Introduction
The ductal epithelial networks that characterize mouse and human mammary tissue appear to comprise an analogous cellular hierarchy: multi-potent mammary stem cells (MaSCs) reside at the apex of the hierarchy and are capable of differentiation along the myoepithelial/basal lineage or the luminal lineage to yield mature ductal and alveolar cells [
1‐
6]. The precise nature of the intermediate cell types remains unclear but two or three distinct luminal progenitor subsets have been prospectively isolated from mouse and human mammary tissue, respectively [
4,
7‐
9]. Several functional studies have used a candidate approach to identify regulators of self-renewal, lineage commitment and differentiation programs (reviewed in [
10]). Furthermore, genome-wide transcriptome analyses [
11,
12] of mouse mammary epithelial subsets have identified a number of potential regulators of mammary gland development. The definition of numerous conserved pathways across species has highlighted those that are likely to be involved in cell-fate decisions and lineage differentiation [
12]. Moreover, the epigenome has been implicated in playing a critical role in regulating such decisions within the epithelial compartment of the normal mammary gland [
13,
14].
There is increasing evidence that microRNAs (miRNAs) regulate a wide range of biological processes, including maintenance of cell identity, differentiation and apoptosis [
15‐
17]. miRNAs, small non-coding RNA molecules that inhibit translation or trigger mRNA decay [
15,
17], have been implicated in both mammary gland development and breast tumorigenesis. In a large-scale study, the expression of 318 miRNAs was assessed during different stages of development, leading to the observation that miRNAs can be expressed in coordinated clusters, and that global miRNA and mRNA expression are significantly lower in lactation and early involution [
18]. In the mouse mammary epithelial cell line, Comma-Dβ [
19], the expression of miR-205 and miR-22 but not let-7 and miR-93 was linked to progenitor-like properties, while miR-200c appears to function within the basal cell compartment of normal breast tissue [
20]. Interestingly, miR-200c targets the mRNA encoding BMI1, a key regulator of the self-renewal of stem cells in multiple tissues. MiR-193b also has been implicated in regulating mammary stem cell activity in vivo and may serve an additional function in controlling the alveolar differentiation during pregnancy [
21]. In the context of breast cancer, many miRNAs have been reported to undergo deregulation, inferring an important role in controlling proliferation versus differentiation decisions. For example, miR-205 is one the most significantly downregulated miRNAs in human breast cancer relative to normal tissue [
22]. Moreover, miRNA signatures that distinguish breast tumors of different subtypes from normal tissue have been described [
23]. To understand the consequences of deregulated miRNA networks, it is essential to characterize the normal expression patterns and roles of miRNAs in the epithelial differentiation hierarchy. Here we sought to determine the global miRNA expression profiles of discrete cellular subpopulations within normal mouse and human mammary tissue. Comparative analyses of miRNA signatures with gene expression and histone modification profiles of the epithelial subsets revealed candidate miRNAs that are likely to execute important roles in mammary epithelial specification and differentiation.
Discussion
This study describes genome-wide miRNA expression profiling of four distinct mouse and human subpopulations that are highly enriched for MaSC/basal, luminal progenitor, mature luminal and stromal cells. The four subpopulations exhibited distinct miRNA signatures that were conserved across species. Evaluation of potential target genes revealed that the top differentially expressed miRNAs likely target lineage-specific mRNAs. The predicted miRNA–mRNA relationships were found to be highly conserved between mouse and human epithelial subtypes. Around 58 % (25 miRNAs) of the top negatively correlated miRNAs in the mouse mammary gland (Additional file
8: Table S8) were conserved in human; these miRNAs are likely to govern important mammary cell fate and differentiation decisions during ontogeny. The top negatively correlated (conserved) mouse miRNAs include miR-30a/d (targets
Runx2) [
57], miR-148a (targets
Met/Snail) [
58], miR-503 (targets
Bcl-2 and
Igf1r, implicated in involution) [
59], miR-203 (targets the transcription factor
p63) [
60] and miR-34a (targets
Dll1 and
CD44, important for stem cell activity) [
61,
62].
There is accumulating evidence that the Wnt and Notch pathways, as well as the Polycomb repressor complex of proteins, play prominent roles in regulating MaSC function [
14,
63‐
65]. In the context of miRNAs that potentially control these pathways, we identified several luminal-restricted miRNAs, including miR-10a, miR-200a/b, miR-203, miR-148a. Conversely, miR-146a, miR-221/222, and miR-205, which have been shown to regulate genes expressed in the ductal and alveolar luminal lineages (e.g.,
Brca1, Gata3, c-kit and
Elf5), were restricted to the MaSC/basal population.
Intriguingly, the primate-specific miRNA cluster (C19MC miRNAs) on chromosome 19 at q13.4 was highly expressed in MaSC/basal cells. Moreover, miR-512 has been implicated in targeting the pro-survival gene
MCL-1 that is expressed at very low levels in this subset [
66]. Our Targetscan analysis further identified the luminal-specific genes
RANK, NOTCH3, ELF5, ESR1, HEY2 and
KIT as potential targets of these primate-specific miRNAs (Fig.
3c). Notably, structural rearrangements of the chromosomal 19q13 region that occur in some thyroid adenomas and adenomatous goiters are associated with aberrant expression of miRNAs in this cluster. In addition, miR-517c and miR-591a are highly expressed in the basal-like subtype of breast cancer [
55], further implicating C19MC miRNAs in carcinogenesis.
It has been presumed that the expression of miRNAs and their host genes largely coincide. However, the expression of miRNAs located within introns or the coding regions of specific genes may be independent of host gene expression and its epigenetic modifications. For example, expression of the BTG4 gene, which harbors the MaSC/basal-specific miRNAs miR-34b and miR-34c, and the TRP3 gene that encompasses miR-204, is not detectable in mammary epithelium. In addition, the MIB-1 gene, which is host to the MaSC/basal-specific miRNAs miR-1 and miR-133a, is expressed at very low levels in all three epithelial subsets (data not shown).
The epigenetic landscape of mammary epithelial cells appears to play an important role in the progressive commitment of MaSC/basal cells to differentiated cells. Not only does the epigenome contribute to gene expression changes [
14], but it tightly correlates with the miRNA expression profiles of the different mouse mammary epithelial subsets. Despite a paucity of information on the TSS of miRNAs, a clear pattern has emerged for histone methylation marks on DE miRNAs: the top DE miRNAs repressed upon restriction of MaSC/basal cells to luminal progenitor cells were enriched for H3K27me3 modifications, while those activated upon commitment were characterized by the presence of H3K4me3 marks. Similar epigenetic patterning held for the luminal progenitor versus mature luminal populations. Overall, the presence of H3K4me3 marks correlated tightly with the expression of both miRNAs and mRNAs, while H3K27me3 modifications negatively correlated with their expression. Moreover, the histone methylase Ezh2 was directly implicated in coordinating H3K27 trimethylation of the regulatory regions of miRNAs whose expression was repressed. Collectively, these data suggest that miRNA expression is regulated by epigenetic modifications and contributes to decisions on proliferation versus differentiation in the mammary gland. It remains to be determined whether steroid hormones also influence the epigenome of regions flanking miRNA loci. In human mammary epithelial cell lines, the expression of the miR-200 family was recently found to be subject to epigenetic regulation, whereby DNA methylation and histone modifications were altered during the transition between stem-like and nonstem states [
67]. DNA methylation of the miR-200c-141 cluster and polycomb group-mediated histone methylation of the miR-200b-200a-429 cluster resulted in repression at these loci [
67]. Moreover, H3K4me3 was found to be associated with active miRNAs in colorectal cancer cell lines, whereas hypermethylation of promoter CpG islands caused epigenetic silencing of miR-124 and mir-34b/c [
68‐
71].
Comparison of miRNA signatures derived for distinct mammary epithelial subsets from normal mammary tissue with those of different breast cancer subtypes further strengthened the molecular links that have been previously defined at the mRNA level. Specifically, the miRNA signature of the luminal progenitor population was most concordant with the basal-like cancer subtype, the mature luminal cell-enriched population was closest to the luminal B subtype, and the signature of the MaSC/basal population was highest in the normal-like subtype of cancer. These findings suggest that defined cell types in normal breast tissue may be predisposed to acquiring oncogenic events that result in specific types of cancer. Notably, there was a strong correlation between the miRNA signatures of the luminal progenitor cell and the basal-like subtype of cancer, also reflected in their corresponding transcriptomes. This cell is the likely ‘cell of origin’ for basal-like cancers that arise in BRCA1 mutation carriers [
2].
Several highly expressed miRNAs have been associated with the development and progression of breast cancer, in which their aberrant expression is presumed to destabilize mRNAs encoding crucial tumor suppressors and differentiation-promoting factors [
72,
73]. Profiling studies of primary breast tumors have revealed differential miRNA expression according to estrogen receptor (ER)/progesterone receptor (PR) or human epidermal growth factor receptor 2 (HER2) status and different tumor stages [
23,
55,
74,
75]. More specifically, the expression of some miRNAs has been linked to histopathological features such as HER2/
neu or ER/PR status (miR-30), metastasis (miR-126 and miR-335) and the EMT (miR-205 and miR-200 family) [
43,
76‐
79]. The luminal subtypes of breast cancer appear to have elevated expression of miR-190b, while basal-like tumors have higher levels of miR-18a/b, miR-9 and the miR-17-92 family and lower levels of miR-29 and miR-190b [
55]. The higher levels of miR-18a/b, miR-9 and miR-17-92 in the MaSC/basal population suggest that a subset of triple negative cancers may harbor an expression signature that more closely resembles that of the stem cell population. Furthermore, the primate-specific, basal-restricted miR-516a and miR-519a were most highly expressed in this subtype of breast cancer (Fig.
4b). Other miRNAs recently implicated in breast cancer include miR-100, shown to target SMARCA5, SMARCD1, and BMPR2 genes, which directly influence tumor cell proliferation [
80], and miR-30c, known to target TWF1 and IL-11 [
81], both of which are expressed in the MaSC/basal lineage. Ultimately, a comprehensive analysis of miRNAs deregulated in breast cancer, together with an understanding of their transcriptional and epigenetic control, may provide novel prognostic or therapeutic tools for breast cancer.
Acknowledgements
We are very grateful to C Perou and K Hoadley for supplying PAM50 subtypes for the TCGA breast cancer samples, and F Vaillant for expert advice on sorting. Human breast tissue samples used in this project were provided by the Victorian Cancer Biobank, which is supported by the Victorian Cancer Agency (VCA). This work was supported by the Australian National Health and Medical Research Council (NHMRC) grants #461221, #490037, #1008440, #1016701, #1054618; NHMRC IRIISS; the Victorian State Government through VCA funding of the Victorian Breast Cancer Research Consortium and Operational Infrastructure Support; the Australian Cancer Research Foundation; Cancer Council SA Project Grant 626956. BP was supported by a NHMRC Peter Doherty Fellowship and VCA Early Career Seed Grant (#ECSG13035), GKS, GJG and GJL by NHMRC Fellowships (#1058892, #1026191 and #1078730, respectively) and JEV by an Australia Fellowship.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
BP designed and carried out the molecular studies, and helped draft the manuscript. YC, YH and GKS carried out the bioinformatic analyses and helped draft the manuscript. AB and GJG carried out the miRNA profiling and contributed to the analysis and interpretation of the data. JMS, TB and PJ performed cell fractionation studies and data analysis. JEV and GJL contributed to design of the study, interpretation of data and writing of the manuscript. All authors read and approved the final manuscript.