Introduction
In approximately 75% of postmenopausal patients, breast cancer is a hormone-dependent disease that relies on the mitogenic effects of estrogen to drive carcinogenesis. Endocrine therapies, including estrogen receptor α (ERα) modulators and aromatase inhibitors (AIs), are the most suitable treatment for ERα-positive (ER+) breast cancer patients. Recently, nonsteroidal AIs (for example, letrozole, anastrozole) that block the biosynthesis of estrogens have proven more effective than the selective estrogen receptor modulator tamoxifen (Tam) in the treatment of postmenopausal patients with ER+ breast cancer [
1]. Despite the demonstrated clinical efficacy of AIs, however,
de novo and acquired resistance still occurs and constitutes a major impediment to successful therapy.
At present, acquired resistance to endocrine therapy is considered to be a progressive, stepwise phenomenon whereby breast cancer cells are converted from an estrogen-dependent phenotype, which is responsive to endocrine therapy, to a nonresponsive phenotype and eventually to an estrogen-independent phenotype. Among the molecular mechanisms involved in the acquisition of endocrine resistance, a switch from steroid signaling to growth factor signaling pathways has been the focus of recent studies, which have demonstrated the activation of the phosphatidylinositol 3-kinase (PI3K)/AKT/mammalian target of rapamycin (mTOR) and/or mitogen-activated protein kinase (MAPK) pathways, both in breast cancer cell lines and in breast tumors [
2-
8]. Activation of these survival pathways may contribute to endocrine resistance via the activation of kinases in an ER-dependent [
9] as well as ER-independent fashion [
2,
10].
MicroRNAs (miRNAs) are short, noncoding RNAs that generally base pair within the 3′ untranslated (3′UTR) region of target mRNAs, causing translational inhibition and/or mRNA degradation. A growing body of evidence favors miRNAs’ being important players in oncogenesis [
11], with some able to act as oncogenes, others as tumor suppressors and others displaying either oncogenic or tumor-suppressive activities, depending on the tissue and tumor context [
11-
13]. Widespread deregulated expression of miRNAs would thus be expected to represent another hallmark of cancer, providing not only biomarkers but also novel therapeutic targets. Recent investigations have revealed that miRNAs are involved in the development of drug resistance, but little is known about the miRNA-driven molecular mechanisms governing the drug-resistant signal transduction network [
14,
15].
Increasing amounts of data support an involvement of miRNAs in estrogen action and/or in endocrine resistance, with most studies dedicated to Tam or fulvestrant resistance. A close cross-talk appears to exist between ERα and specific miRNAs, with several miRNAs found to regulate ERα, which conversely negatively regulates the expression of some miRNAs [
16]. Complexity was further revealed by the observation that miRNAs are also able to regulate ERα activity by repressing the expression of ERα transcriptional cofactors [
16]. Furthermore, some miRNA signatures have been identified by microarray analysis in Tam- or fulvestrant-resistant breast cancer cell lines [
17,
18]. Several miRNAs, such as the miR-200 family [
19], miR-375 [
20], miR-221/222 [
21,
22], miR-15a/16 [
23], miR-101 and miR-519a [
24,
25], have been shown to regulate molecular targets or functional signaling pathways associated with Tam or fulvestrant resistance. To date, only one study research team has investigated miRNA expression profiles associated with AI resistance [
26]. They found that miR-128a, previously identified as being associated with breast cancer aggressiveness [
27], was highly expressed in letrozole-resistant breast cancer cells compared with their sensitive counterpart and targeted the transforming growth factor β (TGF-β) signaling pathway.
Bearing in mind that any one given miRNA can have several targets, some belonging to the same functional network or signaling pathway, and that the 3′UTR of a single gene is frequently targeted by several different miRNAs, our primary aim in this study was to capture a global view of the miRNA expression profiles associated with AI resistance in the hope of identifying common miRNA-targeted specific functional networks. Our second aim was to select the most relevant miRNAs that represent candidate biomarkers or putative therapeutic targets of ER+ breast cancers treated by AIs. In this study, we performed a large-scale investigation of miRNA expression profiles associated with letrozole or anastrozole resistance using two in vitro models of AI-acquired resistance. We report the acquisition of several deregulated miRNAs as a newly discovered alternative mechanism developed by the AI-resistant breast cancer cells to achieve constitutive activation of the AKT/mTOR pathway and to develop AI resistance. We also demonstrate, for the first time to our knowledge, that elevated miR-125b expression levels constitute a novel marker for poor prognosis in breast cancer and that targeting miR-125b in letrozole-resistant cells overcame letrozole resistance.
Discussion
Although AI endocrine therapy provides obvious clinical benefits, the molecular mechanisms involved in resistance to AIs remain poorly described. To gain further insight into the molecular mechanisms underlying the AI resistance, recent whole-genome analyses using AI pretreatment tumor biopsies accrued from patients in AI neoadjuvant studies to delineate the mutational landscape associated with AI response strategies are of great interest [
48]. However, for preclinical studies, few cellular models mimicking resistance to AIs are currently available. Because breast cancer cell lines usually display very low or no expression of endogenous aromatase [
28], the most relevant cells in which to study the response to AIs are cells transfected with the human aromatase gene. Among the cellular models proposed to study AI resistance, aromatase-transfected and long-term estrogen-deprived (LTED) cell lines have been proposed, based on the hypothesis that lack of hormone in the environment could mimic the withdrawal of estrogen that occurs during treatment with AIs [
39,
49]. Most preclinical studies investigating AI resistance have been conducted on LTED models, but it has previously been shown that LTED cells are not totally equivalent to models of endocrine therapy-acquired resistance [
26,
39,
49,
50]. Indeed, genome-wide analysis revealed important gene expression profile differences between the AI-resistant and LTED cell lines and also among the AI-resistant cell lines themselves [
39]. Supporting data came from the observation that activated signaling pathways observed in LTED cells were different from those observed in long-term AI-treated cells [
49,
50]. More importantly, the only study in which miRNA expression profiles associated with AI resistance were investigated clearly demonstrated that specific miRNA profiles could be inversely regulated between AI-resistant cells and LTED cells [
26]. In particular, miR-128a, which was found to regulate TGF-β signaling in letrozole-resistant breast cancer cells, was upregulated in letrozole-resistant cells, but not in LTED cells [
26]. Thus, cellular models established according to a protocol that mimics clinical treatment [
2,
6,
39,
50,
51] (that is, long-term exposure to AIs) may more closely reflect the clinical situation and may be pertinent to investigation of acquired AI resistance. Few of such
in vitro models are currently available [
2,
5,
6,
39,
41].
Deciphering miRNA deregulations as new mechanisms associated with acquired AI endocrine resistance remains poorly investigated, and to date only one previous study team, using an “omics miRNA” approach, identified one particular miRNA associated with letrozole resistance [
26]. In the present study, we aimed to capture a global view of the miRNA expression profiles associated with letrozole and anastrozole resistance in the hope of identifying common miRNA-targeted specific functional networks and relevant miRNAs (candidate biomarkers or putative therapeutic targets of ER+ breast cancers treated by AIs). We performed a large-scale miRNA analysis using two
in vitro cellular models of acquired AI resistance (Res-Let and Res-Ana cells). Importantly, these cells are total populations of AI-resistant cells and may thus also mimic the heterogeneous phenotype and behavior of resistant cells possibly present in the tumors of patients who relapse under AI endocrine therapy.
The AKT/mTOR pathway is known to play a pivotal role in AI resistance [
2,
5,
6,
39,
41]. Among the molecular mechanisms leading to activation of the AKT/mTOR pathway in AI-resistant cells, special attention has previously been given to the upstream ErbB family receptors [
52]. Importantly, in a recent study, researchers found that in breast cancer cells with acquired letrozole resistance, the constitutive activation of the AKT/mTOR pathway could be blocked by PI3K and/or mTOR inhibitors, but not by EGFR/ErbB2 inhibitors [
2]. This observation strongly supports the existence of a mechanism other than activation of receptors upstream of the AKT pathway. Our data show, for the first time to our knowledge, that acquisition of deregulated miRNA expression represents an alternative trick used by ER+ AI-resistant breast cancer cells to activate this crucial survival pathway and thus escape from AI endocrine therapy.
A growing body of evidence developed over the last several years supports the existence, within a given tumor, of separate cell subpopulations that have acquired different mechanisms, all converging to activate the same functional survival network [
53]. Consistent with this idea, the Res-Let or Res-Ana cells develop a deregulated network of several miRNAs, each independently capable of leading (or predicted to lead) directly or indirectly to the activation of the AKT/mTOR pathway. Among the miRNAs identified in our study, deregulated expression of miR-23b [
54], miR-494 [
55], miR-21 [
56-
58], miR-301 [
59] or miR-193a [
60] has previously been shown to induce the activation of the AKT pathway in different cancer contexts. Among the miRNAs predicted to target the AKT/mTOR pathway and commonly deregulated in both Res-Let and Res-Ana cells, we then investigated miR-125b, miR-205 and miR-424 further. Our study has shown, for the first time to our knowledge, that ectopic overexpression of either miR-125b or miR-205, or silencing miR-424 expression, in the sensitive ER+ MCF-7aro cells was sufficient to trigger activation of the AKT/mTOR pathway, to develop
de novo resistance to both letrozole and anastrozole drugs and to induce the selection of stem-like and tumor-initiating cells possessing self-renewing properties.
Several previous studies have demonstrated that any given miRNA could play a dual role (via oncogenic or tumor-suppressive activities), depending on the tissue or the tumor. This is likely due to any given miRNA targeting multiple mRNAs, each of which has a different function in an individual cellular context [
11-
13,
61]. Consistently, miR-424 is known to play a suppressive role in some tumors [
62], whereas high expression levels have been ascribed to chemotherapy resistance [
63,
64]. In accordance with our results, decreased levels of miR-424 led to an activated AKT/mTOR pathway in prostate and colon cancer cells [
65]. MiR-205 also plays a dual role in carcinogenesis [
13]. In its “devil” role (associated with tumorigenesis), and in accordance with our study, overexpression of miR-205 has been shown to lead to a coordinated activation of the AKT signaling pathway in several cancer contexts [
66-
69]; to our knowledge, though, it has never been shown in breast cancer cells. Complexity is further revealed by the observation that, conversely, miR-205 is able to inhibit activation of the AKT pathway [
70]. However, it is noteworthy that this latter study was conducted in ERα-negative breast cancer cells, and one can suggest that the ER status of breast cancer cells provides a cellular context whereby miR-205 exerts opposite roles.
MiR-125b interferes with many different cellular processes and, according to the cellular context, acts as a tumor suppressor (for example, gliomas, ovarian cancer, and hepatocellular carcinoma), as an oncogene (for example, non-small cell lung carcinoma, colon cancer) or as both (brain tumors or prostate cancer) [
12,
61]. In its tumor-suppressive functions, miR-125b has been reported to target
ENPEP,
CK2-α,
CCNJ,
MEGF9 [
71] or the proto-oncogene
ETS1 [
72]. Conversely, in its “devil” role, high miR-125b expression levels confer to breast cancer cells resistance to paclitaxel by targeting
BAK1 [
73] and are detected in miRNA signatures present in Tam- or fulvestrant-resistant breast cancer cell lines [
17,
18]. In a neoadjuvant chemotherapy study in breast cancer, high expression levels of circulating miR-125b were detected in nonresponsive patients [
74]. One of the key findings of our study is that miR-125b overexpression not only enables activation of the AKT/mTOR pathway and AI resistance but also confers an estrogen-independent phenotype to the sensitive ER+ MCF-7aro cells. Importantly, miR-125b is known to directly target the p53 tumor suppressor gene and other genes belonging to the p53 network [
12,
61], and p53 is known to closely interfere and communicate with the cancer relevant PI3K/AKT/mTOR pathway [
75-
78]. Indeed, p53 is able to regulate cell survival by inhibiting the PI3K/AKT/mTOR survival pathway, such as by positively regulating the expression of AKT/mTOR pathway inhibitors [
76,
78] or by negative regulating the expression of
PI3KCA, encoding the p110α catalytic subunit of PI3K [
75,
77]. Thus, inactivation of p53 and subsequent upregulation of
PI3KCA is one of the mechanisms contributing to the pathophysiology of cancer [
75,
77]. Interestingly, transfecting MCF-7aro cells with the mimic of miR-125b was sufficient to lead to decreased p53 expression levels (at both the mRNA and protein levels) and to subsequent increased p110α expression levels (Additional file
9: Figure S5). These data strongly support that p53-driven mechanisms could be involved in the activation of the AKT pathway in the presence of miR-125b. This could thus represent one possible miR-125b-driven mechanism developed by the AI-resistant cells.
Our work also highlights, for the first time to our knowledge, the clinical relevance of miR-125b in breast cancer by demonstrating that high expression levels of this miRNA may represent a marker of poor prognosis. Indeed, assessing miR-125b expression levels (alone or associated with lymph node status) may allow the restratification of patients with breast cancer into outcome-dependent subclasses. Most importantly, high miR-125b expression levels were also associated with earlier relapse in ER+ breast cancer treated with endocrine therapy. The measurement of miRNA expression holds great promise, as miRNA can now be extracted not only from frozen tumor samples but also from formalin-fixed, paraffin-embedded samples as well as from serum [
11]. Expression levels of miR-125b could thus represent a new prognostic marker and candidate predictor of AI response in breast cancer. We aimed to investigate the prognostic value of miR-125b using the miRNA-seq data of The Cancer Genome Atlas (TCGA) breast cancer cohort [
79]. However, the insufficient clinical follow-up (low number of relapse events and short median overall follow-up) actually available in this breast cancer cohort does not allow us to make any statements about the prognostic value of miR-125b (data not shown). When a longer median follow-up for the patients in the TCGA cohort who have not yet relapsed is available to the scientific community, it would be of great interest to assess whether high miR-125b expression levels are also found to be associated with shorter RFS in this cohort.
Importantly, reducing miR-125b expression levels in the letrozole-resistant cells was sufficient not only to block the constitutive activation of the AKT/mTOR pathway but also to overcome letrozole resistance, by resensitizing the resistant cells to AI treatment. In the future, therapeutic strategies aimed at blocking expression of miR-125b in endocrine-sensitive primary breast cancer may reinstate a greater response to AI endocrine therapy in a subset of breast cancers and/or may prevent the emergence of miR-125b-overexpressing breast cancer cell subpopulations evolving toward AI-resistant cells.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
PV, CFD, MV and EG performed most of the biological experiments. LC participated in the design of the study. TB contributed to tumor sample and clinical data collection. JAV performed the biological experiments and statistical analysis and participated in the design of the study. PAC conceived of the study and its design and coordination and wrote the manuscript. All authors critically read the manuscript and approved the final version.