Acquired Resistance of ER-Positive Breast Cancer to Endocrine Treatment Confers an Adaptive Sensitivity to TRAIL through Posttranslational Downregulation of c-FLIP

Over 70% of breast tumors express estrogen receptor (ER), and it is clear that estrogen itself plays a vital role in tumor development and progression (1). As a result, the nonsteroidal antiestrogen tamoxifen has provided improved survival and quality of life for many early and advanced ER+ve disease sufferers (2, 3).

Since the introduction of tamoxifen to the clinic nearly 30 years ago, further endocrine agents have been developed, notably the steroidal antiestrogen fulvestrant (Faslodex) that also promotes ER degradation (4), and aromatase inhibitors (AI) that suppress estrogen production in the body (5, 6). Although AIs are now the gold-standard endocrine strategy in the ER+ve postmenopausal setting (7), tamoxifen remains a pivotal treatment option in ER+ve premenopausal breast cancer patients with 67% of patients responding to tamoxifen as first-line therapy (8–10). Fulvestrant in turn can prove valuable under conditions where tamoxifen or AIs fail (11).

Unfortunately, a large proportion of patients acquire resistance to endocrine treatment following initial responsiveness, and this is associated with reinstigation of tumor growth and disease progression (12, 13). Thus, the majority of patients with advanced disease and up to 30% of ER+ve patients in the adjuvant setting will acquire resistance to the inhibitory effects of their endocrine treatment (12, 13). The acquisition of resistance manifests with a partial epithelial–mesenchymal transition (EMT) and increased migratory and invasive activity both in anti–estrogen-resistant cell line models (14) and in clinical samples (14–17). Furthermore, a subset of tumor-initiating cells with stem-like characteristics (cancer stem cells: CSCs) is enriched in tamoxifen-resistant breast cancers and may demonstrate resistance to endocrine therapies, thus contributing to disease recurrence and metastatic progression (18–21).

It has been suggested that ER expression remains a stable phenotype for many acquired resistance tumors that initially respond to endocrine agents (22) which is corroborated by data showing that a significant proportion of acquired resistance tumors remain sensitive to alternative endocrine treatment. However, as a therapeutic avenue this too has its limitations, as response to additional endocrine therapies declines over time while some patients also lose ER expression during endocrine treatment (23–25). In vitro and xenograft model studies have elucidated that endocrine agents upregulate various “compensatory” growth factor signaling pathways that sustain cell survival and residual proliferative activity, culminating in resistant growth (26, 27). A number of approaches have been examined in patients targeting such signaling mechanisms to overcome endocrine resistance (28); however, responses to such targeted approaches are short-lived for many patients (29, 30). Of these agents, CDK4/6 (palbociclib, ribociclib) and mTOR inhibitors (everolimus) have shown particular promise; however, toxicities associated with these therapies must be managed carefully and predictive biomarkers for response are currently being investigated to help in patient selection (31–33). There thus remains a critical need for superior treatments guided by predictive biomarkers with greater long-term benefits in the management of endocrine resistance in the clinic. TRAIL is an immune-related apoptotic protein that has been shown to specifically target cancer cells while sparing normal cells. Unfortunately, several studies have now shown that despite this cancer specificity, many tumor cell subtypes are inherently resistant to treatment with TRAIL agonists, with no improvement in overall survival rates in clinical trials (34). In breast cancer for example, epithelial-like ER-positive tumor cells are resistant to the proapoptotic effects of TRAIL, whereas cells with a mesenchymal phenotype, including CSCs within epithelial-like tumors, exhibit some sensitivity to TRAIL (35–37). Yet despite these observations, TRAIL agonists have not progressed well in the clinical setting (34, 38). Although tumor resistance has been largely attributed to poor agonist potency, it is probable that a scarcity of adequately stratified patient populations also partly contributes to resistance through a lack of understanding of the underlying mechanisms of resistance (39). The intrinsic TRAIL-mediated apoptosis inhibitor c-FLIP is believed to be one contributing factor behind TRAIL resistance of breast cancer cells as mesenchymal CSC and non-CSCs exhibit redistribution of c-FLIP (37), and removal of c-FLIP blockade hypersensitized both the bulk tumor cell population (40, 41) and breast CSCs to TRAIL in vitro and significantly reduced metastasis in in vivo models (36).

Given the evidence above that endocrine resistance was also associated with mesenchymal-like traits (including CSCs), we postulated that endocrine-resistant breast tumors may respond better to TRAIL therapy than their endocrine-naïve counterparts. If this were the case, we proposed that the increased TRAIL sensitivity may be mediated by c-FLIP. Here, we tested these hypotheses using in vitro and in vivo cell line models and clinical samples of acquired endocrine-resistant breast cancer. We showed that acquired endocrine-resistant breast cancer cell lines and their constituent CSC subpopulation were indeed sensitive to TRAIL treatment, and that this sensitivity correlated with the reduction of c-FLIP levels through a posttranslational mechanism. Furthermore, the TRAIL-mediated reduction in endocrine-resistant CSCs in vitro was replicated in vivo with a marked reduction in metastatic disease. These findings were supported by patient-derived breast cancer tissue models leading to the conclusion that the selective use of TRAIL agonists to target the CSC compartment in breast tumors that have acquired endocrine resistance may improve clinical outcome with potential long-term benefits to patients.

Parental breast cancer cell lines MCF-7ER+HER2- and T47DER+HER2- (obtained from the ATCC) were maintained in RPMI 1640 medium supplemented with 5% FBS, 1% penicillin—streptomycin, and 0.5% l-glutamine. Tamoxifen- and Faslodex-resistant subclones of these cell lines were derived as described (22, 42, 43) and maintained in phenol red–free RPMI 1640 supplemented with 5% FBS (charcoal stripped for MCF-7 derivatives), 1% penicillin–streptomycin, 0.5% l-glutamine, and 100 nmol/L of 4-OH-tamoxifen (Sigma) or fulvestrant (AstraZeneca).

Small-interfering RNAs (siRNA) targeting two unique sequences in human c-FLIP (FLIPi—sense: GGAUAAAUCUGAUGUGUCCUCAUUA, antisense: UAAUGAGGACACAUCAGAUUUAUCC) and a nonspecific scrambled control (SCi—sense: GGACUAAUAGUUGUGCUCCAAUUUA, antisense: UAAAUUGGAGCACAACUAUUAGUCC) RNA were used in reverse transfections (Invitrogen). Cells were trypsinized and resuspended at a density of 1 × 10⁵ and seeded into wells containing 20 μL of 100 nmol/L siRNA in serum-free Optimem (GIBCO) in a volume of 100 μL per well together with 0.3 μL of Lipofectamine (Invitrogen). Cells were cultured in the presence of siRNA for 48 hours prior to subsequent assay.

Cells were treated with soluble human recombinant TRAIL (SuperKillerTRAIL; Enzo Life Sciences) at a concentration of 20 or 100 ng/mL for 18 hours at 37°C in 5% CO₂.

Western blots of cell lysates were performed as described (44). The following antibodies were used: c-FLIP (Enzo Life Sciences—7F10, ALX-8040428; Santa Cruz Biotechnology—5D8, sc-136160), ER-alpha (Santa Cruz Biotechnology; sc-7207), ErbB2 (Abcam; ab2428), GAPDH (Santa Cruz Biotechnology; sc365062), ITCH (AbCam; 99087), and pITCH (Millipore; 10050).

CellTiter blue cell viability assay (CTB; Promega) was performed according to the manufacturer's instructions in a 96-well plate format and fluorescence assessed using a ClarioStar (BMG Labtech) plate reader, whereas Invitrogen live/dead-labeled cells were stained for 15 minutes at 4°C in the dark and subsequently analyzed by flow cytometry (BD Accuri).

Primary and local relapse human breast tumor pathologic and postsurgical core biopsies were obtained following consent from the Cardiff and Vale Breast Cancer Centre, Llandough Hospital under National Institute for Social Care and Health Research (NISCHR) ethical approval (12/WA/0252), and pleural effusions/ascitic fluid metastatic samples obtained through Manchester Cancer Research Center Biobank and The Christie NHS Foundation Trust (studies 12/ROCL/01, 05/Q1402/25, respectively). Cells were processed, maintained, and stored according to the Human Tissue Act with local (Cardiff University) research ethics approval. All animal xenograft procedures were performed in accordance with the Animals (Scientific Procedures) Act 1986 and approved by the UK Home Office (PPL 30/2849). Tumors excised from mice or received from patients were finely minced and then processed in a MACs tissue dissociator using a MACs tissue dissociation kit according to the manufacturer's protocol (Miltenyi Biotec). Primary cells were maintained in 5% CO₂ and 5% O₂ at 37°C. The patient-derived xenograft (PDX) models were generated either as previously described (45) or for the PDX 151 model, and 3 mm chunks of primary tumor material from a 72-year-old patient with local relapse previously receiving endocrine treatment (see Supplementary Fig. S4A) were serially passaged by subcutaneous implantation in JAXTM NOD.Cg-PrkdcscidIl2rgtm1Wj (NSG) mice.

Cell lines and primary tumor samples were initially treated in adherent culture conditions as described and then dissociated into single-cell suspensions and plated (P1) in ultralow attachment 96-well plates (Corning) at a density of 20,000 cells/mL in a serum-free epithelial growth medium (DMEM/F12; Gibco), supplemented with 2% B27 (Invitrogen), 55 μmol/L β-mercaptoethanol, 20 ng/mL EGF (Sigma), 5 μg/mL insulin (Sigma), and 1 μg/mL hydrocortisone. After 7 days, tumorspheres were collected by gentle centrifugation (300 x g), dissociated in 0.05% trypsin and 0.25% EDTA, and reseeded (P2) at 20,000 cells/mL for subsequent passages. %TFU was calculated by dividing the number of spheres formed by the total number of single cells seeded.

The directly conjugated mouse FITC anti-CD24 antibody (BD; 555427), mouse allophycocyanin (APC) anti-CD44 antibody (BD; 559942), mouse APC anti-DR5 (biolegend; DjR2-4), and mouse FITC anti-DR4 (AbCam; ab59047) were used to stain single-cell suspensions at 4°C in the dark for 30 minutes. Control samples were stained with isotype-matched control antibodies, and samples were analyzed using BD Accuri flow cytometer.

To measure ALDH activity, Aldefluor assay (Stem Cell Technologies) was performed on surviving cell populations as previously described (46). Briefly, dissociated single cells were suspended in assay buffer containing ALDH substrate, bodipyaminoacetaldehyde (BAAA), at 1.5 mmol/L and incubated for 45 minutes at 37°C. To distinguish between ALDH+ve and −ve cells, cells were also incubated under identical conditions in the presence of a 2-fold molar excess of the ALDH inhibitor, diethylaminobenzaldehyde (DEAB).

Tumor initiation and growth were assessed by orthotopic mammary fat pad transplantation of dissociated MCF-7 and MCF-7-TAMR cells. Cells were harvested using 1 mmol/L EDTA, washed, and resuspended at a density of 5 × 10⁶ in serum-free L15 media. A 1.5 mg, 90-day slow release 17-β estradiol or Tamoxifen pellet (Innovative Research of America) for parental and TAMR cells respectively was inserted subcutaneously above the right scapula of anesthetized NOD-SCID mice (Harlan). A total of 1 × 10⁶ cells were orthotopically injected directly into the abdominal mammary fat pad, and when tumors reached 5 mm in diameter, the mice were treated with 16 mg/kg TRAIL (IP) daily for 4 days twice, separated by a 6-day treatment holiday. Mice were then monitored, and tumor volume was measured twice weekly. Metastatic tumor burden within lung tissue was then assessed histologically in serial hematoxylin and eosin sections.

Cultured cells were pelleted via centrifuge at 300 rcf for 5 minutes before being resuspended in 350 μL RLT buffer (Qiagen) and placed on ice for immediate extraction or frozen at −80°C for future extraction. RNA extraction was performed using the Qiagen RNEasy Kit following the manufacturer's instructions. The concentration and quality of RNA were analyzed using a nanodrop 3000 spectrophotometer (Thermo Scientific). Previously isolated RNA was synthesized into cDNA using the QuantiTect Reverse Transcription Kit (Qiagen) according to the manufacturer's instructions.

All qRT-PCR experiments were designed to include both target gene probe and ACTB control probe. For each experiment, TaqMan Universal Master Mix II, with UNG (ThermoFischer Scientific), was used. TaqMan master mix was added to target and ACTB control probes together with RNase free H₂O to make individual target master mixes containing all reaction components and plated in 384-well qPCR plates (Applied Biosystems). Plates were then run on a QuantStudio 7 Real-Time PCR machine (Applied Biosystems) set to the following protocol: initial denaturation at 95°C for 10 minutes, followed by 40 cycles of 95°C for 15 seconds (denaturation), and 60°C for 1 minute (annealing/elongation). Relative quantity was determined using ΔΔ^Ct values calculated from the reference sample.

Data are represented as mean ± SEM for a minimum of three independent experiments, unless otherwise stated. Statistical significance between means was measured using parametric testing, assuming equal variance, by standard t test for two paired samples. The χ² test was used to determine the significance of deviation from expected ratios of responders to nonresponders in the clinical cohort.

The development of endocrine therapy resistance in primary breast cancer is associated with poor prognosis and, in cell line models and clinical samples of endocrine resistance, a transition to more aggressive tumor phenotypes that have undergone partial EMT (14–17). The cancer therapeutic TRAIL has been shown to preferentially induce cell death in breast cancer cell lines that exhibit a more mesenchymal phenotype (35). We wished to investigate therefore whether the adaptation to endocrine treatment also increased the sensitivity of breast cancer cells to TRAIL. Initially, four models of acquired endocrine resistance, derived from continuous endocrine treatment with tamoxifen or Faslodex of the luminal ER+ve breast cancer cell lines MCF-7 (42) and T47D (43), were investigated. Acquired tamoxifen-resistant (TAMR) and acquired Faslodex-resistant (FASR) cell lines were significantly more sensitive to the cytotoxic effects of TRAIL than their parental controls. (Fig. 1A and B; Supplementary Fig. S1A). Sensitivity was particularly prominent in the TAMR derivatives. The gain in TRAIL sensitivity was then confirmed in an independent MCF-7–derived tamoxifen-resistant model (ref. 20; Supplementary Fig. S1B). This marked susceptibility to TRAIL could not be explained by the combinatorial effect of TRAIL with either endocrine agent and was a feature confined to the resistant state, as the combination of tamoxifen or Faslodex with TRAIL treatment on the previously untreated MCF-7 cells (cotreated) or on cells pretreated with these endocrine agents for up to 7 days (pretreatment) did not confer TRAIL sensitivity (Fig. 1C). Furthermore, tamoxifen withdrawal (TAMR-W) for 7 days prior to TRAIL treatment also had no effect on the sensitivity of TAMR cells to TRAIL (Fig. 1D). Combined, these results suggested that the substantially increased TAMR sensitivity to TRAIL was due to a stable adaptation resulting from acquired resistance to tamoxifen.

According to the CSC hypothesis, a drug-resistant subpopulation of stem/progenitor-like tumor cells is likely to contribute to promotion of resistance to conventional therapies (18, 21, 47, 48). To determine if these cells demonstrated a comparable sensitivity to TRAIL to the bulk cell cultures in the acquired endocrine-resistant lines, the effect of TRAIL on stem-like activity was assessed by tumorsphere assay, whereas cancer stem/progenitor cell numbers were measured using cell surface markers and ALDEFLUOR assay. Both the MCF-7–derived TAMR and FASR cell lines exhibited an increased capacity to form tumorspheres, which were similar in morphology and size to their parental MCF-7 cell line. However, in contrast to the parental line, these self-renewing tumorsphere-forming cells were significantly more sensitive to TRAIL, with a complete ablation of tumorsphere-forming units (TFU) and colony-forming units in the TAMR cells following treatment with TRAIL (Fig. 2A and B; Supplementary Fig. S2C). Furthermore, this TRAIL sensitivity was independent of the presence of tamoxifen in the media (Fig. 2C), implying that the hypersensitization to TRAIL was inherent to the acquired endocrine (tamoxifen) resistance phenotype. The effect of TRAIL on the CSC compartment in the TAMR line was further supported by flow cytometry using surrogate markers of stem/progenitor cells. Both CD44HiCD24Lo and ALDH+ve subpopulations were significantly elevated in tamoxifen-resistant cell lines and yet remained sensitive to TRAIL-mediated apoptosis (Fig. 2D), although TRAIL failed to further deplete the CD44 population in TAMR cells. A significant reduction in CSC-like activity was also observed in a second cell line model of acquired tamoxifen resistance (T47D-derived). Although the sensitivity to TRAIL was not as profound, TRAIL-treated TAMR cells were still significantly reduced compared with their TRAIL-treated parental counterparts (Supplementary Fig. S2A and S2B) with a 50% reduction in tumorsphere-forming capacity and a 4-fold reduction in CD44+ve cells in T47D-TAMR cells following TRAIL treatment. In summary, these data suggest that sensitivity of breast cancer stem–like cells to TRAIL develops following the acquisition of resistance to antiestrogens, most acutely with tamoxifen resistance.

In order to determine if the TRAIL-mediated reduction in overall cell viability and CSC-like properties in vitro was reflected in tumor growth in vivo, mice bearing established MCF-7 parental or TAMR orthotopic tumors of approximately 5 mm in diameter were subjected to eight i.p. injections of TRAIL over a period of 2 weeks and then monitored for tumor response. Along with the previously observed elevation in CSC-like and proliferative activity of TAMR cells in vitro (Fig. 2B), orthotopic TAMR tumors were found to be more aggressive than parental MCF-7 controls under vehicle control conditions, with an elevated growth rate and appearance of spontaneous metastases to the lung (Fig. 2E and F). MCF-7 tumors responded transiently to TRAIL treatment (20% reduction in tumor volume) but, within 2 weeks of treatment, had regained similar growth kinetics to the vehicle-treated cohort (Fig. 2E). Treatment of TAMR tumors with TRAIL on the other hand resulted in a complete (100%) and sustained regression of all tumors (n = 8) for up to 15 weeks (Fig. 2E). Furthermore, following TRAIL treatment, both the number and size of TAMR metastatic lesions in the lungs were decreased by more than 95% (Fig. 2F), correlating with the profound decrease in CSC-like activity observed in vitro (Fig. 2B). Taken together, these data suggested that tamoxifen-resistant cell response to TRAIL in vivo correlated with the observed in vitro responses whereby tumor regression occurred by debulking tumor mass, and in turn, metastatic tumor burden and tumor relapse were significantly reduced through the elimination of the cancer stem/progenitor pool.

Having seen clear benefits of TRAIL treatment in cell line models of antiestrogen resistance in vitro and in vivo, the efficacy of TRAIL on endocrine-resistant breast cancer was determined in clinical samples ex vivo. Primary cell cultures were derived from pleural effusions or ascites of 13 patients with advanced ER+ve breast cancer, 11 having received endocrine treatment and 2 endocrine-naïve samples (Supplementary Fig. S3A). Cells were plated into tumorsphere culture prior to treatment with TRAIL ex vivo (Fig. 3A). Nine of 11 (82%) of the clinical samples from the advanced metastatic endocrine-resistant patients exhibited a significant reduction in TFUs following treatment with TRAIL (Fig. 3A), whereas neither of the endocrine-naïve metastatic patient samples (90 and 94) responded to TRAIL. Similarly, 7 of 8 (88%) of the endocrine-resistant patient samples cultured as an adherent monolayer were also sensitive to TRAIL, with only 1 of the 2 endocrine-naïve samples exhibiting a significant response to TRAIL (Supplementary Fig. S3B).

A similar relationship was observed in primary cell cultures derived from 13 fresh core biopsies of local relapse ER-positive breast cancer (Fig. 3A). Of two tumor biopsies originating from patients who had relapsed following treatment with tamoxifen (samples 127 and 188), both demonstrated a significant reduction in tumorsphere-forming capacity, whereas only 1 of 11 (9%) of endocrine treatment-naïve primary tumors exhibited such a response to TRAIL.

Overall, 11 of 13 samples (85%) from all clinical groups that had relapsed following endocrine treatment responded to TRAIL ex vivo, compared with just 1 of 13 tumors that had not previously seen endocrine therapy (Fig. 3A–C). Taken together, the findings from these ex vivo models provide preclinical evidence that TRAIL may be valuable in controlling endocrine-resistant breast cancers and potentially provide long-term benefits by targeting the CSC subset.

The potential for TRAIL as a further therapeutic option following acquisition of endocrine resistance was subsequently tested in vivo using two distinct PDX models of endocrine-resistant breast cancer.

For the first PDX model, a primary ER-positive tumor was maintained orthotopically by serially passaging in immunocompromised mice receiving either estradiol (MaCa 3366) or tamoxifen (MaCa 3366 TAMR) for up to 3 years, generating estradiol-dependent and tamoxifen-resistant models, respectively, from the same parental tumor tissue (45). Mice were subjected to systemic administration of 8 i.p. injections of TRAIL over 2 weeks (treatment windows highlighted in pink, Fig. 4) and tumor growth kinetics monitored from the beginning of treatment (100 mm³). In the tamoxifen-sensitive estrogen-dependent control tumors, systemic in vivo TRAIL treatment failed to elicit a significant growth-inhibitory response compared with vehicle only (Fig. 4A). There was also no alteration in the CSC compartment within the tumors (Fig. 4A). In contrast, a transient (10 day) yet significant regression of MaCa 3366 TAMR tumors was observed following first TRAIL administration, compared with a doubling in tumor volume in the vehicle control arm over the same time period, which after a second dose of TRAIL culminated in a sustained and significant reduction in tumor size at 4 weeks compared with untreated controls (P = 0.0045; Fig. 4B). Thus, although TRAIL-treated naïve 3,366 tumors exhibited a 4-fold increase in tumor volume over 4 weeks, TRAIL-treated endocrine-resistant tumors exhibited a 1.8-fold increase over the same time period (Fig. 4A and B). Furthermore, consistent with observations in our cell line models of tamoxifen resistance, there was a significantly higher basal level of CSC-like cells in MaCa 3366 TAMR tumors compared with its endocrine-responsive, estrogen-dependent, counterpart, and TRAIL treatment in vivo significantly depleted the CSC-like activity in the tamoxifen-resistant model (Fig. 4B), whereas no significant difference was observed in the endocrine-sensitive model. In addition, tumorspheres generated from the TRAIL-treated MaCa 3366-TAMR tumors also maintained their sensitivity to TRAIL when treated ex vivo, demonstrating that the tumorsphere-forming subset surviving in vivo TRAIL administration was not a consequence of acquired TRAIL resistance (Supplementary Fig. S4B).

For the second PDX model, a core biopsy (CUbbs 151) obtained from a locally relapsed ER-negative tumor from an ER-positive (original status prior to endocrine treatment) breast cancer patient who had relapsed on the AI anastrazole was used to propagate tumors in recipient immunocompromised mice to model AI-resistant breast cancer (histopathologically characterized in Supplementary Fig. S4A). AIs have become the gold-standard first-line therapy for the majority of postmenopausal ER-positive breast cancer patients, and many of the endocrine-resistant clinical samples used in this study (Fig. 3) had at some point in their treatment received an AI (Supplementary Fig. S3A). In order to determine AI-resistant breast cancer sensitivity to TRAIL, mice bearing CUbbs 151 tumors (PDX 151) were administrated 8 i.p. injections of TRAIL over 2 weeks. This resulted in a significant reduction in AI-resistant tumor growth compared with vehicle control (Fig. 4C), and ex vivo control PDX 151 tumor cells also demonstrated significant sensitivity of CSC-like cells to treatment with TRAIL (Fig. 4D). Furthermore, the sensitivity of PDX 151 CSCs ex vivo was consistent with the sensitivity of the patient biopsy (CUbbs 151) treated ex vivo prior to transplantation (i.e., original biopsy straight from the patient; Supplementary Fig. S4C). This was paralleled by a profound reduction in both the number and size of spontaneous lung metastases in vivo (Fig. 4E), adding additional confirmation that the CSC-like compartment within these tumors was particularly sensitive to TRAIL treatment.

In order to elucidate the underlying cause of the TRAIL hypersensitivity gained both in endocrine-resistant breast cancer cell lines and clinical tumors, key cellular components of TRAIL-mediated cytotoxicity were assessed by Western blot and fluorescent-activated cell sorting analysis in the TAMR and FASR cell lines. These included the TRAIL receptors DR4 and DR5 and the endogenous TRAIL-receptor complex inhibitor c-FLIP, all of which have been shown to influence TRAIL sensitivity in other cancer cell types (49–52). Comparative levels of DR4 and DR5 cell surface receptor expression did not reflect the relative sensitivity of each cell line to TRAIL (Supplementary Fig. S5A). The relative protein levels of the TRAIL pathway inhibitor, c-FLIP (55 kDa isoform identified with two independent antibodies), however, were found to be lower in TAMR cells compared with parental MCF-7 cells (Fig. 5A). Similarly, c-FLIP protein levels were also significantly reduced in TAMR and FASR derivatives of the T47D cell line (Fig. 5A). This reduction in c-FLIP protein levels is attributed to posttranscription regulation of c-FLIP, as no correlation between c-FLIP mRNA levels in either the cell lines or the primary samples tested correlated with TRAIL response (Supplementary Fig. S5B).

Previously, a Jnk-mediated degradation of c-FLIP via the E3 ubiquitin ligase ITCH has been described to explain regulation of c-FLIP protein in primary mouse cells (ref. 53; Supplementary Fig. S5C). Accordingly, phosphorylated ITCH (pITCH) was upregulated in MCF-7– and T47D-derived TAMR and FASR cells (Fig. 5A) compared with parental controls with no parallel increase in total ITCH levels. The proposed link between Jnk and suppressed c-FLIP levels was confirmed by pretreatment of the MCF-7–derived TAMR cells with the Jnk inhibitor SP600125, which after 4-hour exposure partially rescued TAMR cell and endocrine-resistant primary sample viability in the presence of TRAIL (Fig. 5B; Supplementary Fig. S5D; Supplementary Fig. S5E). Furthermore, Jnk inhibition decreased ITCH activation (reduced phospho-ITCH levels) and subsequently restored c-FLIP protein expression of TAMR cells to parental MCF-7 levels (Fig. 5C).

Similarly, untreated MaCa 3366-TAMR tumors exhibited a significantly lower level of c-FLIP protein expression and increased ITCH and pITCH expression compared with its endocrine-sensitive, estrogen-dependent control (Fig. 5D). Western blot analysis of 8 of the advanced metastatic endocrine-resistant clinical samples ex vivo with varying TRAIL response was performed in order to determine if the mechanism of sensitivity in these primary cells was consistent with that of the endocrine-resistant cell lines (Fig. 5E). The activation (phosphorylation) of ITCH and relative c-FLIP levels were subsequently determined by densitometry normalized to GAPDH controls. The activation of ITCH and c-FLIP protein levels were inversely correlated such that increased activation of ITCH in endocrine-resistant samples correlated with a reduction in c-FLIP levels, and, when averaged, cells sensitive to TRAIL had significantly lower levels of c-FLIP and higher levels of ITCH activation (Fig. 5F; Supplementary Fig. S5F). In addition, reduction of c-FLIP protein levels and increased ITCH activation correlated with response to TRAIL (Fig. 5F; Supplementary Fig. S3B). Furthermore, the advanced metastatic endocrine-resistant clinical sample with the lowest activation of ITCH (sample 81, Fig. 5E) and the sample with the highest level of c-FLIP protein (sample 29, Fig. 5E), both of which demonstrated no significant response to TRAIL under adherent conditions, could be substantially sensitized to TRAIL by siRNA inhibition of c-FLIP, supporting the role for c-FLIP in TRAIL sensitivity (Fig. 5G; Supplementary Fig. S5G and S5H). Finally, the endocrine-naïve, TRAIL-resistant, clinical sample 90 could also be sensitized to TRAIL by siRNA suppression of c-FLIP (Fig. 5H; Supplementary Fig. S5I). These data demonstrate that the activation of ITCH and reduction of c-FLIP is one potential mechanism of hypersensitivity of endocrine-resistant clinical samples ex vivo.

There is a clear unmet need for alternative treatments for patients who relapse on endocrine therapy. Relapse rates are high, and following multiple lines of endocrine therapy, there is higher risk of metastasis and poorer overall survival with limited treatment options.

Here, we demonstrate that acquired endocrine-resistant breast cancer cells gain sensitivity to TRAIL treatment compared with their endocrine-responsive parental counterparts. This appears to manifest following the acquisition of resistance and is a stable sensitivity, as administration of endocrine agents in ER-responsive cells does not increase response to TRAIL.

Although other changes in signaling pathways have been identified as a consequence of endocrine resistance—such as EGFR/HER2, MAPK, CDK4/6, mTOR, and Src (54–59)—the data reported here represent to our knowledge the first demonstration of a sensitization to a single agent targeting a proapoptotic pathway. Previous studies have demonstrated increased Jnk signaling following the long-term acquisition of resistance to endocrine therapies (12, 60). Here, we demonstrate that one consequence of this Jnk activation is the ITCH-dependent destabilization of c-FLIP protein, which correlates with TRAIL sensitivity both in cell line models in vitro and in clinical samples ex vivo.

Previously, modest outcomes in phase II clinical trials of TRAIL agonists in terms of overall survival have been partly attributed to the lack of potency of existing TRAIL receptor agonists (34); however, it could also be postulated that a lack of stratification and suitable biomarkers for response also contributes to the poor results seen to date. The data presented here suggest that acquired resistance to endocrine therapy could represent a key subgroup of patients that might exhibit improved responses to TRAIL agonists and that monitoring of c-FLIP, pITCH, and/or p-Jnk levels in tumors could further potentiate the selection process by identifying at least a proportion of TRAIL-responsive tumors. Further exploration of these molecular markers by immunohistochemical analysis of pathologic samples with access to fresh tissue material would be required to determine if these could be used as predictive biomarkers of TRAIL response.

We recently reported that TRAIL can selectively target the CSC-like cells in breast cancer cell lines in vitro following suppression of c-FLIP. Here, we demonstrate that acquired endocrine resistance potentiates sensitivity of the CSC-like compartment to TRAIL in primary-derived breast cancer samples, and that this significantly correlates with a reduction in c-FLIP levels. These observations are particularly significant given that CSCs have been demonstrated to evade conventional chemotherapy agents and potentially endocrine agents, either due to cell quiescence, through the prevention of drug uptake or modulation of intracellular signaling (18, 21, 61–63). Targeted therapies that initiate cell death through cell surface receptors circumvent these resistance mechanisms and therefore may prove to be a more effective strategy for targeting CSCs (64–66). Indeed, this was confirmed in xenograft models of endocrine-resistant tumors, whereby systemic TRAIL treatment resulted in a profound suppression of tumor recurrence (Figs. 2 and 4) and a marked decrease in spontaneous metastasis (Figs. 2F and 4E). Interestingly, response to TRAIL was observed regardless of cellular ER status. In vitro models of TAMR maintain ER positivity following acquisition of resistance, and active ER signaling remains important in cell proliferation. Similarly, MaCa 3366 also maintains ER positivity following acquisition of resistance; however, FASR cell lines and PDX 151 do not maintain their ER positivity. Despite these important differences in cellular signaling, all these models demonstrated sensitivity to TRAIL treatment in vitro and in vivo. Taken together, these data have important implications for potential long-term survival benefit patients who ultimately relapse on any form of endocrine therapy.

Previous studies have proposed that the acquisition of endocrine resistance is accompanied by the appearance of mesenchymal-like properties, which might also explain the increased risk of metastatic disease. This was the premise on which we initially tested for TRAIL sensitivity in endocrine-resistant breast cancer cells, as TRAIL had previously been shown to selectively target breast cancer cell types with a mesenchymal-like phenotype (35). The question still remains as to whether a partial EMT sufficiently explains the strong correlation between endocrine resistance and TRAIL sensitivity in clinical samples observed here. Further studies into the mechanism of Jnk activation and the role of EMT-like processes in TRAIL sensitivity will help elucidate its relationship with acquired endocrine resistance.

In summary, we have shown that acquired endocrine resistance confers sensitivity of breast cancer cells and particularly breast CSCs to TRAIL-mediated apoptosis. These data support further investigation into the selective use of TRAIL agonists as a subsequent treatment for ER+ve patients who relapse on endocrine therapy regardless of their ER status following relapse.

P. Barrett-Lee reports receiving remuneration from Roche Products Ltd. J. Gee reports receiving commercial research grants from AstraZeneca. No potential conflicts of interest were disclosed by the other authors.

Conception and design: L. Piggott, T. Robinson, C. Morris, R. Clarkson

Development of methodology: L. Piggott, I. Fichtner, L. Andera, C. Morris, P. Barrett-Lee, R. Clarkson

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Piggott, A. Silva, T. Robinson, A. Santiago-Gómez, B.M. Simões, M. Becker, P. Young, C. Morris, P. Barrett-Lee, F. Alchami, M. Piva, M.dM. Vivanco, R.B. Clarke, J. Gee

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Piggott, A. Silva, T. Robinson, A. Santiago-Gómez, B.M. Simões, M. Becker, M. Piva, M.dM. Vivanco, R. Clarkson

Writing, review, and/or revision of the manuscript: L. Piggott, A. Silva, T. Robinson, A. Santiago-Gómez, B.M. Simões, L. Andera, C. Morris, P. Barrett-Lee, F. Alchami, R.B. Clarke, J. Gee, R. Clarkson

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Piggott, A. Silva, I. Fichtner, C. Morris, F. Alchami

Study supervision: L. Piggott, P. Barrett-Lee, R. Clarkson

The authors acknowledge the patients and their families for their participation in this study. We acknowledge the infrastructure support from Wales Cancer Bank (www.walescancerbank.com) for the collection of patient samples and Professor Sacha Howell (The Christie NHS Foundation Trust) for access to pleural effusion patient samples.

This study was funded by Breast Cancer Now (charity no. 1160558). A. Silva was supported by Tenovus Cancer Care funding (charity no. 1054015).

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.