The dysregulation of miRNAs has been increasingly recognized as a critical mediator of cancer development and progression. Here, we show that frequent deletion of the MIR135A1 locus is associated with poor prognosis in primary breast cancer. Forced expression of miR-135a decreased breast cancer progression, while inhibition of miR-135a with a specific miRNA sponge elicited opposing effects, suggestive of a tumor suppressive role of miR-135a in breast cancer. Estrogen receptor alpha (ERα) bound the promoter of MIR135A1 for its transcriptional activation, whereas tamoxifen treatment inhibited expression of miR-135a in ERα+ breast cancer cells. miR-135a directly targeted ESR1, ESRRA, and NCOA1, forming a negative feedback loop to inhibit ERα signaling. This regulatory feedback between miR-135a and ERα demonstrated that miR-135a regulated the response to tamoxifen. The tamoxifen-mediated decrease in miR-135a expression increased the expression of miR-135a targets to reduce tamoxifen sensitivity. Consistently, miR-135a expression was downregulated in ERα+ breast cancer cells with acquired tamoxifen resistance, while forced expression of miR-135a partially resensitized these cells to tamoxifen. Tamoxifen resistance mediated by the loss of miR-135a was shown to be partially dependent on the activation of the ERK1/2 and AKT pathways by miR-135a–targeted genes. Taken together, these results indicate that deletion of the MIR135A1 locus and decreased miR-135a expression promote ERα+ breast cancer progression and tamoxifen resistance.
Significance: Loss of miR-135a in breast cancer disrupts an estrogen receptor-induced negative feedback loop, perpetuating disease progression and resistance to therapy.
Graphical Abstract: http://cancerres.aacrjournals.org/content/canres/78/17/4915/F1.large.jpg. Cancer Res; 78(17); 4915–28. ©2018 AACR.
miRNAs are endogenous short noncoding RNAs that repress gene expression posttranscriptionally, primarily via binding to 3′ untranslated regions (3′UTR) of cognate mRNA targets (1). Deregulated miRNA expression plays fundamental roles in breast cancer progression and therapeutic responses (2). Interestingly, a large proportion of miRNA genes are mapped to the genomic instability regions and exhibit aberrant DNA copy numbers in cancer (3–5). The human chromosome 3p has been reported to exhibit frequent loss of heterozygosity in cancers including breast cancer (3, 5–8).
MIR135A1, located at chromosome 3p21.1, has been observed to be frequently lost in breast cancer (3). Notably, miR-135a was reported to be inversely correlated with the proliferation signature in breast tumors (9). In addition, although approximately 70% of human primary breast cancers express ERα and derive benefit from antiendocrine therapies, intrinsic and acquired resistance is common (10). Downregulation of miR-135a has been observed in tamoxifen-resistant breast cancer (11, 12) but its functional involvement remains unexplored.
In this study, we systematically investigated the functional roles of estrogen-regulated miR-135a in ERα+ breast cancer. We demonstrated that downregulated miR-135a expression, through the loss of repression of miR-135a target genes, promotes aggressiveness, estrogen-independent growth and acquired tamoxifen resistance of ERα+ breast cancer cells. Thus, patients with ERα+ breast cancer with MIR135A1 locus deletion and/or low miR-135a expression may represent a subpopulation that is intrinsically resistant to endocrine therapies.
Materials and Methods
Cell culture and breast cancer specimens
If not specified otherwise, all cell lines used in this study were from and cultured as recommended by ATCC. Cells were cryopreserved soon upon receipt and continuously cultured for less than 2 months. MCF-7, T47D, and MDA-MB-231 cells have been authenticated by STR genotyping as described (13, 14). Mycoplasma contamination was monitored routinely using mycoplasma detection set (M&C Gene Technology). Estrogen, tamoxifen, fulvestrant (ICI 182, 780), U0126, and propidium iodide (PI) were purchased from Sigma-Aldrich. AKT inhibitor IV was from Calbiocam. The hormone-deprived fetal bovine serum was prepared using dextran-coated charcoal (Sigma-Aldrich). For estrogen stimulation and tamoxifen treatment, cells were cultured in phenol red-free RPMI-1640 (Gibco) with 5% hormone-deprived serum for 6 days prior to the experiments. The specimens used in this study were collected from the First Affiliated Hospital of Anhui Medical University (Hefei, Anhui, China) with informed consent. The clinical research protocol was approved by the Biomedical Ethics Committee of Anhui Medical University.
miRNA, plasmids, transfection, retrovirus production, and transduction
The miR-135a mimics and siRNAs were from GenePharma. The shRNA plasmids were obtained from The RNAi Consortium (MISSION TRC shRNA library, Sigma). The DNA fragment about 200bps upstream and downstream of the miR-135a-1 stem loop was cloned into pBabe-puro vector to generate miR-135a–expressing plasmid. For miR-135a sponge plasmid, 4 tandem repeats of bulged miR-135a binding sequence were synthesized and subcloned into pBabe-puro vector. For 3′UTR luciferase reporter plasmids, the 3′ UTR of each target was amplified and subcloned into pSiCheck2 vector (Promega). To construct MIR135A1 promoter luciferase reporters, two different lengths of DNA fragments upstream of MIR135A1 gene were each cloned into pGL3-Basic plasmid (Promega). The mutant constructs were generated using QuikChange II XL Site-Directed Mutagenesis Kit (Stratagene). The sequences of cloning primers, miRNA mimics, siRNAs, shRNAs, and miR-135a Sponge are listed in Supplementary Table S1. Lipofectamine 2000 (Invitrogen) was used for all transfections. The production of retrovirus and subsequent generation of stable cells were described previously (15).
qRT-PCR, Western blot, immunofluorescence, and histologic analysis
Total RNA, miRNA, mRNA, and proteins were extracted and analyzed as described (15). The primers and antibodies used are listed in Supplementary Tables S1 and S2, respectively. Immunofluorescence and histologic analysis with hematoxylin and eosin, in situ hybridization, and IHC staining were carried out as previously described (15, 16).
Cell function assays and flow cytometry analysis
All cell function assays, including cell viability, three-dimensional Matrigel culture, soft-agar colony formation, foci formation, wound-healing, and Transwell migration, and invasion assays were carried out as previously described (13, 15, 16). For cell-cycle analysis, cells were collected, fixed with 70% (v/v) ice-cold ethanol at −20°C for 30 minutes, rinsed twice with phosphate-buffered saline (PBS), and then stained with 50 μg/mL PI in PBS containing 100 μg/mL RNase A, and 1% Triton X-100 for 30 minutes at room temperature. The early apoptotic cells were examined using Annexin V–FITC/PI kit (BestBio Biotechnologies). The flow cytometry was performed on BD FACSVerse Flow Cytometer.
Luciferase reporter assays, chromatin immunoprecipitation, and biotinylated miRNA pulldown assay
Luciferase reporter assays and chromatin immunoprecipitation (ChIP) were performed as described (15). DNA enrichment was assessed by qPCR, or regular PCR using PrimeStar HS DNA Polymerase. For RNA pulldown assay, miR-135a targets were pulled down with 3′ biotin-labeled miR-135a mimics and assessed by qRT-PCR with primers around miR-135a-binding sites as described (13). The primers and antibodies used are listed in Supplementary Tables S1 and S2, respectively.
The design and protocol of animal experiments were approved by the Institutional Animal Care and Use Committee, University of Science and Technology of China (USTCACUC1301019). Briefly, 1 × 106 (or 2 × 106 in tamoxifen treatment model) MCF-7 cells mixed 1:1 with Matrigel (BD Biosciences) were injected into the mammary fat pads of mice in the orthotopic model and 1 × 106 MDA-MB-231 cells in PBS were injected into the tail vein of 6-week-old female BALB/c nude mice (Slaccas Co.) in the metastasis model. In orthotopic models with MCF-7 cell–derived tumors, a slow-release pellet containing 0.18 mg of 17β-estradiol and/or another pellet containing 5 mg tamoxifen was subcutaneously implanted into the back of nude mice (Innovative Research of America). The tumor volume was calculated as (length × width2)/2 every 3 days. Five weeks after orthotopic injection or 8 weeks after tail-vein injection, mice were euthanized, and the primary mammary tumors and lungs were collected for histologic analysis.
The copy-number variation and survival information of The Cancer Genome Atlas (TCGA) tumor set were downloaded from cBioPortal (TCGA BRCA provisional, www.cbioportal.org), while miR-135a-1 reads and clinical information were obtained from TCGA (cancergenome.nih.gov). All the GSE files were downloaded from the Gene Expression Omnibus. The METABRIC data were adapted from Kaplan–Meier plotter (17). If not specified otherwise, GraphPad Prism was used for statistical analysis. The χ2 test was performed using SPSS. All experiments were repeated at least 3 times. P < 0.05 was considered statistically significant.
Deletion of the MIR135A1 gene is associated with poor prognosis in breast cancer
To specifically define the DNA copy-number changes of 34 presently known miRNAs located at chromosome 3p, we performed both genomic real-time PCR in 20 primary breast cancer samples and in silico analysis with TCGA invasive breast cancer data (18, 19). As shown in Supplementary Fig. S1A; Fig. 1A, MIR135A1 is one of the most frequently deleted miRNA genes in our cohort of breast cancer specimens (20%) and in TCGA breast cancer samples (32%). We further examined its correlation with the clinicopathologic features of breast cancer in the TCGA dataset (Supplementary Table S3). MIR135A1 locus deletion was observed to be significantly associated with a higher pathologic T stage (P < 0.001), younger patient age (P = 0.006), and the invasive ductal carcinoma histologic type (P < 0.001). Furthermore, Kaplan–Meier survival analyses revealed that TCGA patients with breast cancer with MIR135A1 deletion exhibited a significantly worse survival outcome as compared with those without deletion (Fig. 1B, P = 0.0242).
As an intergenic miRNA gene on human chromosome 3p21.1, MIR135A1 is transcribed as pri-miR-135a-1 and processed to generate the mature miR-135a (20). Analysis of TCGA breast cancer samples revealed a marked decrease in pri-miR-135a-1 reads consistent with genomic deletion of the MIR135A1 gene (Supplementary Fig. S1B, P < 0.001). Mature miR-135a can also be produced from its second precursor, pri-miR-135a-2 on chromosome 12 (20), but in RNA-Seq data derived from TCGA breast tumors, the reads per million of pri-miR-135a-1 were significantly higher than those of pri-miR-135a-2 (Supplementary Fig. S1C, P < 0.001). As pri-miR-135a-1 is the predominant precursor of mature miR-135a in breast cancer samples, our study subsequently focused on pri-miR-135a-1. Patients with breast cancer with lower pri-miR-135a-1 level exhibited significantly shorter survival in the TCGA cohort (Fig. 1C, P = 0.0397). The correlation between pri-miR-135a-1 levels and clinicopathologic characteristics of the TCGA cohort was summarized in Supplementary Table S4, wherein lower pri-miR-135a-1 level was associated with patient menopause status, ERα and PR negativity, and lower HER2 level. Furthermore, we observed a significantly lower expression of miR-135a in our cohort of breast cancer (n = 48) compared with benign breast tissues (n = 28) using qRT-PCR (Fig. 1D, P < 0.001). Meta-analysis based on a combined public breast cancer dataset with matched miRNA and mRNA arrays (Oxford breast tumor set, GSE22216 and GSE22220, respectively; ref. 9) revealed that lower miR-135a but not miR-135b levels are significantly associated with poor distant relapse-free survival (DRFS) of patients with breast cancer (Fig. 1E–F). Similarly, most patients with breast cancer with distant relapse at 10 years were observed to have lower miR-135a expression than those without distant relapse in GSE22216 (Supplementary Fig. S1D). In addition, the miR-135a level is observed to be inversely correlated with histologic grade and tumor size in the Oxford breast tumor cohort, although the correlation between miR-135a levels and tumor size did not reach statistical significance (Supplementary Fig. S1E–S1F). Taken together, these findings suggest that MIR135A1 deletion and reduced miR-135a expression predict for poor survival in breast cancer.
miR-135a is a tumor suppressor in breast cancer cells
A set of 15 different breast cancer cell lines was analyzed for MIR135A1 copy-number and miR-135a expression. MIR135A1 locus deletion was observed in 3 cell lines, ZR75-1, BT-474, and SKBR3 (Fig. 1G). For functional studies, the effects of forced expression or inhibition of miR-135a were examined in both ERα+ (MCF-7, T47D) and ERα- (MDA-MB-231) breast cancer cells. The empty vector, retroviral plasmids harboring either pri-miR-135a-1 sequence or miR-135a sponge with a cassette of bulged miR-135a-binding sequences, was stably transfected into the breast cancer cells. The cells were designated as –Vec, -miR-135a and –Sponge cells, respectively, and the expression levels of miR-135a in these cells were verified by qRT-PCR (Supplementary Fig. S2A). The forced expression of miR-135a resulted in a significantly smaller increase, while the inhibition of miR-135a led to a significantly greater increase in total cell viability of MCF-7 cells over 5 days (Fig. 2A; Supplementary Fig. S2B). Concordantly, forced expression of miR-135a increased the G0–G1 phase fraction, decreased the proliferating (S and G2–M phase) fraction, and increased the percentage of apoptotic cells in MCF-7 cells, while miR-135a inhibition showed the opposite effects (Fig. 2B and C). For 3D Matrigel and soft-agar colony formation, MCF-7-miR-135a and –sponge cells exhibited decreased and increased growth respectively, compared with MCF-7-Vec cells (Fig. 2D and E). Despite differences in the extent of effects, the tumor suppressive effects of miR-135a were in general also observed in T47D and MDA-MB-231 cells (Supplementary Fig. S2B–S2E).
Given that a lower level of miR-135a expression was observed in highly invasive breast cancer cell lines (Hs578T, MDA-MB-231, MDA-MB-435, and SUM-159) compared with relatively less invasive breast cancer cell lines (T47D, MCF-7, and MDA-MB-361; Fig. 1G), we also assessed the possible roles of miR-135a in modulating cell motility. Forced expression of miR-135a decreased, while miR-135a inhibition increased cell migration and invasion of MDA-MB-231 cells as shown in the wound-healing (Fig. 2F–G), and Transwell migration and invasion assays (Fig. 2H–I) respectively. A similar effect was observed in MCF-7 and T47D cells (Supplementary Fig. S2F-G). Epithelial-to-mesenchymal transition (EMT) is a common characteristic of carcinoma cells undergoing migration and invasion. Forced expression of miR-135a in the mesenchymal MDA-MB-231 cells resulted in phenotypic conversion to a more epithelial phenotype, characterized by relatively compact cell colonies, less elongated cellular morphology with fewer numbers of cellular protrusions, and less accumulation of F-actin at the cell periphery (Fig. 2J). Strikingly, analysis of the correlation between the expression of miR-135a and a set of EMT-associated genes revealed that miR-135a levels were inversely correlated with the mRNA levels of mesenchymal markers (FN1, SNAI1, TWIST1, MMP9, and EZH2) and positively correlated with that of epithelial markers (JUP, OCLN, GJA1, and CTNNA1) in the Oxford breast tumor cohort (Supplementary Fig. S2H; Supplementary Table S5). Consistently, forced expression of miR-135a in MDA-MB-231 and MCF-7 cells decreased the protein levels of mesenchymal markers (FN1, Ezh2, and Twist1) and increased those of epithelial markers (γ-catenin and Occludin) compared with the control cells, whereas the inhibition of miR-135a showed the opposite effects (Fig. 2K; Supplementary Fig. S2I). These results suggest that miR-135a inhibits EMT, migration, and invasion of breast cancer cells.
We next examined the roles of miR-135a in breast tumor growth in vivo. MCF-7-sponge cell–derived tumors grew significantly faster, resulting in a 2-fold increase in tumor volume in the fifth week as compared with tumors derived from the MCF-7-Vec cells (Fig. 3A; Supplementary Fig. S3A–S3B). In contrast, MCF-7-miR-135a cell–derived tumors had a 60% reduction in tumor volume as compared with control tumors (Fig. 3A; Supplementary Fig. S3A–S3B). In addition, Ki-67 and TUNEL staining revealed that the tumors derived from MCF-7-miR-135a cells exhibited a lower percentage of proliferative cells and a higher percentage of apoptotic cells compared with control tumors, while tumors derived from MCF-7-sponge cells showed increased proliferation and reduced apoptosis compared with control tumors (Fig. 3B-C). Furthermore, hematoxylin and eosin staining of tumor sections revealed that tumors derived from both MCF-7-miR-135a cells and control MCF-7-Vec cells were well-encapsulated and noninvasive, while those derived from MCF-7-sponge cells were highly invasive with significant local infiltration into muscle and fat pad tissues (Fig. 3D). We further examined the effects of miR-135a on tumor metastatic capacities in tail-vein injection models. MDA-MB-231-miR-135a cells exhibited reduced, while MDA-MB-231-sponge cells yielded increased number and size of lung micrometastases compared with MDA-MB-231-Vec cells (Fig. 3E–F). Moreover, a significantly lower frequency of lung metastasis was observed in mice injected with MDA-MB-231-miR-135a cells than those injected with MDA-MB-231-Vec cells (Fig. 3G). Collectively, miR-135a was demonstrated to act as a tumor suppressor in breast cancer, through decreasing cell proliferation, survival, migration, and invasion in vitro, and inhibiting tumor growth and lung metastasis in vivo.
The estrogen/ERα axis promotes miR-135a expression
We observed that the average expression levels of miR-135a were 5-fold higher in ERα+ breast cancer cells with an integral MIR135A1 gene locus (T47D, MCF-7, MDA-MB-361) as compared with the ERα- breast cancer cells (Hs578T, MDA-MB-231, MDA-MB-435, MDA-MB-453, SUM-159; Fig. 1G). Similarly, analysis of the Oxford breast tumor cohort revealed that the expression level of miR-135a is significantly higher in ERα+ breast tumors than in ERα− breast tumors (Supplementary Fig. S4A), and positively correlates with ERα mRNA (ESR1) levels (Fig. 4A). Next, we measured miR-135a levels in MCF-7 cells transfected with either the empty pIRES vector or pIRES-ERα, and treated with either vehicle control or ERα antagonist, fulvestrant (ICI 182,780). Forced expression of ERα increased miR-135a expression, while fulvestrant treatment, which downregulated ERα levels, also reduced miR-135a expression (Supplementary Fig. S4B–S4C). These data collectively suggest that miR-135a expression in breast cancer is ERα dependent.
Estrogen regulation of miR-135a expression was further investigated in MCF-7 cells. Under estrogen-depleted conditions, the levels of pri-miR-135a-1 and mature miR-135a were significantly decreased in a progressive manner, consistent with that of estrogen-responsive TFF1 (Fig. 4B; Supplementary Fig. S4D). Under estrogen-replete conditions, E2 treatment significantly increased the levels of pri-miR-135a-1 and mature miR-135a progressively over time, similar to E2-stimulated-TFF1 expression (Fig. 4C; Supplementary Fig. S4E). In contrast, the expression of pri-miR-135a-2 was observed to be estrogen-independent (Supplementary Fig. S4D–S4E). Furthermore, estrogen-stimulated miR-135a expression was significantly decreased by ERα antagonism with either tamoxifen or fulvestrant (Fig. 4D). Hence, miR-135a expression is dependent on the transcriptional regulation of pri-miR-135a-1 by the estrogen/ERα axis.
We further determined whether ERα directly binds the promoter region of MIR135A1. The ChIP-sequencing (ChIP-Seq) data from ChIPBase (21) showed ERα enrichment within the region 5 kbps upstream of the MIR135A1 but not MIR135A2 gene (Supplementary Fig. S4F). The rVista 2.0 software was utilized to map the conserved estrogen response element (ERE) and 6 other ERα binding sites (Fig. 4E) in this same region of the MIR135A1 promoter (22). ChIP studies confirmed the direct binding of ERα to the promoter region of MIR135A1 at the conserved ERE site (region 2; Fig. 4F). Estrogen-dependent recruitment of RNA polymerase II to region 4 of MIR135A1 promoter was also observed (Fig. 4F). The estrogen-dependent recruitment of ERα and RNA polymerase II to the TFF1 promoter (positive control) but not to a DNA region at chromosome 12p13.3 with no predicted binding sites (negative control) was verified (Supplementary Fig. S4G).
To study the effect of ERα on the transcriptional activity of MIR135A1 promoter, we cloned two different lengths of the MIR135A1 promoter, namely, Pro S, which contains the conserved ERE sequence (∼400 bps), and Pro L, which contains both the ERE and RNA Polymerase II binding sites (∼3,000 bps), into luciferase reporter plasmids (Fig. 4E). Forced expression of ERα in ERα- HEK-293T cells increased luciferase reporter activity of Pro L but not Pro S, whereas ERα deletion in MCF-7 cells decreased luciferase reporter activities of both promoter constructs (Fig. 4G; Supplementary Fig. S4H–S4I). Furthermore, ERα antagonism by tamoxifen or fulvestrant repressed the E2-stimulated transcriptional activity of Pro L in MCF-7 cells (Supplementary Fig. S4J). Moreover, the mutation of the ERE element in Pro L completely abrogated the estrogen responsiveness of Pro L (Fig. 4H). Taken together, binding of both ERα and RNA Pol II to the promoter region of MIR135A1 is required for its transcriptional activation by estrogen.
miR-135a depletion decreases tamoxifen sensitivity and promotes acquired tamoxifen resistance in ER+ breast cancer cells
As miR-135a expression is ERα regulated, we further studied if miR-135a regulated estrogen responsiveness of ERα+ breast cancer cells. MCF-7 and T47D cells stably transfected with miR-135 sponge or empty vector were estrogen starved prior to treatment with 10 nmol/L E2 or vehicle. Compared with MCF-7-vector cells, MCF-7-sponge cells exhibited a nonsignificant decrease in cell viability, smaller increase in the G0–G1 phase-arrested and apoptotic cell population, and smaller reduction in the capacity for foci formation under estrogen-depleted conditions, suggestive of reduced estrogen dependence (Fig. 5A–D). Similarly, miR-135a inhibition in T47D cells resulted in significant estrogen-independent growth (Supplementary Fig. S5A–S5B).
As reduced cellular need for estrogen has been associated with resistance to antiestrogen therapies of ERα+ breast cancer (23, 24), we determined whether miR-135a modulates sensitivity and acquired resistance of ERα+ breast cancer cells to tamoxifen. Two acquired tamoxifen-resistant cellular models, namely, MCF-7 tamoxifen-resistant (TamR) and T47D TamR, were previously generated (25) and verified (Supplementary Fig. S5C–S5F). miR-135a levels were markedly decreased in TamR cells as compared with their respective parental cells (Fig. 5E), consistent with previous reports (Supplementary Fig. S5G; refs. 11, 12). Decreased miR-135a expression was also observed in TamR breast cancer samples (Supplementary Fig. S5H; ref. 26). Furthermore, whereas miR-135a expression was progressively downregulated by tamoxifen treatment in parental cells, consistent with our finding of ERα-regulated miR-135a expression, it was not significantly altered by tamoxifen treatment in TamR cells (Supplementary Fig. S5I). ChIP analysis showed that ERα bound to the MIR135A1 promoter, and tamoxifen treatment decreased ERα binding to this DNA region in both parental and TamR cells (Fig. 5F). Interestingly, the enrichment of RNA polymerase II in the MIR135A1 promoter was observed and decreased upon tamoxifen treatment in parental cells, but not in resistant cells, indicating transcriptional silencing of MIR135A1 in resistant cells with or without tamoxifen (Fig. 5F). To verify if reduced miR-135a expression contributed to acquired tamoxifen resistance in ER+ breast cancer cells, we forced expressed miR-135a–specific sponge in parental cells and miR-135a in TamR cells (Supplementary Fig. S2A). Notably, forced expression of miR-135a resulted in a greater tamoxifen-mediated decrease in cell viability in MCF-7 and T47D cells (Supplementary Fig. S5J–S5K), indicative of increased tamoxifen sensitivity. In contrast, miR-135a inhibition resulted in a smaller tamoxifen-mediated decrease in cell viability of MCF-7 and T47D cells (Fig. 5G; Supplementary Fig. S5K). Consistently, the inhibition of miR-135a reduced the tamoxifen-induced increase in G0–G1 phase cell-cycle arrest and apoptosis in parental cells (Fig. 5H–I). Importantly, forced expression of miR-135a significantly decreased cell viability with increased G0–G1 phase cell-cycle arrest and apoptosis of TamR cells upon tamoxifen treatment, indicative of partial resensitization toward tamoxifen (Fig. 5G–I). Similarly, the foci formation assays revealed inhibition of miR-135a decreased tamoxifen sensitivity of parental cells, while the forced expression of miR-135a partially overcome tamoxifen resistance in TamR cells (Fig. 5J).
Xenograft studies were also used to determine the effect of miR-135a on tamoxifen resistance in vivo. A total of 2 × 106 miR-135a sponge-transfected MCF-7 parental cells, miR-135a-transfected MCF-7 TamR cells, or their cognate control cells were orthotopically injected into the mammary fat pad of nude mice in the presence of exogenous estrogen supplement. When tumor volume reached 100 mm3, a slow-release tamoxifen pellet was implanted to the treatment group as indicated. The growth of tumors derived from the empty vector–transfected MCF-7 parental cells was inhibited by tamoxifen treatment, while that of tumors derived from miR-135a sponge-transfected MCF-7 parental cells was not significantly (Fig. 5K; Supplementary Fig. S5L), suggesting that miR-135a inhibition decreased tamoxifen sensitivity in vivo. In contrast, whereas the growth of tumors derived from empty vector–transfected MCF-7 TamR cells was unaffected by tamoxifen treatment, that of tumors derived from miR-135a–transfected MCF-7 TamR cells were significantly decreased (Fig. 5K; Supplementary Fig. S5L), suggestive of partial resensitization to tamoxifen in vivo. Consistent with our data, patients with ERα+ breast cancer with higher miR-135a expression showed significantly prolonged DRFS compared with those with lower miR-135a expression in the Oxford breast cancer cohort, which was in accordance with a previous report (Fig. 5L; ref. 27). Notably, patients with ERα+ breast cancer with lower miR-135a expression showed a similar survival curve with ERα− patients (Fig. 5L). In addition, lower miR-135a expression levels correlated with shorter overall survival in METABRIC patients receiving antiendocrine therapies (Fig. 5M).
Identification of direct miR-135a targets including ERα
To elucidate the molecular mechanism(s) utilized by miR-135a in breast cancer cells, we further sought miR-135a target genes. The regulation of ERα pathway genes by miR-135a was also studied. Previously reported target genes of miR-135a were examined (28–33) but none were observed to be downregulated by miR-135a at the mRNA level (Supplementary Fig. S6A–S6B). We thus sought to identify potential novel targets of miR-135a in breast cancer cells. The predicted miR-135a targets (TargetScan, Version 6.2; ref. 34) were compared with either the set of ERα pathway genes or the group of genes, whose expression levels negatively correlate with those of miR-135a, in the transcriptomic profile of breast tumor tissues (GSE22220; Fig. 6A; Supplementary Fig. S6C and Supplementary Table S5). Putative target genes (221) of miR-135a were identified. Among those, the ERα (ESR1), estrogen receptor–related protein alpha ERRα (ESRRA) and nuclear coactivator 1 (NCOA1) are key mediators of estrogen signaling, whereas the pim2 oncogene (PIM2), muscle ras oncogene (MRAS) and lymphocyte cytosolic protein 1 (LCP1) have been implicated in breast tumorigenesis (35–39). Notably, ESRRA, PIM2, MRAS, and LCP1 mRNA levels negatively correlate with that of miR-135a in tumor tissues (Supplementary Fig. S6C).
Consistent with bioinformatics prediction, miR-135a mimic–transfected MCF-7 cells exhibited decreased ESR1, ESRRA, NCOA1, PIM2, MRAS, and LCP1 mRNA expression as compared with scrambled oligonucleotides-transfected cells (Fig. 6B). We further determined whether miR-135a directly binds the 3′UTR of these mRNAs. We introduced mutations into the miR-135a binding sites (MBS) in the 3′UTR regions of these mRNAs and cloned either the wild-type or mutant 3′UTR sequences into luciferase reporter plasmids (Fig. 6C) for cotransfection with miR-135a mimics or scrambled oligonucleotides into MCF-7 cells. Compared with control, miR-135a decreased the wild-type 3′UTR-driven luciferase reporter activities of each of the six genes (Fig. 6D). Furthermore, mutation of the predicted MBS(s) in the 3′UTR abrogated miR-135a–mediated decrease in luciferase reporter activities (Fig. 6D), and enrichment of these six genes was also observed by pulling down mRNAs with biotinylated miR-135a mimics (Fig. 6E), suggesting the direct binding of miR-135a to these site(s). Moreover, forced expression of miR-135a decreased, whereas miR-135a inhibition increased protein expression of these target genes in MCF-7 and T47D cells, and tumors derived from MCF-7 cells (Fig. 6F; Supplementary Fig. S6D–S6G). Thus, ESR1, ESRRA, NCOA1, PIM2, MRAS, and LCP1 are all bona fide targets of miR-135a.
Consistent with our observation of miR-135a–targeted downregulation of ESR1, miR-135a mimics markedly decreased both the basal and estrogen-stimulated ERα transcriptional activity of estrogen response element (ERE4) in MCF-7 cells, while tamoxifen inhibition of ERα abrogated miR-135a–mediated decrease in reporter activity (Fig. 6G). In accordance, miR-135a mimics decreased the expression of the ERα-responsive gene TFF1 in MCF-7 cells (Supplementary Fig. S6H). Therefore, miR-135a, being an ERα-responsive gene, negatively regulates ERα expression and activity in a negative feedback loop.
We further determined whether these miR-135a targets mediate the functionality of miR-135a. The expression of miR-135a targets in MCF-7 cells were depleted using the respective shRNA (Supplementary Fig. S6I–S6J). The depletion of each miR-135a target, except NCOA1, significantly reduced the viability of MCF-7 cells (Supplementary Fig. S6K). Furthermore, the combined silencing of ESR1/ESRRA/NCOA1 or PIM2/MRAS/LCP1 significantly reduced cell viability, migration, and invasion of MCF-7 cells (Supplementary Fig. S6K–S6L). Notably, whereas miR-135a inhibition promoted an estrogen-independent increase in cell viability and hence decreased tamoxifen sensitivity of MCF-7 cells, the depletion of ESR1/ESRRA/NCOA1 or PIM2/MRAS/LCP1 diminished this estrogen-independent growth and partially reversed tamoxifen resistance of MCF-7-sponge cells (Fig. 6H). In addition, TamR ERα+ breast cancer cells with reduced miR-135a levels exhibited elevated expression of miR-135a targets, except ERα, as compared with the respective parental cells (Supplementary Fig. S6M–S6N). The depletion of ESR1, PIM2, MRAS, and LCP1 alone and the combined depletion of ESR1/ESRRA/NCOA1 or PIM2/MRAS/LCP1 significantly resensitized TamR cells to tamoxifen (Supplementary Fig. S6O).
Loss of miR-135a mediates tamoxifen resistance in ERα+ breast cancer cells partially through ERα-ERK1/2-AKT cross-talk
The cross-talk between ERα and growth factor receptor pathways, especially elevated MAPK/ERK and PI3K/AKT activities, has been reported to be pivotal for tamoxifen resistance in breast cancer (23, 24). We sought to determine if our identified miR-135a targets coordinately act through the MAPK/ERK and PI3K/AKT signaling pathways to mediate acquired tamoxifen resistance. Depletion of each of the six miR-135a targets decreased the phosphorylation of ERK1/2 and AKT in MCF-7 cells (Fig. 7A; Supplementary Fig. S7A). In accordance, forced expression of miR-135a reduced, whereas miR-135a inhibition increased the levels of pERK1/2 and pAKT in MCF-7 cells and tumors (Fig. 7B; Supplementary Fig. S7B-C). Consistent with previous reports (25, 40), elevated activities of ERK1/2 and AKT were observed in TamR cells compared with their respective parental cells (Supplementary Fig. S7D–S7E). We further investigated the involvement of the ERK1/2 and AKT pathways in decreased tamoxifen responsiveness mediated by miR-135a loss, using inhibitors of ERK1/2 and AKT (Supplementary Fig. S7F–S7G). The inhibition of pERK1/2 and/or pAKT partially reversed tamoxifen resistance in MCF-7 TamR cells (Supplementary Fig. S7H). The inhibition of the ERK1/2 and/or AKT pathway failed to diminish the growth advantage of MCF-7-sponge cells, suggesting that miR-135a function was not completely ERK1/2 and AKT dependent (Fig. 7C). However, inhibition of ERK1/2 and/or AKT increased tamoxifen sensitivity of MCF-7-sponge cells (Fig. 7D).
It has been well established that ERα can be phosphorylated by ERK1/2 and AKT and phosphorylated ERα has been implicated to regulate ligand-independent activation of ERα and tamoxifen resistance (41–43). Inhibition of ERK1/2 and AKT signaling by either inhibitors or siRNAs attenuated ERα phosphorylation as expected (Supplementary Fig. S7I-L). Forced expression of miR-135a reduced the phosphorylation of ERα at both S118 and S167 sites (Fig. 7E; Supplementary Fig. S7M). Importantly, miR-135a inhibition led to an increased phosphorylation of ERα, similar with that observed in TamR cells (Fig. 7E; Supplementary Fig. S7M–S7O). Increased ERα phosphorylation was also observed in MCF-7-sponge cells grown in estrogen-free and serum-free medium (Fig. 7E; Supplementary Fig. S7M), indicating ligand-independent phosphorylation of ERα by miR-135a inhibition. Moreover, inhibition of the ERK1/2 and AKT pathway fully abrogated the increased ERα phosphorylation in MCF-7-sponge cells (Supplementary Fig. S7P–S7Q), suggestive of ERK1/2 and AKT dependency for ERα phosphorylation by miR-135a inhibition. As ERα phosphorylation is required for full activation of ERα (41–43), we next investigated the effect of ERK1/2 and AKT signaling on ERα-mediated miR-135a expression. Inhibition of ERK1/2 and/or AKT activity by siRNAs or inhibitors markedly attenuated estrogen-stimulated miR-135a expression in MCF-7 cells (Fig. 7F–G). However, in estrogen-free conditions or upon ERα degradation by fulvestrant, ERK1/2, and/or AKT inhibition failed to affect miR-135a expression, indicating that the regulation of miR-135a expression by ERK1/2 or/and AKT is estrogen-ERα dependent (Fig. 7F–G).
MIR135A1 was herein identified as one of the most frequently deleted miRNA genes at the short arm of human chromosome 3 and predicted for poorer survival in patients with breast cancer, suggestive of a tumor suppressive role of miR-135a, which was confirmed by our in vitro and in vivo studies. Our observations are consistent with other previous studies showing that miR-135a suppresses tumor growth and metastasis of breast cancer (44–46). In contrast, miR-135a has been reported to promote breast cancer invasion and migration through mediating the depletion of metastasis suppressor, HOXA10 (33). miR-135a targets a variety of gene transcripts, including those of oncogenes and tumor suppressors, which may account for the observed incongruity in the effects of miR-135a in different cell lines (2). Importantly, the downregulation of miR-135a in TamR ERα+ breast cancer cells as compared with their respective tamoxifen-sensitive parental cells has been consistently observed in previous and our studies (11, 12).
We have herein established a novel regulatory relationship between miR-135a and ERα–ERK–AKT cross-talk. As miR-135a is a novel ERα responsive gene, yet ERα transcript is a miR-135a target, we proposed that ERα transcriptionally activates miR-135a to regulate its own activity in a negative feedback loop in breast cancer. Our study has further provided a mechanistic link between miR-135a and the activation of ERK1/2 and AKT1 in tamoxifen resistance. The elevated activities of ERK1/2 and AKT1 have been demonstrated to play a critical role in the acquisition of tamoxifen resistance (23, 24). The decreased expression of miR-135a resulted in an increased level of the miR-135a target genes (ESR1, ESRRA, NCOA1, PIM2, MRAS, and LCP1), which we have demonstrated to be key mediators of ERK1/2 and AKT1 activation, and subsequent increased ERα transcriptional activity to promote tamoxifen resistance. Thus, miR-135a was identified as the key molecule in orchestrating the signaling cross-talk between ERα and ERK1/2/AKT1 in ERα+ breast cancer cells. In tamoxifen-treated patients with ERα+ breast cancer, the tamoxifen-mediated downregulation of miR-135a expression derepressed ERα–ERK–AKT signaling, leading to decreased tamoxifen response.
In summary, MIR135A1 copy-loss and/or low miR-135a expression can potentially be used as a prognostic biomarker for ERα+ breast cancer aggressiveness, as well as a predictive biomarker for tamoxifen sensitivity or the acquisition of tamoxifen resistance. Hence, the miR-135a–based novel therapeutic approach might be warranted for inhibiting breast cancer progression and overcoming tamoxifen resistance in these patients with ERα+ breast cancer.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: W. Zhang, M. Wu, P.E. Lobie, T. Zhu
Development of methodology: W. Zhang, M. Wu, S. Tan
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): W. Zhang, M. Wu, M. Zhang, X. Zhang, Y. Zhong, P. Qian, S. Tan, K. Ding
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): W. Zhang, M. Wu, M. Zhang, S. Tan, G. Li, K. Ding, P.E. Lobie, T. Zhu
Writing, review, and/or revision of the manuscript: W. Zhang, M. Wu, Q.-Y. Chong, X. Kong, P.E. Lobie, T. Zhu
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W. Zhang, M. Wu, L. Hu, X. Kong
Study supervision: T. Zhu
The authors greatly thank Dr. Suling Liu (USTC, China), Dr. Ping Gao (USTC, China), and Dr. Saraswati Sukumar (JHU, USA) for providing SUM-159 cells, HEK-293T cells, and pIRES-ERα and ERE-Luc plasmids, respectively. This work was supported by The Key Research and Development Program of China (2016YFC1302305) and The National Natural Science Foundation of China (81472494 and 81502282) to T. Zhu and The Shenzhen Development and Reform Commission Subject Construction Project (2017) 1434 to P.E. Lobie. The authors would like to thank Hin-Ching Lo (BCM, USA) for the language review.
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.