MicroRNA profiling of the pubertal mouse mammary gland identifies miR-184 as a candidate breast tumour suppressor gene

We performed miRNA expression profiling on micro-dissected stroma, mature ducts and TEBs of pubertal 5-week-old GFP+ mice to identify miRNAs involved in mammary gland development (Fig. 1a). Each cellular fraction expressed a set of unique microRNAs (Fig. 1b). We identified a set of miRNAs that were specific to the stroma, ducts, TEBs or both epithelial fractions (Fig. 1b; Table 1). miR-31 was the most highly enriched miRNA expressed in the TEBs, approximately 10.7-fold upregulated versus ducts. When we performed unsupervised hierarchical clustering (Fig. 1b), and miR-31 was tightly clustered to several members of the proto-oncogenic miR 17–92 cluster, such as miR-17, miR-18a and miR-19a. Conversely, miR-184 expression was significantly enriched approximately 4.2-fold in the mature ducts compared to the TEBs. Interestingly, miR-184, being the most highly enriched miRNA in the mature ducts was clustered tightly to a subset of epithelial specific miRNAs, which included members of the miR-183 family (miR-183, miR-96) and all members of the miR-200 family (miR-141, miR-200a, miR-200b, miR-200c and miR-429). We validated the expression of miR-31, miR-184, miR-17 and miR-19a in TEBs and ducts by quantitative RT-PCR (Fig. 1c).

To examine whether microRNAs differentially regulated through morphogenesis are also deregulated in cancer, we examined miRNA expression in four mouse tumour models compared to normal total mammary epithelium: spontaneous Tp53 ^−/− tumours [34], transgenic C3 SV40 tag model [35], transgenic MMTV-Neu [36] and transgenic MMTV-PyMT [37]. There are several advantages of using these tumour models: first, each tumour model has been well characterised and the initiating oncogene that drives tumourigenesis in these models is known. Second, each tumour model has its matched normal counterpart as a comparison. Last, these tumour models recapitulate different subtypes of human breast cancer.

miR-31 was highly expressed in all of the murine tumours when compared to the normal murine mammary epithelial cells (Additional file 1: Figure S1D). In contrast, the expression of miR-184 was almost abrogated in the Tp53 ^−/− tumours and the MMTV-Neu tumours, however, miR-184 expression in the C3 SV40 Tag and MMTV-PyMT tumours was comparable to normal mammary epithelial cells suggesting that miR-184 might be specifically silenced in certain breast cancers or that they derive from different cells of origin (Fig. 2a). In addition, we also examined miR-17 and miR-19a, in these murine tumours. Despite the known role of this microRNA cluster in oncogenesis, there was no enrichment of these two miRNAs in murine tumours when compared to the normal mammary epithelial cells (Additional file 1: Figure S1E, F).

We interrogated the relative expression of miR-31, miR-184, miR-17 and miR-19a in a panel of human breast cancer cell lines. Seven basal breast cancer cell lines comprising both basal A and basal B and five luminal cell lines were chosen to represent these breast cancer subtypes [38]. The expression of miR-31 was restricted to the basal breast cancer cell lines compared to the luminal cell lines (Additional file 1: Figure S1A). miR-17 and miR-19a were relatively evenly expressed across cell lines (Additional file 1: Figure S1B, C). Conversely, the expression of miR-184 was undetectable in a majority of cell lines with the exception of MDA-MB-175 (Fig. 2b).

To elucidate its function in cancer, miR-184 was acutely overexpressed in highly proliferative human breast cancer cell lines: MDA-MB-231, BT-549, MDA-MB-436 and HS578T, which express undetectable levels of endogenous levels of miR-184. Let-7a is a well-characterised tumour suppressor miRNA and has been shown to impede proliferation in cancer cell lines, and hence, was used as a positive control in this experiment. The exogenous overexpression of miR-184 resulted in levels comparable to the endogenous levels of miR-184 detected in the MDA-MB-175 (not shown). miR-184 inhibited cell proliferation significantly in all four models (Fig. 2c, d, Additional file 1: Figure S1G). The positive control let-7a also exhibited an anti-proliferative effect, which was of a similar magnitude to the result obtained with miR-184 (Fig. 2c, d).

Given the increase in expression of miR-184 as mammary epithelia differentiated in vivo, we asked whether ectopic miR-184 expression regulates the self-renewal capacity of breast cancer cells. Tumoursphere assays [39] were performed in MDA-MB-231 cells constitutively overexpressing miR-184. This assay requires cells to be in cultured in suspension for an extended period, therefore, miR-184 was overexpressed via retroviral transduction. MDA-MB-231 cells overexpressing miR-184 had a 50 % reduction in tumoursphere-forming potential in primary tumourspheres as well as in secondary tumourspheres when compared to the negative control (Fig. 2e).

After ascertaining the inhibitory effect of miR-184 on proliferation and self-renewal in vitro, we wanted to examine the role of miR-184 in vivo. Control and miR-184 overexpressing MDA-MB-231 cells were separately injected into the mammary fat pad of cohorts of immunocompromised NOD/SCID mice (n = 5). At 10 weeks post transplantation, primary tumours and internal organs were harvested. All the mice in the control group developed solid tumours, however, in stark contrast, none of the mice within the miR-184 cohort had developed any palpable tumours at this time point (Fig. 2f). Under detailed visual examination of the miR-184 cohort by fluorescent microscopy, a small population of GFP+ cells could be detected in the mammary gland. H&E staining identified a bolus of miR-184 overexpressing cells localised to the injection site (Fig. 2g). These results suggested that cells overexpressing miR-184 have impaired proliferation in vivo.

We performed immunohistochemical staining on the tumour sections for phospho-histone H3 and scoring for mitotic figures, both of which measure actively dividing cells. A significantly lower proportion of cells overexpressing miR-184 were positive for phospho-histone H3 positive cells and mitotic figures compared to the control group (Fig. 2h), confirming the suppression of proliferation by miR-184.

We conducted a survival experiment to ascertain if miR-184 expression extended survival. Mice were transplanted with MDA-MB-231 cells expressing miR-184, Let-7a or negative control and aged to ethical endpoint. The control group developed palpable tumours as early as 24 days post transplantation, and these tumours propagated at a faster rate when compared to the miR-184 group (Fig. 3a). In contrast, there was delayed latency in tumour growth in the miR-184 group where mean time to tumour palpation was by day 38. This delay in tumour initiation was also observed in the let-7a cohort and translated to an increase in overall survival. This result suggests that miR-184 impedes tumour initiation and growth.

Lastly, we also examined if miR-184 could reduce metastatic burden in vivo. Visual examination and quantification by fluorescence microscopy at ethical endpoint was used to identify macroscopic metastases in the lung and pancreas, the most frequent sites of metastasis in this model. The majority of control mice developed multiple metastases in the lungs and pancreas. In contrast, only approximately 20 % of mice within the miR-184 cohort developed any metastatic lesions in the lung (p = 0.06) (Fig. 3b). In addition, we also observed a similar reduction in metastatic burden in the pancreas, where only 10 % of mice from the miR-184 cohort developed any pancreatic metastases (Fig. 3c).

Next, we aimed to identify the repertoire of miR-184 targets to explain the impact of miR-184 on proliferation. We capitalised on the evidence that miRNAs can destabilise the mRNA of their targets to use a gene-expression-based approach to target identification [40]. We transfected miR-184 or control microRNA mimics into MDA-MB-231 breast cancer cells and analysed global gene expression changes compared to controls by affymetrix gene arrays.

We observed 1,263 mRNAs and 592 mRNAs that were significantly upregulated and downregulated respectively in cells overexpressing miR-184 compared with the negative control, suggesting dramatic remodelling of gene expression by miR-184 expression. From the profiling data, we filtered and identified the top miR-184 repressed mRNA targets based on their fold change (Table 2). AKT1S1, more commonly termed PRAS40 was the most significantly downregulated gene with approximately 3.1-fold change, followed by LAT1, which was repressed approximately 2.7-fold. To identify direct targets of miR-184 from the profiling data, we adopted a comprehensive seed match analysis method previously described by Melton et al. [30], which interrogates the promoter, 5′ UTR, CDS and 3′ UTRs of transcripts downregulated following microRNA overexpression for the presence of miR-184 seed matches. The analysis revealed highly significant enrichment of miR-184 seed matches within the 5′ UTR and 3′ UTR regions of the downregulated transcripts, suggesting that miR-184 modulates its targets by targeting the UTRs (Fig. 4a). We did not observe any significant enrichment of miR-184 seed regions within the upregulated genes, suggesting that this method identifies bone fide direct targets.

Of the 557 downregulated genes, 193 of the genes possessed a miR-184 seed match in the UTR; of these 193 genes, 135 of them had a miR-184 seed match only in the 3′ UTR, 30 genes contained a miR-184 seed match only in the 5′ UTR and 23 genes possessed a miR-184 seed region in both the 3′ and 5′ UTR. The remaining genes contained miR-184 binding sites in either the CDS and/or UTRs. We asked whether the number of miR-184 seed matches within the downregulated genes correlated with a greater degree of repression. Interestingly, in contrast to previous studies [41, 42], we observed no additive or synergistic effect on mRNA destabilisation when multiple seed matches were detected (Additional file 2: Figure S2A). We also saw no added enrichment in repression when there was multiple miR-184 seed match regions located only in the 3′ UTR (Additional file 2: Figure S2B) or the 5′ UTR (Additional file 2: Figure S2C) or in both the 3′ and 5′ UTR (Additional file 2: Figure S2D). Genes with miR-184 seed match in their 3′ UTR were also significantly more downregulated compared to those with a miR-184 seed match in the 5′ UTR (Fig. 4c). These data indicate that miR-184 represses many of its mRNA targets by targeting the 3′ UTR and that the number of miR-184 binding sites within the 3′ UTR does not correlate with degree of mRNA destabilisation.

Our analysis of putative direct targets identified at least 158 targets that are repressed and contain a 3′ UTR seed match. To look for functional relationships between these putative targets, we applied gene set enrichment analysis (GSEA) [43]. There was an enrichment of genes involved in oxidative stress, PI3K/AKT signalling, apoptosis via NFKB and axon repulsion (Additional file 2: Figure S2E). We focused on the PI3K/AKT pathway for two reasons. First, AKT2 has been shown to be an miR-184 direct target in neuroblastoma [44]. Second, AKT signalling is often dysregulated in breast cancer and AKT2 has been identified as a driver gene in mammary tumourigenesis [45].

Within the PI3K/AKT gene list, there was core enrichment for several genes from our profiling data (marked as yes under the core enrichment column), suggesting that the differential expression of these genes was significant (Additional file 3: Table S1). AKT2, PRAS40 and GSK3A were immunoblotted to validate the microarray result. There was a marked reduction in the total protein expression of AKT2, PRAS40 and GSK3A when miR-184 was overexpressed (Fig. 4d). In order to ascertain whether these genes were targets of miR-184, HEK293T cells were transfected with miR-184 or negative control in combination with 3′ UTR luciferase reporters for CSF1, GSK3A, AKT2 or PRAS40 in addition to ITGB1 (used as a negative control) and a construct containing perfect matches to the miR-184 sequence as a positive control (PMR). There was no repression in the luciferase activity of ITGB1 and a near complete ablation in luciferase activity of the PMR. We observed robust repression (>50 %) in the luciferase activity of reporters carrying the 3′ UTR from CSF1, GSK3A, AKT2 and PRAS40 signifying that miR-184 expression can act via the 3′ UTR to destabilise the mRNA of these targets (Fig. 4e).

AKT is a central node for orchestrating myriad pro-survival and proliferative pathways in oncogenesis [46]. One such downstream effector pathway is the mTOR signalling cascade. The mTOR pathway consists of two mTOR complexes, mTORC1 and mTORC2. The mTORC1 complex comprises RAPTOR, mLST8 and PRAS40 and when activated, initiates cell growth, proliferation and protein synthesis by regulating S6K1 and 4E-BP1 [47]. From the GSEA, there was enrichment for targets genes such as TSC2, RPS6KB2, implicated in mTORC1 mediated protein synthesis in cells overexpressing miR-184, and hence, we asked if miR-184 could be an important regulator of AKT/mTORC1 protein synthesis pathway.

We transfected MDA-MB-231 cells with miR-184 mimics, serum starved them and stimulated the cells with EGF, which is known to initiate a cascade of signalling events to promote cell proliferation, growth and survival. As expected, there was a marked decrease in total AKT2 and PRAS40 expression in cells overexpressing miR-184 and no change in the total protein expression of AKT1 and AKT3. Interestingly, miR-184 also decreased the total protein expression of mTOR (Additional file 4: Figure S3A). In control cells treated with EGF, we observed activation of AKT as both Ser473 and Thr308 were phosphorylated (Fig. 5a). In return, activated AKT inhibited the activities of TSC2 and PRAS40 through phosphorylating them at Thr1462 and Thr246 respectively (Fig. 5a), thus initiating the activation of mTORC1. Conversely when cells overexpressing miR-184 were treated with EGF, we observed a modest increase in p-AKT Thr308, indicating that there was increased AKT activity. Despite the increased AKT activity, the inhibition on TSC2 was completely abrogated, as observed from the dephosphorylation of Thr1462, possibly by miR-184 targeting AKT2. Finally the activating phosphorylation of S6K1 at Threonine-389, a rate-limiting factor in protein synthesis, was attenuated without affecting the activity of 4E-BP1 (Additional file 4: Figure S3A).

We also further examined if this miR-184-regulated signalling cascade was recapitulated in vivo. Within the cohort of mice bearing miR-184-overexpressing tumour cells, the expressions of a number of miR-184 targets (e.g., AKT2, GSK3A and S6K2) were reduced in three out of four mice when compared to the control tumours. Furthermore, we also observed a dramatic repression of mTOR expression accompanied by the dephosphorylation of S6K1 indicating a partial attenuation of the protein synthesis pathway (Additional file 4: Figure S3B).

To directly assay the impact of EGF stimulation and miR-184 expression on protein synthesis, we transfected MDA-MB-231 cells with the miR-184 mimics, and measured the amount of protein synthesised in the cell by incorporating radiolabelled tritiated (H³) leucine into cells in the presence of EGF. miR-184 repressed global protein synthesis dramatically (Fig. 5b). In order to demonstrate that this regulatory phenomenon exhibited by miR-184 was not restricted to the MDA-MB-231 model, we performed the same assay with the HEK293E cells using another growth factor, insulin. We detected that miR-184 similarly repressed global protein synthesis by approximately 30 % in the HEK293E cells stimulated with insulin (Fig. 5c).

We were interested to see if we could reverse the repression on protein synthesis by overexpressing some of the core enriched top candidate genes identified from the GSEA. Therefore we tested if direct targets AKT2, PRAS40 or GSK3A were responsible for the effect of miR-184 on MDA-MB-231 cell protein synthesis by stably overexpressing AKT2, PRAS40 and GSK3A individually or in combination, together with transfection of miR-184 or negative control mimics. Despite the overexpression of these direct target genes individually (Additional file 5: Figure S4A, B, C) or in combination (Fig. 5d, e), protein synthesis was nonetheless suppressed by miR-184 expression in these cell lines. These results suggest that miR-184 regulates the protein synthesis process by modulating expression of genes in addition to AKT2, PRAS40 and GSK3A.

We next investigated the significance of miR-184 expression in breast cancer and its association with clinico-pathological measures. We measured miR-184 expression in snap-frozen primary tumour specimens comprising patient samples diagnosed with luminal, HER2-amplified and triple negative cancers and matched normal tissue. We observed unique expression patterns for miR-184 across these subtypes. There were no significant differences in the expression of miR-184 between luminal cancers and matched normal. However within the HER2 subtype, the patients fell into two groups: normal-like and high miR-184 expression. Last, miR-184 expression was significantly lower in the triple negative subtype compared to the matched normal (Fig. 6a). To verify this result in a bigger patient cohort, we analysed miR-184 expression in the METABRIC cohort of 980 breast cancers [31] and also observed that miR-184 mean expression was highest in HER2-positive breast cancers and significantly lower in basal breast cancers, a subset of TNBC (Fig. 6b). In this cohort, miR-184 expression did not correlate with prognosis (data not shown). However, elevated expression of a signature composed of high-confidence direct targets of miR-184 predicted poor overall survival (Fig. 6c). This effect was observed in two further independent cohorts (Additional file 6: Figure S5). In multivariate analysis, this signature predicted poor prognosis independent of ER status in two out of three cohorts (Additional file 7: Table S2). These data are consistent with a role for miR-184 in suppressing proliferation or metastatic dissemination.

There is emerging evidence for epigenetic silencing by hypermethylation [48] of miRNAs with tumour suppressor properties or growth inhibitory functions in various malignancies [49‐52]. Methyl-capture sequencing was performed to capture a global snapshot of the genome-wide DNA methylation profile of normal mammary tissue, primary TNBC tumours and matched metastatic tumours micro-dissected from the lymph nodes of patients. We observed that there was minimal methylation detected at the miR-184 locus in normal breast tissue. In contrast, there was a pronounced increase in methylation in metastatic tumours in the lymph nodes of three of eight patients (Fig. 6d), suggesting a selective pressure to silence miR-184 during metastatic dissemination, consistent with the capacity of miR-184 to suppress metastasis in animal models (Fig. 3).