drs. Robert Clarke (Georgetown University Medical School, Washington, D. C.) for providing the cell line AK-47 and LCC2, Carolyn Smith for providing the luciferase reporter ERE-E1b-Luc, Hung-Ying Kao for providing the luciferase reporter ERE-TK-Luc, Jovian Y. C. IP for preparing cell lines and tissue samples for the microarray study, and Tori Chan, our former lab manager, for constructing the pcDNA3.1-6xHis-BQ for this project.

This study was supported by grants from the Hong Kong Research Grant Council (HKU 767307M), and the Committee on Research and Conference Grants from the University of Hong Kong (CRCG; 200611159097) and (201109176081).

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.