Introduction
Breast cancer is the most common malignancy that occurs in women globally [
1]. Despite the advancement in therapies, many women will suffer from relapse, acquiring metastatic lesions in distant sites and eventually succumbing to cancer related deaths [
2‐
4]. Currently there is a lack of targeted therapies, in particular towards the triple negative breast cancer (TNBC) subtype. TNBC are aggressive and highly invasive, and these tumours lack estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor 2 (HER2) expression [
5].
A large body of evidence has identified augmented receptor tyrosine kinase (RTK)-PI3K-Akt-mTOR activity in the basal like and TBNCs either due to mutations in RTKs or PIK3CA or loss of phosphatase and tensin homologue (PTEN) expression [
6‐
9]. This pathway has therefore become a major focus of breast cancer drug development, although patient responses to these novel drug compounds in clinical trials have been variable, perhaps due to an incomplete understanding of pathway interactions and feedback loops. Clearly, a thorough understanding of the regulation of these signalling pathways is essential to effectively personalise breast cancer treatment.
microRNAs (miRNAs) are small non coding RNAs that modulate gene expression post transcriptionally. They typically silence their targets by binding their 3′ untranslated regions (UTRs) in a sequence-specific manner [
10]. There are a wealth of experimental data demonstrating the pleiotropy of miRNAs in regulating a multitude of cellular processes, such as embryonic stem cell differentiation [
11,
12], cell fate and lineage commitment [
13‐
15], organogenesis [
16‐
18] and oncogenesis [
19‐
23].
The pubertal developing mammary gland is elaborated through fat pad invasion by terminal end buds (TEBs); poorly differentiated, unpolarised, proliferative and invasive structures [
24] that are enriched for stem and progenitor cell activity [
25], with many cellular and molecular similarities to neoplastic cells [
26]. We previously examined the expression of mRNAs between cellular subsets of the developing mammary gland, found GATA-3 to be specifically expressed in epithelial cells and went on to identify GATA-3 as an important breast tumour suppressor gene [
27,
28].
In this study, we comprehensively profiled miRNA expression in the different cellular compartments within the mammary epithelium of mice. We identify miR-184 as a microRNA associated with epithelial differentiation and demonstrated that miR-184 is silenced and methylated in a subset of TBNC. Further functional characterisation of miR-184 revealed that it is a potential tumour suppressor miRNA in breast cancer; in suppressing cell proliferation, self-renewal in vitro and delaying the formation of metastatic lesions in distant sites in vivo. By performing microarray studies and informatic analysis, we discovered that miR-184 regulates the AKT/mTORC1 pathway by targeting AKT2, TSC2 and PRAS40 in suppressing activity of S6K1 and protein synthesis.
Methods
Expression profiling of TEBs, ducts and stroma
TEBs (n = 4), mature ducts (n = 4), and distal stroma regions (n = 4) were microdissected from mammary glands of anaesthetised 5-week-old β-actin–GFP reporter mice (FVB/n, Jackson laboratory, Bar Habor, Maine, USA) using a Leica fluorescence microscope (Leica microsystems, Wetzlar, Germany). Tissue samples were homogenised in Trizol Reagent (Life Technologies, Carlsbad, CA, USA) with a polytron tissue homogeniser (Thermo Fisher Scientific, Waltham, MA, USA). Total RNA was extracted according to a modified protocol based on the manufacturer’s instructions (Invitrogen). The final RNA pellet was ethanol precipitated and washed in 80 % ethanol. RNA samples consisting of TEBs (n = 3), mature ducts (n = 2) and stroma (n = 3) were sent to the Ramaciotti Centre for Gene Function Analysis for miRNA microarray profiling (University of New South Wales, Sydney, Australia) using the SurePrint Mouse miRNA Array V2 (Agilent Technologies, Santa Clara, CA, USA). All animal work was approved by the Garvan/St Vincent’s Hospital Animal Ethics Committee and conducted in accordance with NHMRC guidelines for the ethical treatment of animals.
Cell lines, retroviral infections
MDA-MB-231 and BT549 cells were maintained in RPMI 1640 supplemented with 10 % FBS and 0.25 % human insulin. HEK293E cells were maintained in DMEM supplemented with 10 % FBS. MDA-MB-231 cells were obtained from the EG & G Mason Research Institute, Worcester, Massachusetts, USA and DNA fingerprinted. BT-549 cell lines were obtained from ATCC. All cell lines used in experiments were cultured at 37 °C in 5 % CO2 and 95 % air. MDA-MB-231 cells were first transduced with the retroviral vector pRQ-rtTA-GFP followed by pRQ-miR184.
Primary tumour samples
Human tumour samples consisting of luminal A (ER+, PR+, Her2−; n = 10), Her2 (ER−, PR−, Her2+; n = 9), triple negative (ER−, PR−, Her2−; n = 7) and matched normal (n = 7) were obtained from the Victoria Cancer Biobank (VCB). Research use of human tissues was approved by the St Vincents Hospital Human Research Ethics Committee (Approval # 08/145). Tumour samples were homogenised using a mortar and pestle. Total RNA samples were extracted using miRVana kit (Invitrogen) and ethanol precipitated with 80 % ethanol.
Transfection with miRNA mimics
For MDA-MB-231 and BT549 cells, miRIDIAN miRNA mimics (Dharmacon, Lafayette, CO, USA) were transfected into cells using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s protocol for transfecting siRNA. Mimics (final concentration of 50 nM) were mixed with 1 % lipofectamine 2000 (v/v) diluted in Opti-MEM transfection medium (Invitrogen) and incubated for 20 minutes. The mimics were added dropwise onto cells in growth medium at a final concentration of 50 nM. Fresh medium was replaced 24 h post transfection. A non-radioactive cell proliferation assay (Promega, Madison, WI, USA) was used to assess the number of viable cells. Three biological replicates were conducted.
MDA-MB-436 and HS578T breast cancer cells were reverse transfected with 40 nM of either miR-184 mimic or scrambled using Dharmafect 4 reagent (Dharmacon) following manufacturer’s instructions; SiTox (Dharmacon) was used as a positive control. The following day, medium was changed and cells were left in culture for an additional 72 h. To assess viability, CellTiter-Glo (Promega) was added directly into the cell medium (1:2 ratio) and left to incubate for 10 minutes; luminescence was read using FLUOstar Omega plate reader (BMG Labtech, Ortenburg, Germany). The experiment was performed in three biological replicates.
Quantitative RT-PCR
Total RNA was extracted using the Trizol reagent (Invitrogen) method with a modification where the final RNA pellet was ethanol precipitated and washed in 80 % ethanol. cDNA was generated using the Taqman MiRNA Reverse Transcription Kit (Applied Biosystems) using the specific primer from TaqMan MiRNA assay (Applied Biosystems) according to the manufacturer’s protocol. Quantitative PCR (qPCR) amplification was run on the 7900 Real-time PCR system (Applied Biosystems). All human and mouse miRNA expression values were normalised to RNU6B and SnoRNA202, respectively.
Immunoblot and 3′ UTR luciferase activity assay
Cells were lysed using complete radioimmune precipitation assay (RIPA) buffer supplemented with complete ULTRA protease inhibitor cocktail tablets (Roche, Basel, Switzerland) and sodium orthovanadate. Anti-Akt2, anti-phospho-Akt (Thr308), anti-phospho-Akt (Ser473), anti-Pras40, anti-phospho-Pras40 (Thr246), anti-Gsk3A, anti-phospho-Gsk3 (Ser21/9), anti-Tsc2, anti-phospho-Tsc2 (Thr1426), anti-mTOR, anti-phospho-mTOR (Ser2448), anti-p70S6k1, anti-phospho-p70S6k1 (Thr389), anti-p70S6K2, anti-4E-BP1, anti-4E-BP1 (Thr37/46) (Cell Signalling Technology, Danver, Massachusetts, USA) rabbit polyclonal antibodies were used in immunoblotting. Luciferase constructs (pLightSwitch_3′UTR) (Switchgear genomics, Carlsbad, CA, USA) containing the 3′ UTR region of Akt2, Pras40, Gsk3a, CSF1 and Itgb1 was individually transfected into HEK293T cells using pGL4.12 (luc2CP) as a normaliser. Luciferase activity was measured by using the dual luciferase assay (Promega).
Tumoursphere assays
MDA-MB-231 cells were cultured in serum-free RPMI 1640, supplemented with B27 (Invitrogen) and 20 ng/ml bFGF (BD Biosciences, Franklin Lakes, NJ, USA), and 4 μg/ml heparin (Sigma Aldrich, St. Louis, MO, USA) and plated at 15,000 viable cells/well in ultralow attachment 6-well plates (Corning Incorporate, NY, USA). Complete serum-free medium was added to the cells every 3 days. Primary tumourspheres were enumerated at day 10. Primary tumourspheres were collected, and were enzymatically dissociated into single cells, re-plated in ultralow attachment 6-well plates (Corning Incorporate) at a density of 1,000 viable cells/well and enumerated at day 10.
Protein synthesis assay
Cells were washed twice and serum starved for 16–18 h prior to EGF stimulation. Cells were stimulated with EGF in serum-free DMEM low glucose without L-ARG, L-LEU, L-LYS, sodium pyruvate and phenol red (Sigma Aldrich) for 1 h: [3H]Leucine (PerkinElmer, Waltham, MA, USA) was added at the same time as EGF to a final concentration of 5 μCi/ml. Cells were washed three times in ice-cold PBS, lysed using RIPA buffer followed by incubating cells with 10 % trichloroacetic acid (TCA) for 10 minutes to precipitate proteins. Pellets were washed three times in 10 % TCA. Pellets were resuspended in 50 nM NaOH with 1 % Triton X-100 at 65 °C for 30 minutes or until the pellet dissolved. The radioactivity of samples was assessed by measuring the scintillation count using the β-scintillation counter. The results were normalised for protein content using bicinchoninic acid (BCA) analysis.
Animal experiments
For primary tumour burden and spontaneous metastasis assays, 1 × 106 MDA-MB-231 cells were injected into the mammary fat pad of 8-week-old female NOD/SCID mice. Mice were culled at the ethical end point, and primary tumour and other organs such as lungs, spleen, lymph node, pancreas and brain were harvested. Metastatic lesions were quantified with a fluorescent microscope within 2 h of harvest.
Immunohistochemistry
Mouse tissues were extracted and fixed overnight at 4 °C in 10 % neutral-buffered formalin (Sigma), and stored in 70 % ethanol at 4 °C. Subsequently, tissues were embedded in paraffin and sectioned. Sections were stained with haematoxylin and eosin (H&E) and phospho-histone H3 (Cell signaling) in accordance with standard protocols. Scoring of phospho-Histone H3 immunostaining and mitotic figures was assessed by a specialist breast pathologist (SO’T).
Gene expression analysis
MDA-MB-231 cells were transfected with miR-184 or control mimics for 48 h before total RNA was extracted using the modified Trizol reagent protocol with an additional ethanol precipitation step. RNA samples were sent off to the Ramaciotti Centre for Gene Function Analysis for gene expression profiling using the affymetrix gene 1.0ST array (Affymetrix) (University of New South Wales, Sydney, Australia). Gene expression analysis was performed using gene pattern.
Statistical analysis
Statistical analysis was performed by using GraphPad Prism v6.0. T tests were performed to determine statistical significance, unless otherwise stated. P <0.05 was considered statistically significant.
Methylation analysis
The MBDCap-Seq experiment was performed by Dr Claire Stirzaker and Dr Jenny Song (Garvan Institute of Medical Research). Analysis of the results was performed by Dr Elena Zotenko (Garvan Institute of Medical Research). Briefly, methylated DNA was isolated using the MethylMinerTM Methylated DNA Enrichment Kit (Life Technologies). Genomic FFPET DNA was sonicated. MBD-Biotin Protein (3.5 μg) was coupled to 10 μl of Dynabeads M-280 Streptavidin according to the manufacturer’s instructions.
MBD biotin conjugated to the magnetic beads was washed three times and resuspended in one volume of 1 × bind/wash buffer. The capture reaction was performed by adding 500 ng to 1 μg sonicated DNA to the MBD-magnetic conjugates on a rotating mixer for 1 h at room temperature (RT). All capture reactions were done in duplicate. The beads were washed three times with 1 × bind/wash buffer. The bound methylated DNA was eluted using single high-salt elution buffer (2 M NaCl). Eluted DNA fraction was concentrated by ethanol precipitation using 1 μl glycogen (20 μg/μl), 1/10 volume of 3 M sodium acetate, pH 5.2 and two sample volumes of 100 % ethanol, and resuspended in 60 μl water.
Preparation of MBDCap-Seq libraries and Illumina sequencing
DNA, 10 ng, was prepared for Ilumina sequencing using the Illumina ChIP-Seq DNA sample prep kit (Illumina, San Diego, CA, USA) according to the manufacturer’s instructions. The library preparation was analysed on Agilent High Sensitivity DNA 1000 Chip. Each sample was sequenced on one lane of the GA11x.
Alignment of MBDCap-Seq data
Sequenced reads were aligned to the hg18 version of the human genome with bowtie [
29]. Sequence reads with three mismatches or more and reads mapping to multiple positions were excluded. Last, multiple reads mapping to exactly the same genomic coordinate were eliminated and only one read was retained for downstream analysis.
miRNA seed match analysis
The seed match analysis was performed as previously described by Melton et al. [
30]. Briefly, ensemble transcripts (hg19) of promoter, 5′ UTR, open reading frame (ORF) and 3′ UTRs and other annotated genes (hg19) were obtained from the UCSC Genome Browser. Relevant miRNA seed match (7mer-1A or 7mer-m8) was conducted on those transcripts using a custom Python script [
30]. Results from seed match analysis were mapped to Affymetrix IDs. Wilcoxon rank sum test was used to determine the
p values in this analysis.
Gene signature score and survival analysis
A stringent 18-gene signature repressed by miR-184 (fold-change >2, Table
2) was assessed for survival analysis using two independent cohorts from METABRIC [
31] and a cohort of women receiving neo-adjuvant chemotherapy [
32]. METABRIC gene expression data were downloaded from the European Genome-Phenome Archive (EGAS00000000083). Gene expression and clinical data from Hatzis et al. were downloaded from Gene Expression Omnibus (GEO) [GEO: GSE25066]. The gene signature score was defined by a weighted average method [
33] for each sample in the METABRIC discovery cohort. Survival curves were estimated using the Kaplan-Meier method, with overall survival used as the outcome metric.
Discussion
In our study, we have discovered miRNAs enriched in different subcellular compartments of the developing mammary ductal structure. These results suggest that some miRNAs are expressed in specific subcellular compartments, such as the TEBs and mature ducts, to modulate cellular processes such as proliferation and differentiation during ductal elongation. We then asked, however, if these miRNAs were differentially expressed and functional in cancer. As a proof of concept, we found TEB-enriched miRNAs to be highly expressed in a panel of breast cancer models whereas miRNAs enriched in the ducts displayed an opposite trend, being lowly expressed in cancer.
One of the miRNAs that followed this expression pattern was miR-184. Upon functional characterisation, there was compelling evidence to suggest that miR-184 is a tumour suppressor in certain cancer subtypes, as it suppressed cell proliferation and self-renewal in vitro and tumour growth in the primary and distant sites. A role in modulating metastasis is supported by analysis of a subset of TNBC patient samples, where miR-184 was epigenetically silenced in lymph node metastases, suggesting silencing of miR-184 can promote metastatic dissemination. Upon interrogating a large breast cancer cohort (METABRIC) dataset, we also observed a significant decrease in miR-184 expression in the ER-negative tumours compared to the ER-positive tumours. In this cohort, very few microRNAs associate with prognosis [
53], as we found for miR-184. However, elevated expression of high-confidence targets repressed by miR-184 predicted poor prognosis, consistent with our evidence from animal models that elevated miR-184 activity improves outcome.
miR-184 is crucial in regulating certain developmental processes such as the differentiation of neural stem cells, germ line cells and corneal epithelial cells [
54‐
56]. Several studies have established that miR-184 is lowly expressed in different malignancies such as childhood neuroblastoma, brain cancers, clear cell renal cell carcinoma and prostate cancer [
44,
57‐
60]. When miR-184 was ectopically overexpressed in vitro, it resulted in cell cycle arrest and apoptosis [
44,
57,
58], as well as impeding neuroblastoma xenograft formation resulting in longer survival in vivo [
61]. Though several downstream targets of miR-184 such as AKT2, NUMBL, SHIP2, NFAT1 have been identified in different cell types, nevertheless there have been no definite reports on the functional role of miR-184 in breast cancer nor detailed analysis of signalling pathways that are potentially modulated by miR-184 [
44,
54,
62,
63].
miR-184 has been previously shown to be controlled by epigenetic mechanisms in development. This was identified in the mouse brain, where the genomic region proximal to the miR-184 locus in adult neural stem cells contains CpG rich sequences instead of canonical CpG islands. Methyl-CpG binding protein 1 (Mbd1) binds to these CpG-rich sequences in the genomic regions surrounding miR-184, and represses the transcriptional activity of miR-184 [
54]. In a separate study, researchers performed bisulphite sequencing on umbilical cord blood graft CD4+ T cells and discovered a small putative CpG island just upstream of the miR-184 locus, and an additional 32 CpG sites present within the adjacent regions of the miR-184 locus making it an ideal target for epigenetic silencing [
63].
We are the first to provide evidence to suggest that the attenuated expression of miR-184 in cancer is potentially a result of epigenetic mechanisms. miR-184 was methylated in a subset of lymph node metastases in TNBC, providing supportive evidence that miR-184 may play a role as a novel mammary tumour suppressor. We hypothesised that the methylation of miR184 in metastatic tissue suggests a selective pressure against maintenance of miR-184 particularly during metastatic dissemination.
Our experimental evidence suggests that the anti-tumourigenic properties displayed by miR-184 are a consequence of miR-184 inhibiting the activity of the PI3K/AKT/mTORC1 pathway, therefore limiting protein synthesis. This study suggests that miR-184 suppresses the protein synthesis pathway by targeting several important members of the AKT/mTOR pathway. miR-184 represses the total levels of AKT2, which relieves the inhibitory function on TSC2, the crucial negative regulator of the mTORC1 pathway. In addition, the reactivation of TSC2 results in the abrogation of S6K1 activity, the effector of the protein synthesis pathway.
Furthermore, this signalling event was partially recapitulated in vivo, where miR-184 repressed several substrates within the AKT/mTORC1 pathway in a majority of tumours. The total expression of mTOR was reduced in the miR-184 cohort, potentially inhibited the formation of the mTORC1 and therefore reducing the activity of S6K1. Interestingly, we also observed a loss of TSC2 expression and an increase in PRAS40 expression in the miR-184 cohort. It is possible that these changes are compensation to circumvent the anti-proliferative effects of miR-184 during tumour progression.
These changes in the signalling pathway correlate with our protein synthesis assay, where we observe less synthesised proteins in cells overexpressing miR-184. However, we do not fully comprehend how miR-184 suppresses protein synthesis. Despite the overexpression of AKT2, PRAS40 and GSK3A in combination, we were unable to rescue the miR-184 mediated suppression of protein synthesis and hence more work needs to be conducted to define the rate limiting targets in protein synthesis and proliferation downstream of miR-184.
Several other targets may explain this phenotype.
S6K2, a member of 40S ribosomal protein S6 kinase family was also repressed by miR-184 at the mRNA and protein level. The 40S ribosomal S6 kinase is a direct substrate of mTORC1 signalling, and when activated it drives cell growth and proliferation by recruiting translational machineries and initiating protein production in cells [
64]. Even though S6K2 is homologous to S6K1, sharing 83 % identical amino acid sequences, they both display unique functions [
65,
66].
CSF1 and
LAT1 were also significantly repressed when miR-184 was overexpressed. Evidence in the literature associates both genes with promoting tumourigenesis and metastasis [
67‐
77]. LAT1 also plays an important role in transporting available amino acids into the cell for protein synthesis, which provides a possible explanation for the inability to revert the defect in protein synthesis even by the overexpression of AKT2, PRAS40 and GSK3A.
As the PI3K/AKT/mTOR axis is such a crucial signalling axis in mediating a myriad of cellular functions essential in both normal development and during carcinogenesis, many research groups have focussed on elucidating the convoluted regulation of these pathways. In recent years, studies have revealed that there is a landscape of microRNAs that specifically regulate various components of the PI3K/AKT/mTOR signalling axis in orchestrating a series of fundamental cellular processes in normal development; including stem cell expansion [
78], wound healing [
79] and smooth muscle and pancreatic beta cell proliferation [
80,
81].
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
AS and YP conceived, designed and interpreted experiments and drafted the manuscript. DEJ, IN, SO’T, SJC, YP, AN, JY, AM, RN, EZ, ST and HKM conceived and/or conducted and interpreted experiments and contributed to writing and review of the manuscript. DR, BE, RR, EZ, WK, ND and MC conceived and conducted bioinformatics analysis and contributed to writing and review of the manuscript. All authors read and approved the paper before submission and agree to be accountable for all aspects of the work.