Background
Breast cancer is the most common cancer in women worldwide. Two-thirds of breast tumors express ER that drives proliferation of mammary epithelial cells and thereby contributes to the etiology and progression of the disease. Consequently, antagonists that directly block ER function or drugs that lower the amounts of the natural ligand of ER, estradiol, have been utilized in breast cancer treatment for decades [
1]. For over 40 years, tamoxifen, a selective ER antagonist, has been the backbone in treating ER-positive breast cancers. Despite of being effective in decreasing mortality,
de novo or acquired resistance frequently occurs [
2]. Some of the mechanisms leading to resistance have been revealed, including mutations in the gene encoding ER [
3‐
5], altered expression patterns of ER or its cofactors [
6,
7], and crosstalk between ER and growth factor receptor cascades such as the EGFR/ERK1/2-pathway [
8]. Consequently, inhibition of ERK1/2 has been reported to restore antiestrogen sensitivity. For example, a study with the MEK inhibitor PD098059, a compound that reduces the phosphorylation and activation of ERK1/2, was shown to inhibit the growth of tamoxifen-resistant cell lines and to restore their sensitivity to therapy [
9,
10]. However, ERK1/2 inhibition has proven efficacy primarily against cells with resistance-provoked overexpression or activation of HER2 [
9]. On the other hand, recent findings suggest that proteasome inhibition might offer a new avenue for overcoming endocrine resistance [
11,
12]. Bortezomib, a proteasome inhibitor, has been investigated as a combination therapy in conjunction with endocrine treatment in a phase II study [
13].
Whilst shRNA- or cDNA-based functional screens [
14,
15] and candidate gene [
16‐
19], or drug [
9,
20‐
23] approaches have been used to study the development and reversal of endocrine resistance, the exact molecular mechanisms remain unknown, and large-scale studies on cells treated long-term with tamoxifen are lacking. Moreover, efforts to find new treatment regimes for overcoming drug resistance have been largely based on a few selected drug candidates, and have only proven to be effective in a fraction of the cases [
1]. Development of primary drug resistance can make the cancer cells susceptible for novel vulnerabilities, hence leading to additional therapeutic opportunities. However, secondary resistances towards other drugs may also arise. Resistance to chemotherapeutics has been linked with estrogen receptor positive breast cancer [
24], but systematic studies on tamoxifen resistance associated co-resistances have not been conducted. Therefore, systematic, large-scale studies to characterize the drug sensitivity profiles of tamoxifen-resistant breast cancer are warranted to reveal new drug vulnerabilities as well as co-resistance patterns in drug-resistant cells.
Here, we report the development and characterization of a panel of seven long-term tamoxifen-treated breast cancer cell lines from four parental strains. Using these resistant cell line models and their isogenic parental counterparts, we, for the first time, performed systematic high throughput drug sensitivity and resistance testing with 279 approved and investigational oncology drugs to reveal potential new drug vulnerabilities and to identify co-resistance patterns acquired with tamoxifen resistance. We further conducted exome-sequencing on each of the isogenic parental-resistant cell line pair to identify point mutations and copy number variations that may contribute to drug resistance. Through integrated network analyses, we uncovered cell- and clone-specific molecular and functional patterns of endocrine resistance, highlighting the underlying molecular diversity, and pointing to several distinct therapeutic opportunities to circumvent it. However, no systematic drug screens with hundreds of oncology compounds on acquired tamoxifen resistance have been conducted.
Discussion
In the present study, we report the development and systematic characterization of seven long-term tamoxifen-treated cell lines, and by pharmacogenomic profiling, determine the drug response profiles and mutational landscapes of these drug-resistant models. Different
in vitro and
in vivo models of endocrine-resistance have been developed to explore common mechanisms behind resistance development [
9‐
11,
19,
20,
22,
40‐
50] However, to our knowledge, this is the first comprehensive drug testing study with hundreds of oncology compounds across a panel of several tamoxifen-resistant models. Using this approach, we identify clone-specific molecular networks reflecting the diversity of pathways leading to endocrine resistance. It is noteworthy that as the availability of clinical data sets on diagnosed acquired tamoxifen resistance with response/survival data are essentially non-existent to date, the resistant/sensitive cell line models and associated data sets presented here form a valuable research resource.
Concurrently with developing tamoxifen resistance, novel drug vulnerabilities emerge. Here, we identified common, cell type-, and cell clone-specific sensitivities. The most important of these are listed in Table
1. Additionally, several of the sensitizing drugs are in clinical trials for treatment of advanced or metastatic breast cancer. However, possible correlation between patient enrollment criteria, observed molecular mechanisms and the sensitivities and co-resistances identified here remains to be investigated.
All resistant cell lines except one (MCF-7 Tam1) exhibited gained sensitivity towards the ERK1/2 inhibitor VX-11E. ERK1/2 inhibition prevents its autophosphorylation [
10,
51] and results in reduced phosphorylation and thereby also decreased activation of ER [
9]. Overactivity of ERK1/2 has been shown to associate with loss of ER, and decreased levels of ER are also seen in the majority of our tamoxifen-resistant cell lines (Additional file
4) [
52]. However, the basal levels of unphosphorylated or phosphorylated EGFR/ERK are not elevated in the tamoxifen resistant lines; rather the opposite, i.e. decrease in basal levels as well as dephosphorylation of ERK1/2 and EGFR is observed (Fig.
7c and data not shown). We therefore anticipate that increased phosphorylation of these signaling proteins does not explain the observed sensitivity towards VX-11E. Interestingly, upon increasing concentrations of VX-11E, slight increase in phosphorylated ERK1/2 is observed in the T-47D Tam clones. However, upon additional inhibition of MEK with selumetinib, this effect is diminished. The effect of VX-11E inducing ERK1/2 phosphorylation has also previously been reported [
53] and might therefore reflect a general mode of action especially as the same effect is observed also in our parental cells (Fig.
7c). It could therefore be speculated that already a short-term tamoxifen treatment causes an effect on the levels of phosphorylated ERK1/2 and that long-term exposure leads to, at least partial, down-regulation of EGFR and ERK1/2. As cells are challenged with increasing concentrations of an ERK1/2 inhibitor (VX-11E), an increase in ERK1/2 phosphorylation is seen, with concomitant cell killing of the tamoxifen-resistant cells observed in the drug screen. However, further studies to elucidate the exact mechanisms are needed.
We also identified bortezomib, a proteasome inhibitor, as a sensitizing agent (Figs.
5,
6 and
7 and Additional file
12, Table
1). A direct role for bortezomib in reversing tamoxifen resistance has not been demonstrated before, although a link between proteasome function and estrogen receptor -mediated transcription has been suggested [
54], and bortezomib has recently been shown to enhance endocrine treatment in cell line models as well as in humans [
11‐
13].
In addition to shared sensitivity to VX-11E and bortezomib in the tamoxifen-resistant cells, we also identified cell line specific drug sensitivities (Table
1, Figs.
3,
4,
5 and
6, Additional files
11 and
12). T-47D Tam1 and Tam2 cells displayed sensitivity towards the SRC-family inhibitor KX2-391 and the dual ABL/SRC-inhibitors dasatinib and ponatinib (Figs.
3 and
4). This is in agreement with recent findings [
20,
55]. Another SRC-inhibitor, SU6656, has also been reported to inhibit growth of tamoxifen-resistant cells [
42], highlighting the potential of SRC-inhibition in overcoming endocrine resistance. Interestingly, dasatinib has been shown to overcome tamoxifen resistance in MCF-7/fibroblast co-culture, and it is currently undergoing clinical trials on metastasized ER-positive breast cancer.
Upon acquiring resistance to tamoxifen, the T-47D Tam1 cells also gained sensitivity towards the BCL-family inhibitors navitoclax and obatoclax (Fig.
3). BCL-2 family proteins BCL-2, BCL2L1, BCL2L10 and MCL1, represented in the network, are major negative regulators of apoptosis, and thus, upregulation of their expression might offer the tamoxifen-challenged cells means to overcome resistance, as well as downregulation of BAD, a proapoptotic regulator. Indeed, BCL-2 has been indicated in tamoxifen resistance, and consequently, a BCL-2 inhibitor, ABT-737, has been reported to restore sensitivity [
56]. Additionally, tamoxifen treated patients with low level of BAD expression had a worse prognosis [
57]. The T-47D Tam1 cells also displayed sensitivity towards RAF-inhibitors BAY 73–4506 and RAF265. This is in line with previous findings on overexpression of
RAF1 promoting tamoxifen-resistant growth [
58]. Both navitoclax and BAY 73–4506 are being investigated for treatment of different cancers, navitoclax for lung cancer and lymphoma, and BAY 73–4506 for metastatic colorectal cancer among others. Our results, and those from others [
56‐
58] suggest that BCL- and RAF-inhibitors might offer means to treat also endocrine-resistant breast cancer.
We also identified sensitizers with preference for the luminal B-derived tamoxifen-resistant cells, BT-474 Tam1 and Tam2. These included the cdk-inhibitors SNS-032 and alvocidib, along with the EGFR-inhibitor gefinitib and the Btk-inhibitor PCI-32765, and several HER2/EGFR-inhibitors (Figs.
5 and
6). Crosstalk between ER and ERBB2/EGFR pathways has been shown to be activated in tamoxifen resistance [
59]. Recently, the EGFR/HER2 dual inhibitor AZD8931 was also suggested to inhibit growth of MCF-7 or T-47D tamoxifen-resistant cells in xenograft models [
60]. It is noteworthy that in our study, the parental BT-474 cells, unlike all others presented here, initially amplify and overexpress HER2, and display some sensitivity towards HER2/EGFR-inhibitors (Additional file
2). Interestingly, as tamoxifen resistance develops, the cells become more sensitive to several of the HER2/EGFR-inhibitors and indeed, the combined use of growth factor receptor kinase inhibitors in conjunction with tamoxifen has been suggested to circumvent endocrine resistance [
45], and combination therapy with antihormone and gefinitib has demonstrated resensitization to tamoxifen in xenografts [
61,
62]. However, our results on decreasing EGFR / phospho-EGFR levels upon acquired resistance (Fig.
7c) indicate that mechanisms other than direct upregulation of the EGFR pathway are responsible for the observed gained sensitivity.
Development of primary drug resistance in cancer treatment frequently results in the emergence of secondary resistances. Here, we discovered that upon acquiring tamoxifen resistance, all of the resistant cells acquired co-resistance towards at least one chemotherapeutic agent, such as paclitaxel, docetaxel, vincristine, vinblastine or topotecan (Figs.
2,
3,
4,
5,
6 and
7, Table
1 and Additional files
5,
11 and
12). Even though chemoresistance has been associated with the estrogen receptor [
24], the co-resistance observed here may rather reflect the slowed-down growth of many of the resistance clones, and may therefore propose a uniform mechanism for paclitaxel resistance (Additional file
3). However, selective co-resistance still occurs, as the cells do not become universally co-resistant against all chemotherapeutics. Nevertheless, general down-regulation of cellular functions is especially evident with the tamoxifen-resistant MCF-7 Tam1 cells, which not only possess an overwhelmingly co-resistant drug response profile, but also down-regulate cell signaling (Additional files
3,
4,
11). Indeed, already a short-term tamoxifen-treatment of MCF-7 cells triggers a predominant down-regulation of gene expression [
63], suggesting that depending on the molecular background, some tamoxifen-resistant cells might exhibit an intrinsically more unresponsive profile.
Interestingly, every single tamoxifen-resistant cell line was also more resistant to the survivin (BIRC5-) inhibitor YM155 than their parental counterparts (Figs.
2,
3,
4,
5,
6 and
7, Table
1 and Additional files
5,
11 and
12), suggesting a role for survivin in development of tamoxifen resistance. Survivin has recently been associated with resistance to chemo- or hormonal therapy, and has been identified to predict poor clinical outcome via ERBB2-mediated overexpression [
47]. Furthermore, siRNA-knock down of
BIRC5 has been shown to enhance cell sensitivity to tamoxifen [
64]. Alternatively, it has been speculated that uptake of YM155 is dependent on cell membrane a solute carriers, encoded by the
SLC35F2 gene [
65]. Upon resistance development, expression of the solute carriers possibly decreases and consequently, less YM155 enters the cells making them resistant to the drug.
As initiation and development of acquired tamoxifen resistance are largely thought to be driven by genetic adaptations [
3‐
5,
7] we inspected the genetic landscape of the drug-resistant cells by exome-sequencing and correlated the findings with our drug profiling data. Tamoxifen-resistant cells accumulated point mutations and copy number changes throughout their genomes, with only some of the changes being common between two resistant clones originating from the same parental cells, implying clonal divergence (Additional files
8,
9 and
10). Whilst many of the genetic aberrations that have been associated with endocrine resistance previously were also recapitulated here, our data as a whole demonstrate that new sensitivities may develop largely independent of the genetic changes, and in fact, antiestrogen resistance can be seen even in the absence of any evident mutations [
66]. Analogous phenomenon has been noted in leukemic cells [
67]. Accumulation of numerous genomic aberrations can trigger resistance development [
20], but mutations can also be carried along as passengers as a result of selection pressure, rather than emerge as true evolutionary drivers [
68]. The data presented here demonstrate that in the majority of cases, no single genetic alteration can be identified as responsible for the drug response, but on the contrary, multiple target genes of the drugs converge into the same response networks, and many of these target genes also harbor genetic changes. Therefore, resistance development is likely to involve complex interactions comprising genetic as well as transcriptional and epigenetic mechanisms, or other adaptive changes in cell signaling.