In the cytoplasm, mature miRNAs are incorporated into the RNA-induced silencing complex (RISC) complex. The miRNA-RISC complex then mediates gene silencing by targeting the 3′ untranslated region (UTR) of mRNAs and induces translational silencing, either by cleaving the mRNA or blocking ribosomal translation of the targeted mRNA. Target mRNAs are recognized by partial sequence similarity in the seed sequence of the mature miRNA [6] and often these 3′UTR target sites are evolutionarily conserved within the miRNA seed region [2, 7]. Once the target mRNA is recognized by the miRNA-RISC complex, it is unclear what mechanisms mediate mRNA repression: mRNA degradation in the P-bodies or inhibition of translation by blocking ribosomes [8–10].

Many tissues, including the skin, blood and gut, contain a small population of lineage-restricted stem cells to maintain the short lived mature cells (reviewed in [25]). While mostly quiescent, these multipotent adult stem cells have the ability to self-renew and are essential for tissue homeostasis because of their ability to differentiate into the cells that make up a tissue. The existence of mammary gland stem cells was first suggested by studies performed over half a century ago in which small fragments of donor mammary tissue were transplanted into the cleared fat pad of recipient mice [26, 27]. These seminal transplantation studies demonstrated that not only could small populations of cells repopulate a full mammary gland, consisting of all the differentiated cell types of the mammary ductal tree, but also that these cells could be serially transplanted and still retain this ability [26, 28].

Identifying markers of normal mammary gland stem cells is essential for their isolation and further functional studies. Studies from two laboratories recently identified such markers within the mouse mammary gland [29, 30]. Stingl and colleagues isolated two subpopulations from the adult mammary gland that exhibited different progenitor-like behavior based on CD49f (α6-integrin)/CD24 (heat stable antigen) profiling. Single cells from the CD49f^hi/CD24⁺ subpopulation from an adult mammary gland were able to regenerate a functional mammary gland in vivo, while the CD49f^low/CD24^hi subpopulation displayed clonogenic potential in vitro [30]. In related studies, the Lineage(Lin)⁻/CD29(β1-integrin)^hi/CD24⁺ subpopulation from the adult mammary gland also was shown to be capable of repopulating the mammary gland [29]. It has been estimated that the stem cell activity within the CD49f^hi/CD24⁺ or Lin⁻/CD29^hi/CD24⁺ subpopulations represents only about 1 in 2000 cells, although this estimate is approximately ten-fold higher than the stem cell frequency by limiting dilution transplantation of unsorted mammary cells [31]. In humans, studies have shown that cells with high aldehyde dehydrogenase (ALDH1) activity [32] or a CD49f⁺/EpCAM⁻ immunnotype have stem cell properties [33].

Like normal mammary gland stem cells, putative breast cancer stem cells have been isolated based on cell surface markers. Using the normal breast stem cell profile of Lin⁻/CD29^hi/CD24⁺, breast cancer stem cells were identified in a p53^null genetically engineered mouse model of breast cancer [54]. Serial and limiting dilution transplantation of the tumor cells isolated by Lin⁻/CD29^hi/CD24⁺ profiling indicated that they were enriched 65-fold in tumor-initiating frequency as compared with the majority of the tumor cells isolated using the other three surface marker profiles. In addition, they had characteristic stem cell properties defined by in vitro assays. Furthermore, the four CD29/CD24 subtypes could be separated based on a 710 mRNA gene signature. Since tumor types can be classified based on miRNA expression patterns [55, 56], it would be expected that cancer stem cells within breast cancers also should have unique miRNA expression profiles.

Originally isolated in 2003, human breast cancer stem cells were identified by FACS sorting for expression of a different set of cell surface markers, identifying the cancer stem cell population as CD44⁺/CD24^−/low [57]. Studies on the CD44⁺/CD24^−/low population in human breast cancer have shown decreased expression of the let-7 family (let-7a to let-7i) compared to differentiated cells. Additionally, induction of these cells to express let-7 was shown to inhibit tumor formation and reduce the metastatic capability of the CD44⁺/CD24^−/low cancer stem cells. The tumor-suppressor activity of let-7 was shown to act via regulation of the oncogenic H-Ras and HMGA2 (High mobility group AT-hook 2) genes (Fig. 1). Thus, tumors expressing low let-7 had higher levels of H-Ras and HMGA2, which correlated with poor prognosis. Conversely, tumors expressing high levels of let-7 were less tumorigenic and had lower levels of metastasis [58].

The long non-coding RNA B-cell Integration Cluster (BIC) was originally identified as a proto-oncogene in B-cell lymphomas because it was a common integration site for the avian leukosis virus [66]. It was later found that BIC encodes for a miRNA, miR-155, that is required for proper B and T cell immunity and for the function of antigen-presenting cells [67]. By miRNA expression profiling, miR-155 was shown to be highly expressed in breast cancers [56], as well as other solid tumors and lymphomas [68] and is one of the miRNAs that defines a 29-miRNA breast cancer-specific signature (including down-regulation of miR-10b, miR-125b and miR-145 and upregulation of miR-21 and miR-155) [44]. In breast cancer cell lines, over-expression of miR-155 led to increased growth and colony formation, increased growth in a cell transplantation assay, and miR-155 was shown to act as an oncomiR by targeting the tumor suppressor gene suppressor of cytokine signaling 1 (SOCS1), a negative regulator of the JAK/STAT pathway[69].

Conflicting reports of miR-205 expression within breast cancer subtypes illustrates the complexity of understanding miRNA function. Studies indicate that miR-205 is highly expressed in normal mammary epithelial cells, and enriched in mouse mammary stem cell populations [35], and it is also developmentally regulated [90]. In prostate and breast cancer cell lines over-expression of miR-205 lead to a reduction of growth, possibly via regulation of the oncogenes HER3, E2F1, E2F5, ZEB1/ZEB2 and PKCe [35, 60, 80, 81]. High expression of miR-205 has been observed in ER+/PR+ HER2+ breast cancers, compared to other subtypes [45] and genomic analysis indicated that miR-205 is within a genomically amplified region in some human breast cancers [42]. Other reports indicate low expression of miR-205 in human breast cancers compared to normal tissue [56], however these studies did not indicate the specific breast cancer subtype(s) analyzed, so loss of miR-205 expression in these samples should not be considered a universal breast cancer feature. Loss of miR-205 expression in breast cancer may be a late event during breast cancer progression. In fact, invasive breast cancer cell lines have lower expression of miR-205 than non-invasive breast cancer cell lines [60], and human metastatic breast cancer biopsies had low expression of miR-205 compared to non-metastatic lesions [44]. This is also supported by recent studies demonstrating that miR-205 is a negative regulator of EMT, a process required for vascular invasion and metastasis, by means of silencing ZEB2/SIP1 and ZEB1, repressors of E-cadherin and regulators of EMT [60]. Thus, over-expression of miR-205 may lead to different phenotypes depending on cell type.

The process of EMT has been associated with tumor cell invasion and metastasis. As mentioned above, the miR-200 family, as well as miR-205, regulate EMT through targeting Zeb1 and Zeb2. While miR-205 was shown to suppress cell growth in MCF7 cells, it was shown to inhibit invasion and metastasis in MDA-MB-231 cells [81]. Most of these studies have used in vitro assays to show that miR-200 family members can suppress the invasive properties of cells in culture [60–62]. Unexpectedly, over-expression of miR-200 in the non-metastatic murine mammary tumor cell line 4TO7 enables these cells to metastasize to the lung and liver [99]. This may be due to the fact that some tumors may have the requirement of a mesenchymal to epithelial transition (MET) occurring for efficient tumor colonization at the metastatic site. Further studies using in vivo metastasis models are surely warranted to elucidate the role of the miR-200 family in metastasis. Furthermore, miR-200c targets class III β-tubulin (TUBB3), and restoration of miR-200c expression in cells with low miR-200 family levels results in increased chemosensitivity to microtubule-directed agents [100]. As a pro-metastatic miRNA, miR-9 targets E-cadherin, loss of which results in loss of epithelial characteristics, cell migration and invasion, ultimately leading to an EMT [101]. Further studies on breast cancer cells demonstrated miR-9 is directly regulated by MYC and MYCN, and over-expression of miR-9 resulted in an EMT, increased proliferation, enhanced metastasis, and induced angiogenesis when transplanted [101].

Breast cancer metastasis suppressor 1 (BRMS1) is a protein that suppresses metastasis in multiple tumor types. By analyzing the miRNA expression profiles of metastatic MDA-MB-231 and MDA-MB-435 cells compared to BRMS1-transduced non-metastatic counterparts, Welch and colleagues found that BRMS1 coordinately regulates a number of these metastamirs [102]. BRMS1 decreased metastasis-promoting miRNAs: miR-10b, miR-373, and miR-520c. Metastasis suppressing miRNAs: miR-146a, miR-146b, and miR-335 were up-regulated following BRMS1 overexpression. This shows that BRMS1-containing SIN3/HDAC complexes may be recruited to and regulate miRNA promoters in addition to coding genes involved in regulating cancer metastasis.

miR-10b, which is highly expressed in metastatic breast cancer, is also involved in breast cancer metastasis [103]. This control is due to a feedback loop where the transcription factor Twist regulates transcription of miR-10b via direct E-box binding, which in turn represses translation of HOXD10 via a highly conserved miR-10b binding site in the 3′UTR. Decreased levels of HOXD10 then resulted in increased expression the pro-metastatic gene, RHOC, although a direct connection here has not been established. miR-10b has also been shown to directly repress Tiam1 expression in breast carcinoma cells [104]. This inhibits Tiam1-mediated Rac activation suppressing migration and invasion. Furthermore, over-expression of miR-10b in non-metastatic breast cancer cells promoted metastasis, while inhibition of miR-10b in metastatic breast cancers decreased metastatic phenotypes of cells in vitro. However, systemic administration of miR-10b antagomirs to mice bearing highly metastatic 4T1 cells was not able to reduce growth of the primary lesion, but was shown to markedly suppress formation of lung metastases, representing a promising approach for an anti-metastasis agent [105].

By comparing miRNA expression profiles between MDA-MB-231 parental cells and derivatives with higher metastatic ability, Tavazoie, et al., identified miR-335 and miR-126 as low in more metastatic lines. Both miR-126 and miR-335 were significantly down-regulated in the metastatic breast cancers, and this decreased expression was correlated with poor prognosis. Induction of miR-335 expression is sufficient to suppress lung and bone metastasis in a metastatic breast cancer cell line. SOX4, which is critical in cell migration, was shown to be a functional target of miR-335, as well as the extracellular matrix component tenascin C [106]. A miR-335 six-gene target signature (SOX4, PTPRN2, MERTK, PLCB1, COL1A1 and TNC) is associated with poor metastasis-free survival in a large patient dataset. Additionally, miR-126 expression reduced overall tumor growth and proliferation. This inhibition of cell cycle progression was shown to be possibly through inhibition of Insulin Receptor Substrate-1 (IRS-1) [107]. Other studies of the metastatic MDA-MB-231 cell line indicated that miR-146 inhibits invasion and metastasis of cells by downregulating NFkB, in addition to targeting IRAK1 and TRAF6 [108].

Valastyan et al. demonstrated that miR-31, which is expressed in normal mammary cells, is specifically lost in metastatic breast cancer cell lines. Overexpression of miR-31 expression in metastatic cells can inhibit multiple steps in metastasis, and in non-metastatic cell lines, deletion of miR-31 is sufficient to confer metastatic properties. [109]. In a subsequent study, the authors demonstrate that while miR-31 is predicted to modulate the expression of >200 mRNAs, its regulation of metastasis can be accounted for primarily by three targets, integrin α5 (ITGA5), radixin (RDX), and RhoA [110]. Interestingly, each of these targets affects distinct steps in the metastatic process and re-expression of these three genes concurrently abrogates metastasis suppression by miR-31.

miR-373 and miR-520c were identified as promoters of tumor invasion and metastasis through a genetic screen using MCF7 cells subjected to a trans-well migration assay. These miRNAs can stimulate the normally non-metastatic MCF7 cell line to display metastatic behavior of cell invasion and migration, interestingly by targeting CD44 [111].

In MDA-MB-231 cells, unlike MCF7 cells, miR-21 does not affect growth of the primary tumor. It was however shown that miR-21 promotes cell invasion and lung metastasis in this model [73]. This is likely through the down-regulation of multiple targets including several other tumor suppressor genes.

miR-145 was recently shown to suppress invasion and experimental metastasis assays using metastatic breast cancer cell lines MDA-MB-231 and the LM2-4142 lung metastatic subline [112]. This was in part due to miR-145 directly targeting MUC1, which in turn down-regulates β-catenin, Cyclin D, and Cadherin 11.

Liu et al. found that the miR-17–92 cluster was expressed at higher levels in metastatic breast cancer cell lines. The expression of the cluster was diminished by treatment with a ROCK inhibitor (Y27632). They further showed that blockade of miR-17 using anti-miR-17 molecules decreased breast cancer cell invasion in vitro and lung metastasis from the orthotopic site [113]. In contrast, another group found that miR-17/20 (miR-17-5p and miR-20a) were reduced in highly invasive breast cancer cell lines and node-positive breast tumors, and in vitro miR-17/20 could inhibit migration and invasion of neighboring cells via a unique mechanism of heterotypic secreted signaling [114].

Recently, Baffa et al. performed a miRNA microarray analysis of 43 matched primary tumors (13 breast, 10 lung, 10 bladder, and 10 colon cancers) and corresponding lymph node metastasis to identify deregulated miRNAs in metastasis. They identified 32 differentially expressed miRNA [115]. These included a number of the above mentioned metastasis-regulating miRNAs including the upregulation of miR-10b and miR-21 and the down-regulation of miR-141, miR-200b, miR-200c, and miR-205 in metastatic samples.

Using a large familial study population of BRCA1/2 mutation negative cases, Tchatchou et al. analyzed the impact on breast cancer risk of 11 miRNA target site SNPs located in the 3′UTR of cancer associated genes [125]. There was a significant association with familial breast cancer risk for a variant affecting a putative miR-453 binding site in the 3′UTR of ER. The protective effect was stronger in pre-menopausal women and in high-risk familial cases. The binding of the miRNA was predicted to be stronger when the protective allele is present, potentially leading to lower levels of ER and making biological sense with the role of estrogen and ER in breast cancer progression. Nicoloso and colleagues analyzed SNPs associated with breast cancer risk for their ability to modify miRNA binding sites [126]. They identified SNPs in TGFB1 and XRCC1 that could modulate their expression by differential interaction with miR-187 and miR-138, respectively. A genome-wide bioinformatics analysis of human HapMap data predicted approximately 64% of transcribed SNPs can increase or decrease the binding energy of putative miRNA::mRNA duplexes. To assess whether SNPs affecting miRNA target sites are implicated in breast cancer susceptibility, the authors conducted a case-control population study. They observed that occurrence of miRNA target SNPs found in BRCA1 and TGFR1 significantly varied among populations with different risks of developing breast cancer. These predictions and biological effects were validated using in vitro luciferase reporter assays as well as testing the effects of overexpression of the two interacting miRNAs (miR-638 and miR-628-5p) on protein levels in cell lines with different genotypes.

Antisense oligos that bind to the mature miRNA sequence have been exploited to inhibit specific miRNAs. Antisense oligos are 17–22 nucleotides long and are designed to have the exact nucleic acid complement of the mature miRNA sequence such that the targeted miRNA will bind to the miRNA inhibitor by means of Watson-Crick base pairing, and not to its endogenous target mRNA. One of the first methods used for inhibiting specific miRNAs was the use of 2′-O-Methyl antisense oligos. The unique design of the 2′-ribose modification made the oligo resistant to cellular nucleases, such that the oligo would have an increased stability and half-life as compared to a conventional oligo, and the inhibitor-miRNA interaction was thus irreversible [129,130]. Originally used for detection of miRNA expression by in situ hybridization and Northern Blot analysis, locked nucleic acids (LNAs) were also used as miRNA inhibitors in vitro [131, 132]. Like the 2′-O-Methyl oligos, LNAs are resistant to cellular nucleases, but the conformational change design has given the LNA/RNA interaction increased stability [133]. The use of these 2′-O-Methyl and LNA antisense oligos were effective in vitro, but the need to deliver antisense oligos in vivo has lead to the design of more advanced oligos. The addition of a cholesterol modification to the 3′ end of a 2′-O-Methoxyethyl antisense oligo, now called “antagomirs”, helped facilitate the uptake of this miRNA inhibitor into cells in vivo [134]. This method allowed injection of antagomirs intravenously, and was used to target several miRNAs in many tissues, with the exception of the brain, likely due to the blood-brain barrier [134]. Although the mechanism is not fully understood, an anta-gomir-miRNA duplex will result in the degradation of the target miRNA and subsequent recycling of the antagomir, allowing it to act on more target miRNAs [135]. Therapeutically, antagomirs have been used in mouse models of human diseases and have been shown to revert phenotypes associated with over-expression of miRNAs [134, 136–139].

As mentioned previously, the oncogenic miR-21 was found to be up-regulated in human breast cancers compared to normal tissue. The use of an antisense oligo to miR-21 in a breast cancer cell culture model was successful at inhibiting tumor growth when treated cells were transplanted into recipient mice. As expected, when miR-21 was inhibited, resulting tumors had a longer latency, were proliferating slower, had increased apoptosis and had increased susceptibility to anti-cancer drugs [140].

A second more recent approach is the use of miRNA mimics to restore a miRNA in diseases where a miRNA is lost. Mimics are generally chemically synthesized artificial miRNAs, designed to be stable and less susceptible to endogenous RNases, that can be delivered systemically. In mouse models of heart disease, several miRNAs were deregulated following a myocardial infarction. In particular, miR-29 expression was significantly lower in both mouse models and human samples of myocardial infarction. Cell culture models were employed to study the effects of replacing miR-29 with a miRNA. These resulted in the reversal of cardiac fibrosis, a result of myocardial infarction; however, these studies have not yet been validated in vivo [141]. Additionally, a miRNA mimic for the well-studied tumor suppressor miRNA let-7 was used therapeutically in mouse models of lung cancer. Expression of let-7 is frequently lower in many cancers, and is thought to act as a tumor suppressor by modulating expression of multiple oncogenes, including Ras, Myc and HMGA2 [142–144], as well as cell cycle genes Cyclin D2, Cdk6 and Cdc25 [145]. Addition of a let-7miRNA mimic to lung cancer cell culture models was shown to decrease proliferation by induction of cell cycle arrest and cell death in vitro, as well as in mouse lung cancer xenograft models. Additionally, genetically engineered mouse models of lung cancer were also inhibited by the let-7 mimic [146].

Chemotherapeutic resistance in primary and metastatic breast cancer presents a major clinical challenge, as many patients will become resistant to multiple cytotoxic agents, termed mutlidrug resistance. The use of miRNA-based therapy to enhance anticancer drug activity was recently investigated in the NCI-60 human cancer cell line panel. These investigators used the NCI-60 cell line panel and miRNA mimics and inhibitors of let7i, miR-16, and miR-21, and found that altering miRNA levels could affect the efficacy of a number of the anticancer agents tested [147]. These results provide a potential for using miRNAs in a clinical setting to improve current chemotherapy strategies. Additionally, in a cisplatin-resistant MCF-7 cell culture model of human breast adenocarcinoma, several miRNAs were found to be disregulated, including miR-345 and miR-7, which both target the human multidrug resistance-associated protein 1 (MRP1) [148]. This disregulation may, in part, account for some of the acquired drug resistance in these cells.