A discontinuous pattern of LOH at chromosome 3p has been reported in 87% of primary breast cancers. Despite the identification of several tumor suppressor genes in this region, there has yet to be a detailed analysis of noncoding RNAs including miRNAs in this region. In this study, we identified 16 aberrant miRNAs in this region and determined several that are frequently lost or amplified in breast cancer. miR-128-2 was the most commonly deleted miRNA. Embedded in the intron of the ARPP21 gene at chromosome 3p22.3, miR-128-2 was frequently downregulated along with ARPP21 in breast cancer, where it was negatively associated with clinicopathologic characteristics and survival outcome. Forced expression of miR-128 impeded several oncogenic traits of mammary carcinoma cells, whereas depleting miR-128-2 expression was sufficient for oncogenic transformation and stem cell-like behaviors in immortalized nontumorigenic mammary epithelial cells, both in vitro and in vivo. miR-128-2 silencing enabled transforming capacity partly by derepressing a cohort of direct targets (BMI1, CSF1, KLF4, LIN28A, NANOG, and SNAIL), which together acted to stimulate the PI3K/AKT and STAT3 signaling pathways. We also found that miR-128-2 was directly downregulated by SNAIL and repressed by TGF-β signaling, adding 2 additional negative feedback loops to this network. In summary, we have identified a novel TGF-β/SNAIL/miR-128 axis that provides a new avenue to understand the basis for oncogenic transformation of mammary epithelial cells. Cancer Res; 72(22); 6036–50. ©2012 AACR.
Mammary epithelial oncogenic transformation and progression is a multistep process that purportedly involves varied genomic alterations and epigenetic modifications (1). These aberrations activate oncogenes, inactivate tumor suppressor genes (TSG), and fundamentally convert a normal cell into a symptomatic cancer cell clone, which is rendered with growth advantage and expands in an unregulated manner, leading to local invasion and distant metastasis (2). In the past 4 decades, significant progress has been made in the discovery of chromosomal aberrations in human breast cancer. In familial breast cancer, BRCA1 at 17q21 and BRCA2 at 13q12–13 have been isolated as breast cancer susceptibility genes, and their mutations account for the majority of hereditary breast cancer (3). LOH has been discovered on chromosomes 3p, 6q, 7p, 11q, 16q, and 17p in breast cancer, indicating TSGs may harbor in these regions (4). Of special interest, 87% of primary breast cancers exhibit a discontinuous pattern of LOH at chromosomes 3p and several genes in this region have been identified as TSGs, such as DUTT1 and FHIT (5). However, although large-scale genomic and molecular genetic technologies allow us to efficiently identify numerous breast cancer-related protein-coding genes in the abnormal chromosomal regions, detailed analyses of noncoding RNAs locating in these regions are largely not determined.
miRNAs are small, endogenous noncoding RNAs consisting of 21 to 24 nucleotides, which repress gene expression posttranscriptionally via recognizing complementary target sites in the 3′UTRs of cognate mRNAs (6). Accumulating evidence has revealed that miRNAs are involved in the regulation of varied biologic processes and their dysregulation is associated with various human diseases, including cancer (7). It has been estimated that more than half of known human miRNAs reside in, or close to, fragile chromosomal sites that are susceptible to deletion, amplification, or translocation in the establishment and progression of tumor (8). Moreover, systematic miRNA expression profilings have observed that characteristic miRNA signatures are associated with prognosis and progression of certain human cancers (7, 9).
Our recent study (10) observed significantly diminished expression of let-7g, embedded in the WDR82 gene on chromosome 3p21.1, in breast cancer compared with normal tissue. Previous studies have consistently reported that LOH is frequently discovered in primary breast cancers at the chromosome 3p allele, especially at 3p21.3, 3p22–24, 3p21.2–3p21.3, 3p25, 3p14.2, 3p14.3, and 3p12 (5, 11, 12). For these reasons, we undertook to uncover specific miRNAs in these aberrant regions, with the hope that these breakpoint-associated miRNAs may provide insights into the causal mechanisms of breast cancer development.
Materials and Methods
All human breast cancer cell lines and nontumorigenic human breast epithelial cell lines used in this study were obtained from the American Type Culture Collection (ATCC) and cultured in conditions as ATCC recommended.
Cell function assays
All functional assays such as total cell number, soft agar colony formation assay, growth on 2-dimensional or in 3-dimensional matrigel, in vitro cell motility (migration, invasion, and wound healing) assays were carried out as described (13). For monolayer cell proliferation, 3,000 cells were seeded into 100 μL media (2% FBS) and viability was quantified every 24 hours. Colony scattering assays were carried out as previously reported (14). All the images were taken under a microscope (Olympus).
Protein extraction and Western blot analysis, RNA extraction and quantitative reverse transcription-PCR
These procedures were carried out as described previously (10). The sequences of primers used in the real-time PCR experiments and the characteristics of the antibodies are listed in Supplementary Tables S7 and S8, respectively.
Aberrant miRNAs on chromosomal 3p in human mammary carcinoma
To screen for miRNAs within the aberrant regions at the chromosome 3p allele, especially at 3p21.3, 3p22–24, 3p21.2–3p21.3, 3p25, 3p14.2, 3p14.3, and 3p12, we initially searched the miRbase 18.0 database (15) on the basis of their precise locations on the human chromosome and identified 34 presently known miRNAs (Supplementary Table S1). We then carried out genomic real-time PCR to determine changes in DNA copy numbers in 20 archived breast cancer specimens compared with 4 breast tissue specimens from patients with benign breast diseases (normal; Supplementary Fig. S1A–F). As summarized in Fig. 1A, we uncovered 11 novel miRNAs with genome copy losses (beyond 20% of the tested cases) in breast cancer. Intriguingly, we also detected 5 miRNAs with obvious genome copy gains. Among the 11 frequently lost miRNAs, we identified miR-128-2 as the most frequently lost miRNA (deleted in 75% cases; Fig. 1A), even higher than let-7g (60% deletion rate). This observation attracted our interest to determine the specific functional roles and regulatory mechanisms of miR-128-2.
miR-128-2 is frequently downregulated together with ARPP21 in human breast cancer and is associated with poor survival outcome
miR-128-2 is an intronic gene, which resides within the 17th intron of the ARPP21 gene on human chromosome 3p22.3 (Fig. 1B). The 21-kDa cAMP-regulated phosphoprotein (ARPP21), also known as Regulator of Calmodulin Signaling, which is enriched in brain, has been described as a candidate TSG and is frequently deleted in mammary carcinoma (16). We first detected the expression of ARPP21 using quantitative reverse transcription (qRT)-PCR and observed a significant reduction in breast cancer (n = 21) compared with normal specimens (n = 15; Fig. 1C). To assess whether deregulation of miR-128-2 expression is due to loss of DNA copy number, we next examined the expression levels of pri-miR-128-2 and mature miR-128. We found that both pri-miR-128-2 and mature miR-128 were remarkably downregulated (Fig. 1D and E), and their expressions were significantly concordant with that of the host gene ARPP21 (Supplementary Fig. S1G and H), indicating that miR-128-2 is often co-expressed with ARPP21.
To further explore the clinicopathologic implications of miR-128 in the progression of breast cancer, we determined the specific abundance of mature miRNA-128in breast cancer specimens (n = 85) and normal breast tissue specimens (n = 18) by in situ hybridization (ISH). We observed significantly decreased expression of miR-128 in mammary carcinoma compared with normal specimens (P = 0.039, Fig. 1F and G). In normal mammary tissue miR-128 was predominantly localized to carcinoma cells with limited expression in stroma. Moreover, we further analyzed the relationships between miR-128 expression and clinicopathologic characteristics of breast cancer and the results are summarized in Supplementary Table S2. A statistically significant association was observed between lower expression of miR-128 and clinical stage (P = 0.036), lymph node metastasis (P = 0.020) and Ki-67 labeling index (P = 0.027).
We next examined the relationship between miR-128 expression and survival in the breast cancer patient cohort with 5-year follow-up (n = 85) using Kaplan–Meier survival analyses. We found that patients with lower expression levels of miR-128 exhibited significantly worse survival outcomes compared with those with higher expression levels of miR-128 (Fig. 1H and I, Supplementary Table S3). Thus, these results showed that miR-128-2 is frequently downregulated together with ARPP21 in human breast cancer and the expression level of miR-128 may be a potential prognostic indicator for patient survival.
Forced miR-128 expression reduces oncogenicity, migration and invasion of mammary carcinoma cells in vitro
To determine the functionality of miR-128 in breast cancer cells, we first carried out a preliminary screen for expression levels of ARPP21, pri-miR-128-2 and mature miR-128 in 2 mammary epithelial cell lines and 8 mammary carcinoma cell lines by qRT-PCR (Fig. 2A, and Supplementary Fig. S2A). We observed remarkably higher expression levels of ARPP21, pri-miR-128-2 and mature miR-128 in 2 immortalized but otherwise normal HMECs, in contrast to lower or undetectable expression levels in mammary carcinoma cell lines. In addition, we observed a strong correlation between the expression level of mature miR-128 and ARPP21 (Pearson coefficient = 0.8749) and a further stronger correlation between the expression level of pri-miR-128-2 and ARPP21 (Pearson coefficient = 0.9689; Supplementary Fig. S2B and C).
We next assessed the potential effect of miR-128 on the phenotype of the MDA-MB-231 cells by transfecting a synthetic miRNA mimic (Fig. 2B). Forced miR-128 expression decreased both the total cell number and colony formation number in soft agar by approximately 60% compared with cells transfected with scrambled control oligonucleotides (Fig. 2B). Smaller colonies and a 2-fold decrease in branching colonies were consistently observed with MDA-MB-231 miR-128 cells cultured in 3-dimensional matrigel (Fig. 2C). In addition, forced expression of miR-128 in MDA-MB-231 cells produced a “cobble stone” like epithelial cellular morphology in monolayer adherent culture as compared with the control cells (Fig. 2D). In colony-scattering assays, we observed a significantly larger proportion of compact cells (3-fold) and smaller proportion of scattered cells (3-fold) with forced expression of miR-128 in MDA-MB-231 cells (Fig. 2E). Finally, forced expression of miR-128 in MDA-MB-231 cells led to a retarded wound closing (Fig. 2F) and reduction in cell migration by 3-fold and invasion by 2.5-fold compared with the control cells (Fig. 2G).
Conversely, miR-128 depletion in mammary epithelial MCF-10A cells by transfection of specific ASO (Supplementary Fig. S2D) significantly increased both the total cell number and colony formation number in soft agar compared with cells transfected with control oligonucleotides (Supplementary Fig. S2D). Moreover, miR-128 depletion in MCF-10A cells increased the number of branching colonies by 3-fold in matrigel (Supplementary Fig. S2F). Furthermore, MCF-10A cells with reduced expression of miR-128 exhibited a spindle like and a more scattered, elongated, and mesenchymal cellular morphology in monolayer adherent culture compared with the control cells (Supplementary Fig. S2E). In colony-scattering assays, we observed a 4-fold larger proportion of scattered cells and 2-fold smaller proportion of compact cells in MCF-10A cells with diminished expression of miR-128 (Supplementary Fig. S2G). Finally, miR-128 depletion in MCF-10A cells resulted in a more rapid wound closing (Supplementary Fig. S2H), and an obvious augmentation in cell migration by 2-fold and invasion by 2.5-fold compared with the control cells (Supplementary Fig. S2I).
To determine whether miR-128 impacts on EMT, we examined the expression of epithelial and mesenchymal markers by Western blotting. MDA-MB-231 cells transfected with miR-128 expressed higher levels of the epithelial markers, E-cadherin and plakoglobin (γ-catenin) and lower levels of the mesenchymal markers, FN1 and vimentin. Conversely, MCF-10A cells transfected with miR-128 ASO expressed lower levels of epithelial markers and higher levels of mesenchymal markers (Supplementary Fig. S2J).
To exclude cell type specific effects, the parallel experimental approaches were deployed in MCF-7 with forced expression of miR-128 and conversely with HBL-100 with reduced expression of miR-128. We observed concordant effects in the different 2 cell lines (Supplementary Fig. S3A-L). Besides, we used TERT immortalized HMEC cells as another model of normal immortalized mammary epithelial cells (17). We infected TERT-HMEC cells with pBabe-miR-128 sponge or control retroviruses and carried out colony formation assay. We observed larger colonies and a 3-fold increase in colony number of TERT-HMEC cells transfected with pBabe-miR-128 sponge compared with control cells (Supplementary Fig. S3M). Thus, these data indicate that forced expression of miR-128 dramatically alters the morphology and impairs oncogenicity and invasiveness of mammary carcinoma cells in vitro.
Forced miR-128 expression impairs mammary carcinoma xenograft growth in vivo
To further determine the function of miR-128 in mammary carcinoma cells in vivo, we constructed 1 retroviral vector expressing miR-128 by harboring pri-miR-128-2, which was transduced in MDA-MB-231 cells, and another vector inhibiting miR-128 expression by inserting a cassette of bulged miR-128 sequences as described (18), which was transduced in nontumorigenic MCF-10A. The 2 stable cell lines, denoted MDA-MB-231 pri-miR-128-2 exhibited elevated miR-128 expression by 2.8-fold and MCF-10A miR-128 sponge reduced miR-128 expression by 4-fold, respectively (Fig. 3A).
We next injected 5 × 106 MDA-MB-231 pri-miR-128-2 cells and MDA-MB-231 control cells subcutaneously into the flanks of female immunodeficient nude mice. Both groups formed palpable and measurable tumors, however, with a 5.2-fold increase of miR-128 expression, tumors formed by MDA-MB-231 pri-miR-128-2 were smaller than those formed by control cells (Fig. 3B and C). Tumors formed by MDA-MB-231 pri-miR-128-2 cells were well confined and differentiated (upper right panel, Fig. 3D), whereas those formed by MDA-MB-231 control cells were poorly encapsulated and highly invasive (upper left panel, Fig. 3D). Furthermore, local infiltration into muscle, and tumor emboli in blood vessels, were observed in tumors formed by control cells (upper left panel, Fig. 3D). Moreover, pulmonary metastases were readily detectable in the lungs of host mice 3.5 weeks after subcutaneous injection with MDA-MB-231 control cells. In contrast, no metastases were detected in mice injected with MDA-MB-231 pri-miR-128-2 cells (Fig. 3D). In addition, Ki-67 or PCNA staining and TUNEL or Caspase-3 assays were carried out on tumor sections, and the MDA-MB-231 pri-miR-128-2 tumors exhibited fewer Ki-67-labeled or PCNA-labeled cells (Fig. 3E, Supplementary Fig. S4A) and higher percentage of TUNEL-positive or Caspase-3-labeled cells (Fig. 3F, Supplementary Fig. S4B) compared with MDA-MB-231 control tumors.
Alternatively, 5 × 106 MCF-10A miR-128 sponge cells and MCF-10A control cells were subcutaneously inoculated into the flanks of host mice. Consistent with prior reports (19), the nontumorigenic MCF-10A cell line is not capable of initiating tumor formation. However, MCF-10A miR-128 sponge cells readily formed visible and measurable tumors (Supplementary Fig. S4C and D), suggesting that miR-128 depletion promoted tumor formation in vivo.
Forced expression of miR-128 represses mammary carcinoma stem cell-like behavior
Dysregulated cancer stem cells have been postulated to be crucial for cancer initiation and progression (20). By extension, we wondered whether miR-128 could influence cancer stem cell-like behavior. We therefore transiently transfected miR-128 mimics into MDA-MB-231 cells and miR-128 ASO into MCF-10A cells. We then cultured these cells in conditioned mammosphere culture media in ultra-low attachment 6-well plates. MDA-MB-231 cells with forced expression of miR-128 exhibited both significantly decreased number (43%) and size of mammospheres compared with control cells (Fig. 4A). Conversely, we observed a slightly yet significant increase (2.6-fold) in mammospheric growth of MCF-10A cells with miR-128 depletion (Fig. 4A). To exclude the possibility that the increase in mammosphere number arose from the aggregation of quiescent cells, we carried out self-renewal assays. We observed decreased numbers of secondary and tertiary mammospheres generated from MDA-MB-231 cells with increased miR-128 expression, whereas MCF-10A cells with decreased miR-128 expression exhibited increased numbers of secondary and tertiary mammospheres compared with their corresponding control cells (Fig. 4B). Recent studies have identified ALDH1 enzymatic activity as a potential marker for breast cancer stem cells (21). To this end, we carried out the ALDEFLUOR assay to determine whether miR-128 modulated the ALDH1 positive cell population. Forced expression of miR-128 in MDA-MB-231 markedly reduced the percentage of ALDH1 positive cells by 3-fold whereas depletion of miR-128 expression increased the percentage of ALDH1 positive cells by approximately 3.5-fold (Fig. 4C). Moreover, we have sorted MDA-MB-231-ALDH1+ and MDA-MB-231-ALDH1- populations, and observed decreased expression of miR-128, CD44 and CD133 and increased expression of CD24 in the MDA-MB-231-ALDH1+ population comparing with MDA-MB-231-ALDH1- population (Supplementary Fig. S4E and F). We next determined whether miR-128 modulated the CD44+CD24neg/low population. Forced expression of miR-128 in MDA-MB-231 reduced the percentage of CD44+CD24neg/low population by 1.5-fold whereas depletion of miR-128 expression increased the percentage of CD44+CD24neg/low population by 1.3-fold (Supplementary Fig. S4G).
To ascertain the influence of miR-128 on tumor-initiating capability in vivo, we subcutaneously injected MDA-MB-231-pri-miR-128-2, MCF-10A miR-128 sponge cells and their corresponding control cells at a series of limiting dilutions from 5 × 105 to 5 × 103 cells into nude mice and identified their ability to form visible tumors. After an incubation period of 6 weeks, as summarized in Fig. 4D, in mice injected with 5 × 104 cells, the tumor incidence derived from MDA-MB-231 pri-miR-128-2 cells (4/6) was lower than control cells (6/6). Strikingly, in mice injected with 5 × 103 cells, 4/6 mice inoculated with MDA-MB-231 control cells formed tumors, whereas no tumor was detected in mice inoculated with MDA-MB-231 pri-miR-128-2 cells. Conversely, no tumors were observed in mice injected with any numbers of MCF-10A control cells or 5 × 103 MCF-10A miR-128 sponge cells. However, tumors were formed in mice injected with 5 × 105 MCF-10A miR-128 sponge cells (2/6) and 5 × 104 MCF-10A miR-128 sponge cells (1/6), suggesting that depletion of miR-128 expression confers normal mammary epithelial MCF-10A cell with tumor-initiating capacity. Collectively, these observations indicate that miR-128 dramatically interfere with mammary carcinoma stem cell-like behavior, both in vitro and in vivo.
To determine a potential association between miR-128 and ALDH1 expression in patients with breast cancer, we carried out qRT-PCR on a cohort of 24 breast cancer specimens to determine the expression levels of miR-128 and ALDH1 mRNA. We observed that the expression of miR-128 was inversely associated with ALDH1 mRNA (Pearson coefficient = −0.4065; Fig. 4E). In addition, we carried out IHC staining on a cohort of 47 breast cancer specimens to determine the expression levels of ALDH1. Combined with the expression levels of miR-128 by ISH, we observed that the expression of miR-128 was inversely and significantly associated with ALDH1 expression (Pearson coefficient = −0.327, P = 0.025).
miR-128 directly targets a cohort of stem cell factors
To determine the mechanism by which miR-128 exerts tumor suppressor properties, we initially employed 3 algorithms to predict the potential mRNA targets of miR-128—MiRanda (22), Pictar (23), and TargetScan (6). On the basis of the sequences in their 3′ UTRs, 98 mRNAs were predicted to be targets by all of the 3 programs (Fig. 5A, Supplementary Table S4). To further narrow the possible downstream effectors of miR-128, we next screened for the mRNA expression profile of an array of genes, which have been widely known in cancer initiation and progression, by qRT-PCR in 2 sets of paired cells. (MDA-MB-231 cells transfected with either miR-128 mimics or control and MCF-10A cells transfected with either miR-128 ASO or scrambled sequence control). Interestingly, as shown in the heat map (Fig. 5B), the data derived from the 2 complementary screening systems were highly concordant, and indicated that miR-128 modulates numerous key genes functionally involved in neoplastic progression (Supplementary Table S5). By combination of the 2 strategies, we identified a cohort of 6 genes (BMI1, CSF1, KLF4, LIN28A, NANOG, and SNAIL), all of which are known transcriptional regulators implicated in stem cell development (24) and contain 3′UTR elements that are partially complementary to miR-128 (Supplementary Fig. S5A-G).
To confirm whether the 6 candidate genes are direct targets of miR-128, we cloned full-length fragments of their 3′UTRs into the luciferase reporter plasmid psi-CHECK2. We then transfected the miR-128 mimic with different 3′UTR constructs into MDA-MB-231 cells and observed that miR-128 markedly repressed the relative luciferase activities of the wild-type 3′UTR of BMI1, CSF1, KLF4, LIN28A, NANOG, and SNAIL (Fig. 5C). Furthermore, mutation of the partially complementary miR-128 site(s) in these 3′UTRs abrogated responsiveness to miR-128 (Fig. 5C). In addition, we measured the effects of miR-128 on expression levels of the 6 targets. Forced expression of miR-128 sharply reduced the protein levels of all 6 targets by 60% to 90%, whereas inhibition of miR-128 resulted in an obvious increase by 4-fold to 10-fold (Fig. 5D). To further ascertain whether miR-128 represses these 6 targets in vivo, we examined their expression by IHC staining on formalin-fixed paraffin-embedded sections derived from tumors formed in the xenograft studies. The 6 proteins were expressed at low, almost undetectable, levels in tumors derived from MDA-MB-231 pri-miR-128-2 cells compared with their control counterparts, respectively (Fig. 5E). Thus, BMI1, CSF1, KLF4, LIN28A, NANOG, and SNAIL are all bona fide targets of miR-128.
miR-128 diminishes mammary carcinoma cell tumor-initiating capacity via AKT and STAT3
Given our identification of these 6 stem cell factors as coordinate targets of miR-128, we reasoned there may be common signaling pathways involved in mediating the effects of miR-128. BMI1 has been reported to act as part of the polycomb repressor complex, which plays a role in stem cell renewal and malignant transformation through activation of AKT pathway (25, 26). CSF1 is known as macrophage colony stimulating factor 1, which regulates progenitor cell differentiation by AKT phosphorylation (27). KLF4, LIN28A, and NANOG are all well-known stem cell factors and reported to possess fundamental roles in iPS reprogramming, in which AKT and STAT3 pathways are activated to promote the induction of normal somatic cells to pluripotent stem cells (28, 29). SNAIL (SNAI) is a zinc finger transcriptional repressor, which has been described to induce EMT via repression of E-Cadherin and activation of several signaling pathways, including STAT3 and AKT (30, 31). Consequently, we examined the expressions levels of AKT and STAT3 by Western blotting. Diminished p-AKT and p-STAT3 activities were observed in MDA-MB-231-miR-128 cells, whereas elevated p-AKT and p-STAT3 activities were detected in MCF-10A cells with miR-128 depletion compared with their respective control cells (Fig. 6A). Moreover, we carried out IHC staining to examine levels of p-AKT and p-STAT3 on sections of xenograft tumors. We observed decreased p-AKT and p-STAT3 in MDA-MB-231-pri-miR-128-2 cells compared with control cells (Fig. 6B), showing that miR-128 deprives mammary carcinoma tumor-initiating suppression via AKT and STAT3 phosphorylation both in vitro and in vivo.
To further show that p-AKT and p-STAT3 activities were required for the tumor-initiating capability of miR-128, MCF-10A miR-128 sponge cells were treated with either the specific AKT inhibitor IV and/or the STAT3 inhibitor III, WP1066, or specific siRNAs targeting AKT and/or STAT3. We observed markedly diminished p-AKT and p-STAT3 activities (Fig. 6C) and expression of AKT and STAT3 (Supplementary Fig. S6C) in MCF-10A miR-128 sponge cells, which subsequently abrogated miR-128 depletion–promoted tumor-initiating capacity (Fig. 6D, Fig. S6D). Thus, miR-128 modulates mammary carcinoma tumor-initiating capability via PI3K/AKT and STAT3 pathways.
miR-128-2 is directly repressed by SNAIL and attenuated via TGF-β
We next undertook to determine how miR-128 expression is specifically regulated. Given our observation that miR-128 was expressed at low levels in breast cancer cells and directly repressed multiple stem cell factors, we hypothesized whether any transcription factors might regulate miR-128 expression. To this end, we searched conserved transcription factor binding sites by use of rVista 2.0 (32), and identified 2 conserved E-box motifs, 1 at -991 bp (E-box 1, CACATG) and the other at -26 bp (E-box 2, CACATG) relative to the transcription start site (+1) of the human miR-128-2 stem-loop, respectively (Fig. 6E). SNAIL family (SNAIL, SLUG, and SMUC), ZEB (ZEB1 and ZEB2), and the bHLH (E47 and TWIST1) families have been reported to bind to E-box sequences (CANNTG) present in the promoters of genes (30). To further determine which factors could impact on miR-128-2 expression, we constructed a 3-kb fragment upstream of human miR-128-2 stem-loop and inserted the fragment into the luciferase reporter plasmid pGL3. We cotransfected this plasmid together with different vectors expressing SNAIL, SLUG or TWIST1. We observed reduced luciferase activity by SNAIL and a milder decrease by SLUG, whereas no change was observed with TWIST1 (Supplementary Fig. S6A).
We next cotranfected the SNAIL expressing vector with different constructs, which contained 3-kb fragment upstream of human miR-128-2 stem-loop, into MCF-10A cells and observed that SNAIL significantly repressed the relative luciferase activity of miR-128-2 (Fig. 6F). In addition, mutations of either of the 2 E-Box motifs abrogated responsiveness to SNAIL (Fig. 6F), indicating that 2 E-Box sites are necessary and functional for SNAIL regulation of miR-128. Conversely, depletion of SNAIL by specific siRNA increased the activity of luciferase construct of miR-128-2, whereas mutations in either of the 2 E-Box sites abolished the increase (Supplementary Fig. S6B). We next carried out ChIP assays to validate whether SNAIL controls miR-128-2 expression by binding to the 2 putative binding sites. With an amplicon at a distance of 2,116 base pairs upstream of the human miR-128-2 stem-loop as negative control, ChIP revealed that SNAIL bound to both E-box 1 and E-box 2 (Fig. 6G). Concordantly, forced expression of SNAIL decreased the expression levels of mature miR-128, pri-miR-128-2 and ARPP21 (Fig. 6H).
We further examined the mammosphere-forming ability of the MCF-10A cells with forced expression of SNAIL and/or miR-128. MCF-10A cells with forced expression of SNAIL formed 8-fold more mammospheres than MCF-10A control cells, and the increased numbers of mammospheres was abrogated by transfection with miR-128 (Fig. 6I). Thus, these findings indicate that SNAIL enhances tumor-initiating capability of mammary epithelial cells by repressing miR-128-2 expression through directly binding to the 2 E-Box sites.
We further sought to determine whether certain growth factors secreted into the tumor microenvironment may contribute to promotion of mammary carcinoma tumor-initiating capacity through repression of miR-128 expression. Previous studies revealed that TGF-β is a potent inducer of SNAIL, which can endow mammary epithelial cells with stem cell-like characteristics (33, 34). Treatment of MCF-10A cells with TGF-β resulted in reduced miR-128 expression (Fig. 6J), sustained over time (Fig. 6K). In addition, to verify the specificity of TGF-β treatment, we used SB431542, a specific TGF-β1 receptor kinase inhibitor (35). We observed that SB431542 abrogated the TGF-β induced reduction in miR-128 expression (Fig. 6L), suggesting that TGF-β decreased miR-128 expression by enhancing expression levels of SNAIL. Functionally, we observed that TGF-β-treated MCF-10A cells exhibited an 11-fold increase in mammosphere formation compared with untreated control cells (Fig. 6M). The increase in mammosphere numbers stimulated by TGF-β was abolished by forced expression of miR-128 (Fig. 6M). Thus, TGF-β promotes mammary epithelial cell oncogenic transformation, to some extent, by miR-128-2 downregulation.
Genome instability has been proposed as one of the hallmarks of cancer, which generates random mutations and confers selective advantage on cancer cells, enabling their outgrowth and eventually increasing cancer risk (36). The fact that 50% of human miRNAs reside in or around fragile chromosomal sites prompted us to uncover these breakpoint-associated miRNAs and clarify their functions. In this study, we have identified 16 miRNAs, which are localized at chromosome 3p and often aberrantly expressed in human breast cancer. Of the 11 miRNAs, which are subject to chromosome loss, several of them have been well documented. Our group has observed that let-7g, located on chromosome 3p21.1, is remarkably reduced in invasive ductal carcinoma and may be considered as a prognostic biomarker for breast cancer (10). In addition, miR-26a-1 and miR-135a-1 have been reported to be frequently lost in a variety of human malignancies, including breast, prostate, liver, brain, pancreatic, colon, and nasopharyngeal carcinoma. Moreover, miR-885-5p is downregulated upon loss of the 3p25.3 region in neuroblastoma (37). Intriguingly, we also found that 5 miRNAs (Fig. 1A) were amplified in breast cancer compared with normal breast tissues. miR-191 has been validated to be amplified and is viewed as a candidate oncogenic miRNA in liver (38), stomach (39), and pancreas (40). LOH comprises not only monosomy, caused by loss of 1 allele, but also homozygous disomy and homozygous polysomies, caused by the concurrent loss of 1 allele and endomultiplication of the other allele (41). Thus, we have identified a series of aberrant miRNAs at chromosome 3p, which exhibits frequent LOH and a discontinuous pattern of allele loss in breast cancer. Further endeavors will unravel their precise roles in the pathogenesis of breast cancer.
Among the eleven downregulated miRNAs, miR-128-2 was the most frequently deleted (75% deletion) in our cohort. Previous reports have identified miR-128 as a TSG. miR-128 was first identified to be enriched in brain and frequently reduced in glioma cells (42), and dysregulation of miR-128 confered a variety of malignant hallmarks on neural cells, including sustained proliferative signaling (43), resisting apoptosis (44), activating invasion and metastasis (45), and promoting self-renewal (25). A recent study has proposed that reduced miR-128 in breast tumor-initiating cells produced chemotherapeutic resistance (46). Consistently, we observed that forced expression of miR-128 in invasive MDA-MB-231 cells inhibited self-renewal and tumor-initiating capacities whereas depletion of miR-128 endowed normal mammary epithelial MCF-10A cells with a malignant phenotype and initiated tumor formation in vivo. Moreover, we further provided proof that miR-128 could directly target a cohort of stem cell factors that have comprehensive roles in stem cell development. In addition, we showed that miR-128-2 is directly regulated by SNAIL, suggesting a potential mechanism of how miR-128-2 is precisely controlled in embryogenesis and tumorigenesis. Finally, we observed miR-128 is involved in the TGF-β signaling pathway, in which TGF-β promotes breast cancer initiation and progression, partially, via miR-128-2 repression. Thus, we have extended our current knowledge by delineating the role of miR-128 in tumor-initiation, implying miR-128 is a candidate molecular target for clinical prognosis or therapy.
It is of interest to note miR-128 could directly target a series of 6 genes: BMI1, CSF1, KLF4, LIN28A, NANOG, and SNAIL, all of which are critical determinants of normal stem cell maintenance and renewal (24), and contribute to tumor formation. One explanation as to why depletion of miR-128 triggers oncogenic transformation in normal mammary MCF-10A cells is due to the liberation of these 6 fundamental factors, which subsequently modulate PI3K/AKT and STAT3 pathways enabling cells to undergo oncogenic transformation. Our hypothesis is further supported by the observations that miR-128 modulates numerous key genes functionally involved in EMT and neoplastic progression (Supplementary Table S5).
The SNAIL family of zinc-finger transcription factors consist of SNAIL1 (SNAIL), SNAIL2 (SLUG), and SNAIL3 (SMUC), all potent inducers of EMT and important predictors of breast cancer metastatic potential and recurrence (47). Expression of SNAIL is precisely regulated by a complex signaling network at the transcriptional and posttranscriptional level. Here, our findings supplement this network with the novel description of 2 positive feedback loops (Fig. 7). First, SNAIL directly binds to E-Box sites upstream of miR-128-2 and represses miR-128-2 expression, which in turn further increases the expression of SNAIL via direct targeting by miR-128-2. Secondly, SNAIL indirectly enhances the expression of BMI1, CSF1, KLF4, LIN28A, and NANOG by inhibiting miR-128-2, which cooperatively increases the activity of AKT and STAT3. It has been reported that AKT upregulates the expression of SNAIL with (48) or without (49) the activation of NF-κB. The STAT3 pathway was also recently been reported to increase the expression of SNAIL (50). Thus, we have indentified another positive feedback loop between SNAIL and the AKT or STAT3 cascade. Furthermore, TGF-β pathway has been showed to induce EMT and endow mammary epithelial cells with stem cell-like traits mainly via transcriptionally inducing SNAIL expression (33, 34). Our results show that TGF-β promotes EMT and breast cancer initiation at least in part by miR-128-2 downregulation, revealing a cascade (TGF-β→TGF-β1 Receptor→SNAIL→miR-128-2→BMI1/CSF1/KLF4/LIN28A/NANOG→pSTAT3/pAKT→EMT/Stem cell-like traits) exists in the breast cancer oncogenic transformation triggered by TGF-β.
Collectively, in this study, our data reveal that decreased miR-128-2 expression due to allele loss on chromosome 3p22.3 can endow nontumorigenic mammary epithelial MCF-10A with malignant phenotypes and stem cell-like behaviors both in vitro and in vivo via de-repression of a cohort of stem cell factors and activation of the PI3K/AKT and STAT3 pathways. In addition, miR-128-2 is downregulated by SNAIL and repressed in the TGF-β signaling pathway. The SNAIL/miR-128 axis provides a novel understanding of the mechanism of mammary epithelial oncogenic transformation.
Disclosure of Potential Conflicts of Interests
No potential conflicts of interest were disclosed.
Conception and design: P.-X. Qian, P.E. Lobie, T. Zhu
Development of methodology: P.-X. Qian, A. Banerjee, Z.S. Wu
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): P.-X. Qian, A. Banerjee, X. Zhang, H. Wang, W. Zhang, P.E. Lobie, T. Zhu
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P.-X. Qian, A. Banerjee, Z.S. Wu, V. Pandey, W. Zhang, P.E. Lobie, T. Zhu
Writing, review, and/or revision of the manuscript: P.-X. Qian, A. Banerjee, P.E. Lobie, T. Zhu
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): V. Pandey, X.-F. Lv, S. Tan, P.E. Lobie, T. Zhu
Study supervision: P.E. Lobie, T. Zhu
We thank Muh-Hwa Yang and Kou-Juey Wu (Institute of Clinical Medicine, National Yang-Ming University, Taipei, Taiwan) for kindly providing the SNAIL, SLUG and TWIST1-expression plasmids. TERT-HMEC cell line was kindly gifted from Hines WC (University of New Mexico School of Medicine, New Mexico).
This work was supported by the National Key Scientific Program of China (2012CB934002, 2010CB912804), Chinese Academy of Sciences (XDA01040410), National Natural Science Foundation of China (30971492, 81101597), Chinese Academy of Sciences Visiting Professorship for Senior International Scientist (2010TZS03), and Cancer Science Institute of Singapore.
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.