ERα-XPO1 Cross Talk Controls Tamoxifen Sensitivity in Tumors by Altering ERK5 Cellular Localization

Abstract

Most breast cancer deaths occur in women with recurrent, estrogen receptor (ER)-α(+), metastatic tumors. There is a critical need for therapeutic approaches that include novel, targetable mechanism-based strategies by which ERα (+) tumors can be resensitized to endocrine therapies. The objective of this study was to validate a group of nuclear transport genes as potential biomarkers to predict the risk of endocrine therapy failure and to evaluate the inhibition of XPO1, one of these genes as a novel means to enhance the effectiveness of endocrine therapies. Using advanced statistical methods, we found that expression levels of several of nuclear transport genes including XPO1 were associated with poor survival and predicted recurrence of tamoxifen-treated breast tumors in human breast cancer gene expression data sets. In mechanistic studies we showed that the expression of XPO1 determined the cellular localization of the key signaling proteins and the response to tamoxifen. We demonstrated that combined targeting of XPO1 and ERα in several tamoxifen-resistant cell lines and tumor xenografts with the XPO1 inhibitor, Selinexor, and tamoxifen restored tamoxifen sensitivity and prevented recurrence in vivo. The nuclear transport pathways have not previously been implicated in the development of endocrine resistance, and given the need for better strategies for selecting patients to receive endocrine modulatory reagents and improving therapy response of relapsed ERα(+) tumors, our findings show great promise for uncovering the role these pathways play in reducing cancer recurrences.

The nuclear hormone estrogen receptor (ER)-α is present in approximately 75% of human breast cancers and is considered one of the most critical predictive factor of breast cancer prognosis (1). ERα is targeted by endocrine therapies, which are generally well tolerated (1). Tamoxifen is one of the most effective therapeutics when a patient is diagnosed with ER-positive (ERα [+]) breast cancer. Although several recent trials showed that a combination of aromatase inhibitors (AIs) and ovarian suppression was effective in premenopausal breast cancer treatment, AIs also have significant adverse side effects in some postmenopausal women, such as increased joint pain, bone fractures, and cardiovascular disease risk (1–3) and the American Society for Clinical Oncology still recommends the use of tamoxifen for premenopausal women (3). Therefore, tamoxifen remains an important endocrine therapy agent in both pre- and postmenopausal women and is expected to continue to be widely used for some time.

The benefit of endocrine therapies is limited, as demonstrated by tumor recurrence in 30% of ERα (+) patients. The recurrence of cancer in ERα-negative (ERα [−]) patients is higher in the first 5 years after the diagnosis, yet for ERα (+) patients, there is a consistent long-term risk of death due to recurrent breast cancer and even a greater risk after 7 years (3). In fact, most breast cancer deaths occur in women with recurrent ERα (+) metastatic tumors (5, 6). Recurrence appears to result from the up-regulation of cytoplasmic-initiated/ERα-dependent kinase pathways that provide an alternative mechanism to support tumor cell proliferation and survival and renders tumor cells resistant to endocrine therapies (7–11). Hormonal regulation of breast cancer involves crucial inputs from estrogens, acting through ERs, and growth factors, operating through growth factor receptors that regulate downstream protein kinase pathways. The relative regulatory inputs to/from these two pathways are thought to underlie the degree to which the cancers remain more indolent and responsive to endocrine therapies vs the acquisition of resistance to these therapies. In the latter situation, the up-regulation of protein kinases serves as a hallmark of endocrine-resistant breast cancer (12–15). Also, whereas the presence of ERα is usually associated with a more favorable prognosis (16–18), it is increasingly appreciated that not all ERα (+) breast cancer patients have a good outcome. A significant subset of patients with ERα (+) breast cancers, specifically those patients characterized as having a luminal B molecular subtype, also have high levels of protein kinase activity, do not respond to tamoxifen as efficiently, and have much less favorable disease-free survival, despite often being treated with additional chemotherapy (19–22).

ERK5, a member of the MAPK family, is present in most human breast tumors and is overexpressed in approximately 20% of these tumors (23). It was also recently identified in a kinase screening study as a major factor that regulates circulating tumor cell invasiveness (24). We have recently identified ERα as the key factor responsible for the activation and regulation of the subcellular localization of ERK5 (25). In this previous work, we showed that its nuclear localization is abrogated by treatment with ERK5 or ERα inhibitors. Based on this information, we used a combinatorial approach in which we took advantage of our tamoxifen-sensitive and tamoxifen-resistant cell culture models, an animal model and data from patient samples to delineate the role of nuclear transport pathways, particularly exportin 1 (XPO1), in tamoxifen sensitivity and endocrine therapy resistance. We identified high levels of XPO1, the major nuclear exporter of the tumor suppressor proteins, as a biomarker for tamoxifen resistance and evaluated its inhibition as a novel means to enhance the effectiveness of endocrine therapies. Our findings suggest that a higher expression of selected nuclear export pathway proteins results in decreased residence times of important nuclear factors that might be involved in proper transcriptional responses to tamoxifen treatment, thus conferring resistance to tamoxifen. Enhanced export to the cytoplasm results in key proteins communicating with other components of the cancer cell machinery involved in cell motility and enhanced kinase signaling, which function to increase aggressiveness of these cells. Our results show that inhibition of nuclear export machinery would improve therapy responsiveness and delay the development of hormone targeted treatment resistance and recurrence.

Materials and Methods

Cell culture, adenovirus, small interfering RNA, and ligand treatments

MCF-7 cells were grown in MEM with non-essential amino acids (NEAA) salts (Sigma), supplemented with 10% calf serum (HyClone) and 100 μg/mL penicillin/streptomycin (Invitrogen) (26). T47D cells were grown in DMEM with 10% calf serum (HyClone) and 100 μg/mL penicillin/streptomycin (Invitrogen). BT474 was cultured in American Type Culture Collection-recommended Hybri-care medium with 10% inactivated fetal bovine serum (FBS), sodium bicarbonate, and antibiotics. MDA-MB-468 and MDA-MB-134 cells were grown in Leibovitz's medium with 20% calf serum (HyClone) and 100 μg/mL penicillin/streptomycin (Invitrogen). HCC1500 cells were cultured in American Type Culture Collection-formulated RPMI 1640 media with 10% FBS, sodium bicarbonate, and antibiotics. All cell lines were obtained from American Type Culture Collection. For experiments with 4-OH-TAM treatment, the cells were maintained in the corresponding phenol red-free medium for at least 3 days and were then seeded at a density of 3 × 10⁵ cells per 10-cm tissue culture dish (Corning) for 2 days before adenovirus infection. Recombinant adenoviruses were obtained from Applied Biological Materials Inc and were used to generate the MCF-7 cells expressing dominant-negative ERK5 or the overexpression XPO1 (AdXPO1) as described previously (25). Adenovirus with a cytomegalovirus construct was used as an infection control (AdCMV).

Gene expression in MCF7 and BT474 cells

Cells were seeded in six-well plates at a concentration of 2.5 × 10⁴ per well in treatment medium (phenol red free medium with 5% charcoal-dextran FBS, NaHCO₃, and antibiotics). The medium was changed on days 2 and 4, and cells were treated on day 6 for 24 hours with ligands: vehicle (Veh; EtOH), 17β-estradiol (E2; 10⁻⁸ M), 4-hydroxytamoxifen (4-OH-TAM; 10⁻⁶ M); Selinexor (SXR; 10⁻⁷ M) + Veh, SXR + E2, and SXR + 4-OH-TAM. Total RNA was extracted with TRIzol reagent (Life Technologies) according to the manufacturer's protocol. The concentration and quality of RNA was determined with a BioTek Cytation 5 plate reader. Reverse transcription was conducted with Moloney murine leukemia virus reverse transcriptase (BioLabs). Quantitative PCRs (qPCRs) were done with Fast Start universal SYBR Green reagent (Roche) using Applied Biosystem Step One Plus qPCR system. The primer sequences were obtained from the Integrated DNA Technology web site. 36b4 was used to normalize the gene expression level. The relative difference in gene expression level was calculated using the δδcycle threshold method.

In vivo xenograft study in mice

Tumor xenograft studies were performed using the BT474 cell line in immunocompromised female mice based on previously reported protocols (27, 28). We used 6-week-old BALB/c athymic, ovariectomized, nude female mice from Taconic Biosciences. After 1 week of acclimatization, we implanted subcutaneously 0.72 mg, 60-day release E2 pellets from Innovative Research of America to maintain a uniform level of estrogen. The next day we injected subcutaneously into both right and left flank of each mouse 2.5 × 10⁷ BT474 cells resuspended in 50% PBS and 50% Matrigel. Once all the animals harbored tumors of approximately 200 mm³, we randomized five animals to each treatment group. Half of the mice were implanted with vehicle pellets and the other half were implanted with 25 mg, 60-day release TAM pellets. We then randomized each group for vehicle or SXR injection. We performed biweekly injections (Monday and Friday) for 4 weeks. Each mouse was housed individually. Animals were monitored daily by the veterinarians for any signs of starvation, dehydration, stress, and pain. We monitored total weight, food intake, and tumor size using a digital caliper biweekly. Tumors were removed from euthenized mice at the end of the experiment or at the time when tumor size reached 1000 mm³ and then stored at −80°C for further analysis.

Immunofluorescence microscopy and data analysis

MCF-7, MCF-7 TAM R, and BT474 cells were treated with Veh (0.1% EtOH) or 1 μM 4-OH-Tam in the presence or absence of 100 nM SXR for the indicated times. Cells were then washed in PBS and fixed on glass coverslips in 4% paraformaldehyde for 30 minutes and washed two times for 5 minutes in PBS. After incubation in acetone for 5 minutes, another PBS wash was performed and then cells were incubated with antibodies against XPO-1 (Santa Cruz Biotechnology; 1:500), ERα (Santa Cru Biotechnology; 1:1000), ERK5 (Bethyl; 1:2000), or phospho-ERK5 (Upstate, Millipore; 1:500). The next day, the cells were incubated with goat antimouse Alexa 568 or goat antirabbit Alexa 488 secondary antibodies. These slides were mounted and stained using Prolong Gold antifade with DAPI (Molecular Probes) to identify the nuclei.

BT474 xenograft samples were paraffin embedded and sectioned (4–5 μm). After rehydration, antigen retrieval, and blocking, the slides were incubated with XPO1 antibody (Santa Cruz Biotechnology; 1:100). The next day, the slides were incubated with goat antimouse Alexa 568 secondary antibody. These slides were mounted, and stained using Prolong Gold antifade with 4,6-diamidino-2-phenylindole (DAPI) (Molecular Probes) to identify the nuclei.

Samples were imaged using a 63×/1.4 oil DIC M27 objective in a Zeiss LSM 700 or 710 laser-scanning confocal microscope (Carl Zeiss). The images were obtained in a sequential manner using a 488-Ar (10 mW) laser line for phosphorylated ERK5 (pERK5) signal (500–550 nm emission) and 555 nm (10 mW) laser line for ERα (600–650 nm emission). The individual channels were obtained using a sequential scanning mode to prevent bleed-through of the excitation signal. Laser power, gain, and offset were kept constant across the samples and scanned in a high resolution format of 512 × 512 or 1024 × 1024 pixels with two/four frame averaging.

Further quantification of the images was performed in Fiji software (http://fiji.sc/wiki/index.php/Fiji) (29). Briefly, images were converted to eight bits for segmentation for each channel. Images for all channels were background subtracted using a rolling-ball method, with a pixel size of 100. Images were segmented using the DAPI channel. The DAPI images were contrast enhanced using the Otsu algorithm. To split touching nuclei and produce the final nuclear masks, the watershed algorithm was used. The resulting objects that had an area of less than 20 pixels and were close to edges were considered noise and were discarded. The DAPI image was selected as the mask, and the signal from pERK5 and/or ERK5 signal was quantified in the nucleus. Three frames per treatment were quantified. Experiments were repeated two times.

Cell Proliferation, cell cycle progression, invasion, motility, and soft agar assays

For proliferation assays, cells were seeded on 96-well plates at a confluency 2000 cells/well (except MDA-MB-134: 5000 cells/well and HCC-1500: 7000 cells/well) in no-phenol red media with 5% charcoal-dextran (CD) FBS. Cells were treated on day 2 and day 5 with indicated concentrations of (Z)-4-hydroxytamoxifen (Sigma-Aldrich; number H7904) and/or SXR (Selleckchem; number S7252). On day 7 a water-soluble tetrazolium-1 (WST-1) assay (Roche) was used to quantify the cell proliferation ratio. Absorbance was measured at 450 nm using the BioTek Cytation5 plate reader (30). Invasion assays used BDBioCoat Matrigel invasion chambers (BD Biosciences) with 10% FBS as the chemoattractant in the lower chamber as described (30). Motility and soft agar colony formation assays were performed as described (30). Briefly, sterile 2× MEM was mixed 1:1 with 1.2% low-melting-point agarose solution and disposed to six-well plates to form a 1.5-mL bottom base agar layer. Plates were immediately placed into the refrigerator for half an hour. Cells were diluted in 2× MEM and mixed 1:1 with sterile 1% low-melting-point agarose suspension cooled down to 37°C. Then 1.5 mL of the cell-agarose suspension was placed on top of the bottom base agar layer. The final concentration was 6500 cells/well. The plates were placed at 4°C for 15 minutes, and then 1 mL of full growth media was added for cells to recover. Twenty-four hours later, the media were changed to treatment media (5% CD FBS, NaHCO₃, and antibiotics) with the following ligands: Veh (EtOH), E2 (10⁻⁸ M), TAM (10⁻⁶ M); SXR (10⁻⁷ M) +V eh, SXR + E2, and SXR + TAM. Cells were treated once or twice a week, aspirating old media and adding the fresh one. When the colonies were big enough (14–28 d), 500 μL of Giemsa stain was added for 1 minute to each well. Wells were washed three times with PBS. Colonies were counted under the microscope.

Cell cycle distribution was assessed by flow cytometry on ethanol-fixed, ribonuclease-treated and propidium iodide-stained cells. The 10⁶ cells were seeded in 10-cm dishes in reduced serum media (5% CD FBS with antibiotics), left overnight to attach, and treated the next day with ligands (Veh [0.1% EtOH], E2 10⁻⁸ M, 4-OH-TAM 1 μM, SXR 10⁻⁷ M, SXR + E2, and SXR + 4-OH-TAM) for 24 hours. After the treatment, the cells were collected in PBS with 0.1% FBS, washed twice, and fixed with 70% (vol/vol) ice-cold ethanol for 24 hours at −20°C. The next day, after washing twice with PBS, the cells were resuspended in 1 mL of PBS with 50 μg/mL propidium iodide and 0.5 μg/mL ribonuclease A. After 30 minutes of incubation at room temperature, at least 10⁴ cells were analyzed on a BD LSR II flow cytometer. The percentage of cells in S, G₁, and G₂ cell cycle phases was analyzed with FCS Express 4 software.

Tumor sample processing and RT-qPCR analysis of XPO1

Formalin-fixed, paraffin-embedded (FFPE) human breast cancer patient tumor samples obtained at the time of the surgical resection were retrieved from the archives of the Department of Pathology at Carle Foundation Hospital in accordance with an institutional review board-approved research protocol. Total RNA was extracted from 40 patient samples using FFPE RNA isolation kit (Invitrogen) following the manufacturer's protocol. RNA was reverse transcribed into cDNA using gene-specific primers and Superscript reverse transcriptase (Invitrogen). Thirty-three of the 40 samples were found to have an adequate quality and quantity of cDNA to permit further analysis. A molecular subtype assignment of samples was based on 55-gene RT-qPCR analysis of the PAM50 gene panel using the StepOne Plus system (Life Technologies) as earlier described (31). Quantitative RT-PCR for the XPO1 gene was performed in parallel on all 33 molecular subtyped human patient samples.

Western blot analysis in cell lines

On day 1 the cells were seeded on 10-cm plates at a concentration of 100 000 per plate in no-phenol red media with 5% CD FBS. The media were changed on day 3. On day 6 whole-cell lysates were collected in radioimmunoprecipitation assay buffer (Thermo Scientific) with 1× Complete protease inhibitor (Roche). Samples were sonicated and boiled in sodium dodecyl sulfate loading buffer and then resolved on precast gels (Bio-Rad Laboratories) and transferred to a nitrocellulose membrane. Membrane was blocked in Odyssey blocking buffer (LI-COR). The following antibodies were used: pERK5 Thr218/Tyr 220 (number 3371; Cell Signaling), Lamin B1 (ab16048; Abcam), NUP153 (A301–789A), KPNA3 (A301–626A), NUP205 (A303–935A), RANGAP1 (A302–026A), KPNA2 (A300–483A), XPO1 (A300–469A), ERK5 (A302–656A), all from Bethyl Laboratories, and ERK2 antibody (D-2 sc-1647; Santa Cruz Biotechnology) were used in 1:1000 dilution. β-Actin (Sigma; SAB1305546) was used in 1:10 000 dilution. The secondary antibodies were obtained from Odyssey and were used at a working concentration of 1:10 000. The images of the membranes are obtained by a LI-COR Odyssey CLx infrared imaging device and software (32). To compare the levels of nuclear receptor signature genes in different cell lines, we normalized the signal from the signature protein to the β-actin signal. The average values and SEM from three experiments were reported.

For the preparation of nuclear/cytoplasmic fractions, cells were seeded at 2 × 10⁶ cells per 10-cm dishes. The media were changed at day 2. At day 4, the cells were treated with Veh (0.1% EtOH) or 1 μM 4-OH-TAM for 45 minutes. Whole-cell samples were suspended in cytosolic extraction buffer (10 mM HEPES, pH 7.9; 10 mM KCl; 0.1 mM EDTA; 0.1% Nonidet P-40; and protease/phosphatase inhibitors), and they were centrifuged at 12 700 rpm for 10 minutes at 4°C. Cytosolic fractions were transferred into a fresh tube and kept on ice. The pellets were washed with cytosolic extraction buffer without Nonidet P-40 and phosphatase/protease inhibitors three times and dissolved in lysis buffer (0.5 M EDTA; 1 M Tris HCl; pH 8; 10% sodium dodecyl sulfate; 10% Empigen; and phosphatase/protease inhibitors). After a 10-minute incubation on ice, the nuclear fractions were sonicated once at 20% amplitude for 5 seconds and centrifuged at 12 700 rpm for 10 minutes. Supernatants (nuclear fractions) were transferred into fresh tubes.

Tumor data sets and data analysis

The log2 median-centered intensity expression values for signature genes were obtained from the Oncomine database (33). Hierarchical clustering of data was performed and displayed using Cluster 3.0 (34) and Java Tree View (35) software for analysis and visualization. Patients were stratified according to the average expression value of the genes in the signature, and data from all of the patients were used for computation of the Kaplan-Meier curves (36) by the Cox-Mantel log-rank test (37) and the Gehan-Breslow-Wilcoxon test (38) in GraphPad Prism version 6.04 for Windows (GraphPad Software, www.graphpad.com).

Inclusion criteria for breast cancer data sets included a report of at least one time-to-outcome event (survival or recurrence) as a continuous variable, reporting of ER and triple-negative status and the reporting of a form of staging or grading. If multiple forms of staging or grading were reported, the N stage was used; if only the grade was reported, the levels of the grade were set equal to the levels of the staging. Menopause status and treatment were collected if reported.

Analysis consisted of a Weibull accelerated failure time model (39) fit singly for each of the gene expression values of interest and included the ER status, triple-negative status, and stage as covariates. The model considered either time to death from disease or time to recurrence, with right censoring for all other outcomes. Each model was fitted both within and across all studies; the latter models included the study as a fixed-effect covariate. Subset analyses were conducted to consider the interaction between the gene expression effect and, separately, menopausal status and treatment. Because the number of data sets available for the subset analyses were small, the combined model fitting included the data source as a fixed effect only. Multivariate models were fitted using Bayesian information criterion (40)-based backward selection. All models were fit using the survival package (41) in R 3.0.3 accessed through Revolution R Enterprise 7.2.0 (2014, Revolution Analytics, Inc). To assess the specificity and sensitivity of the tested markers, receiver-operating characteristic curves were generated using the verification package (42) in R 3.0.3.

Results

ERK5 nuclear localization is reduced during the course of tamoxifen resistance

To characterize the state of the cellular localization of active ERK5 upon 4-OH-Tam treatment, we monitored the localization of ERK5 and pERK5 in a cell culture model of breast cancer endocrine resistance. We grew the initially tamoxifen-responsive MCF-7 cells in continuous 4-OH-TAM treatment for 100 weeks. We and others have earlier demonstrated that this model mimics the development of endocrine resistance and for the characterization of molecular changes associated with the tamoxifen-resistant phenotype (43). When we monitored the localization of ERK5 in cells that were in 4-OH-TAM for 10, 50, and 100 weeks, we found that ERK5 and pERK5 localization to the nucleus was lost progressively compared with parental MCF-7 cells. (Figure 1A). This was consistent with an increased cytosolic localization of pERK5 as the resistance progressed (Figure 1B). Moreover, ERK5 signaling contributes to motility and invasiveness in TAM-resistant breast cancer cells similarly as in ERα (−) cell lines (Figure 1, C and D), whereas in parental MCF-7 cells, ERK5 mainly regulated cell cycle progression as we previously reported (25).

Figure 1. ERK5 nuclear localization is lost in MCF-7 cells that are tamoxifen resistant.

A, ERK5 (upper panel) and pERK5 (lower panel) immunostaining after 45 minutes of 4-OH-TAM treatment in tamoxifen-sensitive MCF-7 cells and tamoxifen-resistant MCF-7 TAM R cells that were kept in 4-OH-TAM for the indicated times. MCF-7 cells or MCF-7 TAM R cells at different stages of resistance progression were treated with 1 μM 4-OH-TAM for 45 minutes, and immunofluorescence microscopy was performed with an antibody specific to ERK5 or pERK5. Nuclei were stained with DAPI. Three fields per slide were quantified (n = 3). A one-way ANOVA model was fitted to assess the contribution of tamoxifen resistance progression on 4-OH-TAM treatment-induced ERK5 or pERK5 nuclear localization. When the main effects were statistically significant at α = .05, pairwise t tests with a Newman-Keuls correction were used to identify the time that ERK5 or pERK5 localization was significantly different from parental MCF-7 cells. *, P < .05; **, P < .01; ***, P < .001; ****, P < .0001. B, Nuclear localization of pERK5 decreases as tamoxifen resistance progresses. MCF-7 parental cells or MCF-7 TAM R cells that were kept in media containing 4-OH-TAM for 50 or 100 weeks were fractionated, and whole-cell lysate and cytosolic or nuclear fractions were subjected to Western blot analysis using pERK5, Lamin1b (as nuclear fraction marker), and β-actin (as cytosolic fraction marker) antibodies. C, ERK5 activity regulated migratory potential in MCF-7 TAM R cells. MCF-7 TAM R cells were infected with AdCMV or dominant-negative AdERK5 for 24 hours and then seeded on the upper chamber of a transwell system for migration assays (B) or invasion assays (C). The number of cells that migrated/invaded to the bottom side of the chambers was counted. Values are presented as mean ± SEM from two independent experiments. AdERK5, adenovirus construct for ERK5.

Derivation of nuclear transport signature

Because the lack of nuclear ERK5 localization in tamoxifen-resistant cells suggested a potential role for nucleocytoplasmic transport machinery in the development of resistance, we monitored the expression of genes important for nuclear transport. To study the differential expression of nuclear transport genes that are most pertinent to clinical outcomes in breast cancer patients, we used the data from publicly available tumor data sets (Figure 2A). We first generated a comprehensive list of nuclear transport genes using the web-based gene set enrichment analysis database. Next, we used Oncomine tumor data sets to select genes that were differentially expressed in ERα (+) vs ERα (−) tumors (n = 49 genes). Interestingly, the expression of these 49 genes were overall lower in ERα (+) tumors. Furthermore, we used survival and outcome data analysis to generate a final list of 13 genes, some of which were previously shown to be modulated in different pathological conditions (Figure 2B).

Figure 2. Derivation of nuclear transport signature and relation to tumor outcome.

A, Flow chart presenting derivation of nuclear transport signature. B, List of nuclear transport signature genes and heat map of expression of each gene in different breast cancer subtypes. The root tree shows similarities between different breast cancer subtypes when it comes to expression in signature pattern. The analysis of publicly available tumor data was performed using web-based Geneanalytics software using 13 genes as the input and with the following variables: 250 months, PAM50 subtype, median split, comparison with tamoxifen, any chemo and end point recurrence-free survival (RFS), or distance metastasis-free survival (DMFS). C, Comparison of recurrence-free survival of patients who are stratified, depending on high and low expression of nuclear transport signature in different breast cancer subtypes using analysis from panel B. D, Nuclear transport signature predicts recurrence of free and DMSF of tamoxifen treated patients. Recurrence-free survival and metastasis-free survival analysis using our nuclear transport gene list and patient gene expression and survival/recurrence data from GSE2034 (48), GSE20711 (49), GSE6532 and GSE6532_1 (50), GSE7390 (51), and GSE9195 (52, 53) using Geneanalytics software and GraphPad software.

Of the 13 selected genes, nine encoded proteins that are either structural components of nuclear pore complex, such as NUP153 and tetratricopeptide repeat motif (TPR), or other proteins that play a role in the transport of the cargo by interacting with the cargo in the cytoplasm or nucleus, such as XPO1 and KPNA2. XPO1 is an exportin that binds to the nuclear export sequence of cargo proteins, among them the major tumor suppressor proteins like p53, p21, pRb, FOXO, survivin, and inhibitory-κB (44) and exports them out of the nucleus and is already being evaluated in multiple later stage clinical trials in patients with relapsed and/or refractory hematological and solid tumor malignancies (45). KPNA2, KPNA3, and KPNA5 are karyopherins, belonging to the importin-α family and are involved in nuclear protein import. KPNA2 has recently been identified as a target of estrogen signaling in breast cancer cells (46). IPO5 is an importin that is important for the nuclear transport of proteins. Several of these proteins, including NUP153 and TPR, were shown to exhibit transcription-dependent mobility in the cell, suggesting an important role for these proteins in ERα-mediated transcription (47). The expression level of all these genes was highest in basal-like tumors followed by Her2 (+) tumors. Interestingly, luminal B type tumors, which are ERα (+) but more aggressive than luminal A tumors due to the increased kinase signaling have similar overexpression of the signature genes as basal and Her2 (+) tumors, which are also more aggressive, are harder to treat, and have a poorer prognosis. (Figure 2B). More than 50% of breast cancer patients had a high expression of nucleocytoplasmic transport signature in luminal B, Her2 (+) and basal subtypes (Supplemental Figure 1A).

Next, we monitored the predictive power of these 13 genes in six tumor data sets including GSE2034 (48), GSE20711 (49), GSE6532 and GSE6532_1 (50), GSE7390 (51), and GSE9195 (52, 53) using the Geneanalytics tool (http://geneanalytics.duhs.duke.edu/). High expression of signature genes predicted a worse survival when we did not distinguish between the cancer subtypes. Next, we tested whether the impact of the gene signature on survival increased when we divided samples based on their subtypes. When we divided our data set based on PAM50 scores, the luminal B subtype with a high expression of these genes had the worst outcome (Figure 2C). Interestingly this was not the case for Her2-overexpressing tumors because a lower expression of the signature genes predicted a worst outcome in this subtype (Figure 2C). The predictive power of our gene signature improved further when we divided luminal B samples based on tamoxifen treatment status (Figure 2D). Those patients who had low expression levels of nuclear transport genes (Figure 2D, black line, low) had survival similar to those of high expression (Figure 2D, red line, high) but responded favorably to tamoxifen (Figure 2D, green line, low-TAM). However, those patients with a high expression level of signature gene levels did the worst (Figure 2D, red line, high), and treatment with tamoxifen slightly improved survival, but the improvement was not statistically significant at an α = .05 (Figure 2D, blue line, high-TAM). Recurrence and distant metastasis-free survival of luminal B patients with a high expression level of nuclear transport signature genes did not improve with tamoxifen treatment (Figure 2D). Thus, the average expression of these genes predicted those patients in the luminal B subtype that would respond to tamoxifen well. These results show that we selected the most significant nuclear export regulators that have been implicated in disease-free survival and distant metastasis-free survival of tamoxifen-resistant breast cancers. Based on these data, we propose that during the development of resistance to tamoxifen, nuclear export components are selectively up-regulated.

Characterization of XPO1 and the other signature genes in tumor data sets

To further understand the contribution of each gene to the predictive power of the identified nucleocytoplasmic transport signature, we characterized the effect of individual genes on patient outcomes using 10 publicly available tumor data sets. The selection criteria, data sets, and results for each gene individually are depicted in Supplemental Figure 2 and Supplemental Table 1. The expression levels of the signature genes were highly correlated, and the correlation levels of the genes are shown for the hormone-treated survival (Supplemental Figure 3) and recurrence (Supplemental Figure 4) data sets.

Survival

We used publicly available data from 10 studies to assess the individual contribution of the expression of our set of 13 genes to overall survival time (Supplemental Figure 2). We used the two publicly available data sets (54, 55) that reported any treatment information to assess the role of treatment in the relationship between gene expression values and survival. By fitting individual accelerated failure time models to each gene, we determined that increased XPO1 expression increased mortality rates in women receiving a combination of chemotherapy, hormone treatment, and radiation and also was associated with increased mortality in women receiving chemotherapy combined with tamoxifen treatment (Supplemental Figure 5A).

We then used these publicly available data to assess the influence tamoxifen treatment had on the association between expression of these 13 genes and the overall survival. In two of these data sets reported elsewhere (54), 2450 women were treated with tamoxifen, of which 732 women died of cancer, on average 2133 days after diagnosis. Using this subset of the data to fit individual accelerated failure time models for each gene, we obtained statistically significant associations between all genes and survival time (Supplemental Figure 6). XPO1 increased survival after tamoxifen in two different data sets (Supplemental Figure 5B). Thus, we conclude from the analysis of publicly available survival data that the specific genes whose expressions are significantly associated with survival depend on factors including hormone treatment and chemotherapy.

Recurrence

There were 12 studies providing information on time to recurrence (Supplemental Figure 7). Eight of the 13 signature genes, fitted individually in accelerated failure time models with no covariates, were found by the fixed-effect model (ie, the model fitted across the 12 studies) to have expression values significantly associated with time to recurrence (Supplemental Figure 8): IPO5, KPNA3, NUP153, RANBP2, RANGAP1, TOMM22, TPR, and XPO1. Two data sets included only women who received hormone therapy (50), both of which were used to analyze the effect of the 13 signature genes in tamoxifen-treated women. In these data sets, 164 women were included and 41 suffered a recurrence at, on average, 1970 days after diagnosis. Increased expression of XPO1 was associated with an increased overall recurrence and recurrence after tamoxifen in different data sets (Supplemental Figure 5, C and D). We conclude from the analysis of publicly available recurrence data that the specific genes whose expressions are significantly associated with recurrence depend on factors including hormone treatment and chemotherapy. Whereas the list of all genes associated with recurrence differed from the list of all genes associated with survival, XPO1 was associated with both outcomes.

Nucleocytoplasmic transport signature genes are regulated by 4-OH-TAM treatment in ERα (+) breast cancer cell lines

To further characterize the molecular basis of the increase in expression of nuclear transport genes and their effect on tamoxifen response, we verified our findings in various cell lines that corresponded to different subtypes of breast cancer. To examine whether ERα ligands induced the expression of nucleocytoplasmic genes, we treated MCF-7 cells with 10 nM E2 or 1 μM 4-OH-TAM for 24 hours and examined the mRNA expression of the genes using qPCR. Eight of the 13 signature genes were stimulated 1.5-fold or more at this time point by 4-OH-TAM treatment (Figure 3A). Moreover, using published ERα chromatin immunoprecipitation-sequencing data (56) in MCF-7 cells and MCF-7 Tam R cells, we showed that these genes had at least one ERα binding site within 100 kb of their promoters, suggesting that direct ERα recruitment after ligand treatment was responsible for the increase in gene expression (Figure 3B). In addition, short- and long-term treatment of MCF-7 cells with 4-OH-TAM increased the protein expression of several of the signature genes including XPO1, RANGAP1, NUP205, NUP153, and KPNA2 (Figure 3C). When we compared the expression of the protein expression of the same genes in tamoxifen-sensitive MCF-7 and T47D cells with that of in tamoxifen-resistant BT474, MDA-MB-134, and HCC-1500, we found that the protein level of four of these factors (XPO1, RANGAP1, NUP205, and NUP153) were higher in the tamoxifen-resistant cell lines consistently (Figure 3D). Based on this evidence, we propose that nucleocytoplasmic transport signature is likely to be a marker for the risk of failure of endocrine therapies and might also be a key element controlling the localization of key signaling molecules that underlie the development of endocrine resistance.

Figure 3. Expression of nuclear transport signature genes are stimulated by 4-OH-TAM in ERα (+) breast cancer cells.

A, ERα ligands increase mRNA expression of signature genes in MCF-7 cells. MCF-7 cells were treated with 10 nM E2 or 1 μM 4-OH-TAM for 24 hours. Total RNA was isolated and mRNA expression of signature genes were analyzed using qPCR. B, ERα recruitment to gene-regulatory regions of nuclear transport signature genes upon tamoxifen treatment in MCF-7 cells or in MCF-7 TAM R cells. Dendograms for ERα occupancy in MCF-7 TAM R cells (black) and MCF-7 cells that are treated with 4-OH-TAM (blue) from the University of California, Santa Cruz, Santa Cruz, California) genome browser. BED files are obtained from another report 4). C, Protein expression of nuclear transport signature genes are increased with ligand treatment in MCF-7 cells and MCF-7 TAM R cells. Western blot analysis of XPO1, RANGAP1, NUP205, NUP153, and KPNA2 and ERK2 as the loading control for MCF-7 parental cells or MCF-7 TAM R cells that were kept in media with 4-OH-Tam for indicated times. The experiment was repeated three times. Representative results are shown. D, Comparison of signature gene expression in different cell lines. ERα (+) MCF-7, T47D, BT474, MDA-MB-134, and HCC-1500 cells were cultured as described. Western blot analysis of XPO1, RANGAP1, NUP205, NUP153, and KPNA2 and β-actin as the loading control for all cell lines was performed. Representative results from the Western blot analysis are shown. The experiment was repeated three times. Signal from each antibody is calculated and normalized relative to the β-actin signal. Level of proteins in cell lines is reported relative to MCF-7 cells. For proteins that increase in tamoxifen-resistant cells relative to MCF-7 cells average relative expression and SEM is plotted. Ctrl, control.

XPO1 inhibitor SXR resensitizes tamoxifen-resistant breast cancer cell lines

To test the feasibility of targeting the nuclear export pathways, we focused on XPO1, an exportin that binds to nuclear export sequence of cargo proteins and exports them out of the nucleus. In our analysis with tumor samples, high XPO1 values are associated with a poor outcome in all women who are treated with hormone therapy with or without chemotherapy. Moreover, XPO1 was previously indicated in the nuclear export of ERK5 (57). In addition, XPO1 is already being targeted in other therapy-resistant cancers, including leukemias (58) and prostate cancers (59, 60), with a highly specific small molecule inhibitor, SXR (61). SXR is orally active and is generally well tolerated. It has manageable side effects including nausea, fatigue, and anorexia that improve over time on treatment. Even in patients who remained on therapy for more than 8 months, no significant cumulative drug toxicities have been identified (62).

Because we found a correlation between XPO1 expression and the failure of the tamoxifen treatment in our preliminary analysis (Figure 2D), we hypothesized that the treatment of tamoxifen-resistant breast cancer cells with an XPO1 inhibitor would improve the sensitivity of these cells to endocrine therapies. To test our hypothesis, we initially performed dose-response studies in various cell lines to identify the ideal dose to treat these cells (Figure 4A). We also treated 4-OH-TAM responsive, MCF-7, and resistant, BT474, HCC-1500, MDA-MB-134, and MCF-7 TAM R cell lines with increasing doses of 4-OH-TAM and/or SXR. In all of the cell lines, cotreatment with 4-OH-TAM and SXR caused a left shift in the nonlinear regression curves, suggesting an improved response to any of the agents when the combination is applied (Figure 4B). Based on these results, we used SXR at 10⁻⁷ M in the rest of our experiments. Next, we treated tamoxifen-responsive MCF-7 and T47D cells and tamoxifen-resistant MCF-7 TAM R, BT474, MDA-MB-134, and HCC-1500 cells with increasing doses of tamoxifen in the presence or absence of 10⁻⁷ M SXR. Inhibitor treatment decreased proliferation in tamoxifen-sensitive cells and blocked tamoxifen-induced proliferation in tamoxifen-resistant cell lines (Figure 4C).

Figure 4. XPO1 inhibitor resensitizes tamoxifen-resistant breast cancer cells to tamoxifen.

A, ERα (+) breast cancer cells are more sensitive to SXR treatment. Dose response for the SXR in ERα (+) MCF-7, T47D, and BT474 cells and ERα (−) MDA-MB-453 and MDA-MB-468 cells. IC₅₀ values were calculated using a nonlinear regression analysis in GraphPad software. B, Impact of single and combination of 4-OH-TAM and SXR treatments in various ERα (+) cell lines. Cells were treated with increasing doses of 4-OH-TAM and/or SXR. Cell numbers were examined using the WST-1 assay at day 7. Values are the mean ± SEM from at least two independent experiments. C, Impact of XPO1 inhibitor on cell proliferation. Cells were treated with increasing doses of 4-OH-TAM in the presence or absence of 10⁻⁷ M SXR. Cell numbers were examined using the WTS-1 assay at day 7. Values are the mean ± SEM from at least two independent experiments.

XPO1 is increased in luminal B subtype patient tumor samples and human breast cancer cell lines and increased XPO1 levels increase cell proliferation of tamoxifen-resistant cell lines in the presence of 4-OH-TAM

First, we verified our findings in patient tumor samples (Figure 5A). In our studies in an independent cohort of human breast cancer patient tumor clinical samples, we found that XPO1 and other signature genes had higher mRNA levels in the luminal B molecular subtype of ER-positive breast cancers that is characterized by significantly worse disease-free survival compared with ER-positive cancers of the luminal A molecular subtype (Figure 5A) (19–22). Our experiments in MCF-7 and BT474 cells indicated that XPO1 is important for G₁-S phase transition in MCF-7 cells, in which in BT474, XPO1 activity modulates 4-OH-TAM-induced cell proliferation by regulating both G₁-S and G₂-M phase transition because SXR reduces the percentage of cells in both the S and G₂ stages (Figure 5B). Of note, 4-OH-TAM treatment increased the percentage of cells in G₂-M transition, and treatment with SXR reduced the number of cells in this phase (Figure 5B). When we tested whether XPO1 inhibition modulated anchorage-independent growth of BT474 cells, which proliferate in the presence of 4-OH-TAM, we observed that treatment with SXR decreased the number of colonies formed. The colonies that formed in the presence of E2 and 4-OH-TAM were larger and had a more dispersed shape, which was abrogated by SXR treatment (Figure 5C). Overexpression of XPO1 using an adenovirus system resulted in at least a 2-fold increase in the level of XPO1 mRNA and protein in MCF7 and BT474 cells increased active ERK5 levels in both of the cell lines (Figure 5D). Moreover, XPO1 overexpression increased the proliferation of both cell lines, which was blocked by the XPO1 inhibitor (Figure 5E).

Figure 5. XPO1 is increased in luminal B subtype patient tumor samples and human breast cancer cell lines, and increased XPO1 levels increase cell proliferation of tamoxifen-resistant cell lines in the presence of 4-OH-TAM.

A, Verification of XPO-1 levels in patient tumor samples. mRNA from tumor FFPE samples were isolated. After the quality of RNA was verified, qPCR was run for XPO1 and 36B4 as control. A one-way ANOVA model was fitted to assess the contribution of subtype to expression of XPO1 mRNA. When the main effects were statistically significant at α = .05, pairwise t tests with a Newman-Keuls correction were used to identify the subtype that had the highest level of XPO1 expression. ***, P < .01. B, Impact of XPO1 inhibitor on cell cycle progression. Cells were treated with Veh or 1 μM 4-OH-TAM in the presence or absence of 10⁻⁷ M SXR. Cell numbers were examined using the FACS analysis. C, Anchorage-independent growth of tamoxifen-resistant BT474 cells treated or nontreated with XPO1 inhibitor (SXR). Colony formation was visualized with Giemsa staining. On the right are representative pictures of spheroid, single colonies formed by BT474 cells. A two-way ANOVA model was fitted to assess the contribution of ligand (Veh, E2, or tamoxifen) and inhibitor (Ctrl and SXR) treatment on anchorage-dependent growth of BT474 cells. When the main effects were statistically significant at α = .05, pairwise t tests with a Bonferroni correction were used to identify whether treatment were statistically different from each other. **, P < .01; ****, P < .0001. D, XPO1 overexpression in MCF-7 and BT474 cell increase ERK5 activation. MCF7 cells (left panel) or BT474 cells (right panel) were infected with AdCMV as control or AdXPO1. XPO1mRNA overexpression was detected by qPCR. XPO1 protein overexpression was assessed by Western blot of immunofluorescence analysis. A t test was applied to assess whether AdXPO1 virus infection resulted in statistically significant overexpression of XPO1 in each cell line. *, P < .05 E, XPO1 overexpression in MCF-7 and BT474 cells increase cell proliferation. MCF-7 cells (left panel) and BT474 cells (right panel) were infected with AdCMV as control or AdXPO1, and then cells were treated with increasing doses of 4-OH-TAM in the presence or absence of 100 nM SXR. Ctrl, control; FACS, fluorescence-activated cell sorter.

XPO1 activity is required for tamoxifen preferential gene regulation in breast cancer cell lines and regulates subcellular localization of pERK5

To determine whether XPO1 was required for ERα-mediated gene transcription in response to estrogen and tamoxifen, we treated MCF-7 cells in the presence and absence of SXR, a bioavailable small inhibitor that binds covalently to XPO1 and inhibits nuclear export. E2-mediated gene regulation events were not affected by the inhibitor treatment; however, the stimulation of tamoxifen preferential gene FOXM1 and YWHAZ (43) was lost in the presence of the inhibitor (Figure 6A). To test whether the 4-OH-TAM-mediated localization of pERK5 was different in tamoxifen-resistant and -sensitive cell lines, we treated MCF-7 cells and BT474 cells with Veh or 4-OH-TAM for 45 minutes and then isolated cytoplasmic or nuclear fractions. Western blot analysis of each fraction revealed that 4-OH-TAM treatment resulted in an increase in nuclear pERK5 in MCF7 cells, whereas active ERK5 was in cytoplasm in BT474 cells after 4-OH-TAM treatment (Figure 6B). To test whether this localization disparity in two cell lines was due to XPO1 activity, we treated BT474 cells with Veh or 4-OH-TAM in the presence or absence of SXR for 45 minutes and then subjected cells to immunofluorescence analysis. This experiment showed that in the presence of 4-OH-TAM, active ERK5 was completely extranuclear and treatment of the cells with the SXR relocated ERK5 back into the nucleus (Figure 6C).

Figure 6. Impact of XPO1 activity on 4-OH-TAM preferential gene expression and subcellular localization of pERK5.

A, MCF-7 cells were treated with 10 nM E2 and 1 μM 4-OH-TAM for 24 hours. Expression of PgR, pS2, and tamoxifen preferential genes, FOXM1 and YWHAZ, was analyzed with real-time qPCR using 36b4 as internal control. The normalized fold expression was calculated with the δδcycle threshold method relative to Veh samples. A two-way ANOVA model was fitted to assess the contribution of ligand (Veh, E2, or tamoxifen) and inhibitor (Ctrl and SXR) treatment on gene expression. When the main effects were statistically significant at α = .05, pairwise t tests with a Bonferroni correction were used to identify whether treatment was statistically different from each other. **, P < .01; ***, P < .001; ****, P < .0001. B, 4-OH-TAM treatment increases nuclear pERK5 in MCF-7 cells but cytoplasmic pERK5 in BT474 cells. MCF7 cells and BT474 cells were treated with Veh (0.1% EtOH) or 1 μM 4-OH-TAM for 45 minutes. Nuclear and cytoplasmic fractions were prepared as described. The increase in pERK5 levels in nuclear or cytoplasmic fractions were assessed by Western blot analysis. The experiment was repeated three times and representative results are displayed. C, XPO1 inhibition relocalizes pERK5 into the nucleus in the presence of 4-OH-TAM in BT474 cells. BT474 cells were treated with Veh (0.1% EtOH) or 1 μM 4-OH-TAM in the presence or absence of 100 nM SXR for 45 minutes. Localization of pERK5 was monitored using immunofluorescence analysis. Ctrl, control.

XPO1 inhibitor SXR resensitizes tamoxifen-resistant tumor xenografts to tamoxifen treatment

To verify our findings from cell line experiments in an in vivo system, we performed experiments using BT474 cell xenografts in immunocompromised mice as previously reported (27, 28). In this experiment BT474 cells formed bigger tumors in the animals that are treated with TAM compared with the tumors in animals treated with vehicle (Figure 7, A–C). In fact, XPO1 mRNA and protein expression was also higher in these tumors that were treated with TAM (Figure 7, B and C). Inhibition of XPO1 using biweekly SXR injections blocked tumor growth, but once the injections were stopped, these tumors started to grow back (Figure 7C, blue line). Conversely, we saw a complete disappearance of tumors in animals that received TAM + SXR, and these tumors did not come back weeks after the injections are stopped (Figure 7C, purple line), suggesting that a combination of SXR and TAM is more effective in resensitizing the tumor cells to tamoxifen. We also monitored food consumption and weight of these animals and did not see any statistically significant change in these parameters at α = .05, which suggested that the effects that we observed are due to the inhibition of XPO1 but not due to a decrease in food consumption (Supplemental Figure 9A). Of note, the body weight of animals that received SXR + TAM treatment increased after we stopped the injections, suggesting improved overall health of these animals after tumor disappearance (Supplemental Figure 9B). These results showed the feasibility of inhibiting XPO1 in the wild-type BT474 xenograft model. Based on this evidence, we believe that the nucleocytoplasmic transport signature is likely to be an important element in controlling the localization of key signaling molecules that underlie the development of endocrine resistance.

Figure 7. XPO-1 inhibitor SXR resensitizes BT474 xenograft tumors to treatment with tamoxifen.

Four- to 6-week-old BALB/C nude mice were ovariectomized and after 1 week were implanted with E2 pellets. The next day the animals were subcutaneously injected with 2.5 × 10⁷ BT474 cells in 50% Matrigel. As tumors reached 200 mm³ in the animals, Veh or tamoxifen pellets were implanted subcutaneously. Starting from the next day, the animals were injected ip with Ctrl or 25 mg/kg SXR biweekly for 4 weeks (n = 5 animals per treatment group). A, Tamoxifen induces growth of BT474 xenografts in nude mice. Picture of representative tumors from Veh and tamoxifen-treated animals. B, Tamoxifen treatment increases XPO-1 expression in tumors. Tumors from Veh or tamoxifen-treated animals were harvested. RNA was extracted and cDNA was synthesized. Expression of XPO1 mRNA was assessed using qPCR assay. 36B4 primers were run as control. A t test was used to assess whether the expression of XPO1 mRNA in tumors from Veh or tamoxifen-treated mice were different. *, P < .05. C, Expression of XPO1 protein in tumors from Veh or tamoxifen-treated animals were assessed using immunofluorescence. XPO1 signal/cell intensity was calculated and normalized by dividing the total XPO1 signal by the cell number from each field. A t test was used to assess whether the expression of XPO1 protein in tumors from Veh or tamoxifen-treated mice were different. *, P < .05. D, SXR resensitizes BT474 xenografts to tamoxifen treatment. Tumor size was measured biweekly using a caliper. The tumor volume was calculated using the following: (length × width²) × 3.14/6. A two-way ANOVA model was fitted to assess the time-dependent contribution of ligand (Veh, tamoxifen) and inhibitor (Ctrl and SXR) treatment on tumor volume. When the main effects were statistically significant at α = .05, pairwise t tests with a Bonferroni correction were used to identify whether treatments were statistically different from each other. *, P < .05; **, P < .01; ***, P < .001; ****, P < .0001. Ctrl, control.

Discussion

In our studies, we found that as ERα (+) breast tumors acquire resistance to tamoxifen, a group of nuclear transport proteins including XPO1 will be up-regulated, increasing ERK5 export from the nucleus. Thus, ERα, which is in the nucleus, will not have the partners to elicit proper transcriptional responses to tamoxifen and ERK5, which partners with other cytoplasmic proteins, will now contribute to tumorigenicity and tamoxifen resistance (Figure 8).

Figure 8. Model.

As ERα (+) breast tumors acquire resistance to tamoxifen, a group of nuclear transport proteins including XPO1 will be up-regulated, increasing ERK5 export from nucleus. Thus, 1) ERα, which is in the nucleus, will not have the partners to elicit proper transcriptional responses to tamoxifen and 2) ERK5, which partners with other cytoplasmic proteins, will now contribute to increased tumorigenicity and tamoxifen resistance.

The development of tamoxifen resistance is a major limitation to the effectiveness of treatment of hormone-responsive breast cancer. Whereas ERα presence is usually associated with a more favorable prognosis, it is increasingly appreciated that not all ERα (+) breast cancer patients have a good outcome. A significant subset of patients with ERα (+) breast cancers, such as those patients characterized as luminal B cancers that contain ER but also have high levels of protein kinase activity, have a much less favorable disease-free survival. Blocking the activity of ER using selective ER modulators such as tamoxifen or raloxifene, or the ER-degrading agents fulvestrant or AIs, which reduce estrogen production, has proven highly effective in targeted treatment of hormone-responsive breast cancers (1, 63, 64). Because of the inherent differences in subtypes, luminal B-type tumors are less responsive to antitumor activities of tamoxifen. Our bioinformatics analysis suggest that in this subtype, nuclear transport proteins are up-regulated and when combined with the higher kinase activity in these tumors, localization of key kinases to cytoplasm might explain the reduced effectiveness of tamoxifen in these tumors.

Our findings suggest that the expression of nuclear transport-related genes in ERα (+) tumors might be used to select those patients that would favorably respond to tamoxifen. Moreover, based on our cell line and tumor xenograft studies, by combining the XPO1 inhibitor with tamoxifen treatment, we can improve the effectiveness and duration of tamoxifen treatment. Our approach was built upon our initial findings that estrogens increase the nuclear localization of key signaling molecules like ERK5, and the absence of ERα renders ERK5 extranuclear (25). When outside the nucleus, ERK5 enhances the actin cytoskeleton reorganization and thus contributes to cell aggressiveness and motility, which are characteristics of breast cancers that are resistant to endocrine therapies. This study validated the hypothesis that nuclear export proteins level could be used as a marker for the risk of recurrence. Establishing XPO1 as a target for inhibition would enhance the effectiveness of endocrine therapies by maintaining tamoxifen sensitivity. Targeting the localization of key signaling molecules to cellular localizations in which they can be more efficiently used by ERα to increase efficiency of tamoxifen or other endocrine agents has a promise of higher efficacy and lower toxicity.

Cancer cells of different tumor types have been shown to be more sensitive to XPO1 inhibition than normal cells, including myeloma, in which ratjadone, another XPO1 inhibitor, is shown to be selective and kill myeloma but not normal cells (65, 66). The inhibition of XPO1 in cervical cancer using another small molecule inhibitor, LMB, demonstrated the higher sensitivity of XPO1 inhibition in the cancer vs the normal cells (67).

Further research will be necessary to establish a prognostic test that can be used to identify those ERα (+) patients most likely to respond favorably to tamoxifen and allow identification of those patients who would benefit from XPO1 targeting agents to engender improved tamoxifen sensitivity. Gene expression and immunohistochemistry studies need to be performed to determine baseline values of XPO1 and how it relates to ERK5 localization and tamoxifen responsiveness in the tumors. In addition, XPO1 has other targets in the tumor cells that might modulate responses to antiestrogens such as p53, p21, pRb, or Forkhead box O. XPO1 inhibitors might resensitize tamoxifen-responsive tumors to tamoxifen by modulating localization of these other factors as well. This could have a broad translational importance in the prevention and treatment of late-stage cancers. For example, in cancers in which ERK5 is localized to the cytoplasm, cellular aggressiveness can be down-regulated by the pharmacological inhibition of XPO1, which results in a decreased nuclear export, thus allowing the return of ERK5 into the nucleus, in which it contributes to transcription and effective tamoxifen responsiveness. Thus, important advances in the therapy of late-stage disease and avoidance of complications associated with broad kinase inhibitors could ultimately be expected. Furthermore, our findings might be applicable to other cancers, including therapy-resistant leukemia, prostate cancers, and triple-negative breast cancer, for which highly selective XPO1 inhibitors are already in clinical trials (45, 65, 68). In addition, the findings from our research might contribute to a broader understanding of how XPO1 might modulate the localization of proteins important for the activity of other nuclear receptors including androgen receptor and progesterone receptor.

Our research represents a new and substantive departure from the status quo by shifting the focus to modulating the localization of key proteins rather than modulating their actual activity. Most current research efforts in the therapy resistance field have focused on a delineation of the underlying mechanisms that lead to increased activity of selective signaling pathways. Undoubtedly, interrogating and targeting the end-point kinases in tumors is highly relevant, and these studies led to the development of combination therapies involving phosphatidylinositol 3-kinase inhibitors or mammalian target of rapamycin pathway inhibitors together with endocrine agents. However, resistance to these combination therapies also occurs, and in such cases, the cancer that develops is considerably more aggressive due to hyperactivation of compensatory mitogenic signaling pathways (69). Moreover, these kinase inhibitors have many adverse side effects. More recently, ERα mutations that decrease the sensitivity of the receptor to selective ER modulators and selective estrogen receptor degradors were identified in more than 30% of the metastatic, but not primary, tumors (70–74). However, in two-thirds of ER-positive metastatic tumors, the mechanism of therapy resistance is not attributable to ER mutation, and alterations often cannot be targeted effectively.

Our findings strongly suggest that this approach will be effective in allowing the relocalization of key proteins, such as ERK5, to the nucleus to improve transcriptional response to tamoxifen that would otherwise function to regulate invasiveness and aggressiveness in the cytoplasm. This approach is expected to open new research horizons, particularly in the biology of luminal B and basal-like breast cancers, which are more aggressive and resistant to current therapies. By using an integrative computational and experimental approach, we have generated evidence that XPO1 appears to play key roles in drug resistance in breast cancer. We have postulated that these pathways have not previously been focused on in breast cancer primarily because the effects that we describe pertain primarily to luminal B and basal like type breast cancers. Thus, specific breast cancer subtypes may well have been overlooked in previous studies designed primarily to analyze the overall effect in all breast cancers independent of the molecular subtype. Our findings delving into deciphering the mechanistic details of this relationship and testing the efficacy of targeting these pathways in the clinic show great promise for ultimately delivering novel diagnosis and treatment strategies for therapy-resistant ERα (+) luminal B and ER (−) basal subtype tumors.

In summary, our study reported here is the first attempt in the field to define the causal role of the nuclear export pathways in tamoxifen resistance and explore the feasibility of targeting these pathways to improve the response to tamoxifen and decrease the risk of recurrence.