Register for Mechanobiology 2016!

Epigenetic modulation of the miR-200 family is associated with transition to a breast cancer stem-cell-like state
Yat-Yuen Lim, Josephine A. Wright, Joanne L. Attema, Philip A. Gregory, Andrew G. Bert, Eric Smith, Daniel Thomas, Angel F. Lopez, Paul A. Drew, Yeesim Khew-Goodall, Gregory J. Goodall


The miR-200 family is a key regulator of the epithelial–mesenchymal transition, however, its role in controlling the transition between cancer stem-cell-like and non-stem-cell-like phenotypes is not well understood. We utilized immortalized human mammary epithelial (HMLE) cells to investigate the regulation of the miR-200 family during their conversion to a stem-like phenotype. HMLE cells were found to be capable of spontaneous conversion from a non-stem to a stem-like phenotype and this conversion was accompanied by the loss of miR-200 expression. Stem-like cell fractions isolated from metastatic breast cancers also displayed loss of miR-200 indicating similar molecular changes may occur during breast cancer progression. The phenotypic change observed in HMLE cells was directly controlled by miR-200 because restoration of its expression decreased stem-like properties while promoting a transition to an epithelial phenotype. Investigation of the mechanisms controlling miR-200 expression revealed both DNA methylation and histone modifications were significantly altered in the stem-like and non-stem phenotypes. In particular, in the stem-like phenotype, the miR-200b-200a-429 cluster was silenced primarily through polycomb group-mediated histone modifications whereas the miR-200c-141 cluster was repressed by DNA methylation. These results indicate that the miR-200 family plays a crucial role in the transition between stem-like and non-stem phenotypes and that distinct epigenetic-based mechanisms regulate each miR-200 gene in this process. Therapy targeted against miR-200 family members and epigenetic modifications might therefore be applicable to breast cancer.


Epithelial-derived tumours contain a heterogeneous population of cells which can be characterized by differences in histopathology and functional properties including proliferative and apoptotic responses to therapies and capacity for anchorage-independent growth (Hanahan and Weinberg, 2011). Recent evidence supports the existence of a cellular hierarchy within epithelial tumours. At the apex of this hierarchy is a tumour-initiating cell (T-IC) or cancer stem cell (CSC) population that can self-renew and differentiate to progeny cells, thus resulting in the observed cellular and functional heterogeneity within epithelial tumours (Polyak and Hahn, 2006; Reya et al., 2001). CSCs have been prospectively isolated from a variety of solid tumours, including breast (Al-Hajj et al., 2003; Ginestier et al., 2007), brain (Singh et al., 2004), colorectal (Dalerba et al., 2007; Ricci-Vitiani et al., 2007), head and neck (Prince et al., 2007), pancreatic (Li et al., 2007), prostate (Patrawala et al., 2007; Patrawala et al., 2006), melanoma (Schatton et al., 2008), and bladder (Chan et al., 2009). The characterization of these rare tumorigenic cells has the potential to provide important prognostic and therapeutic value for epithelial cancers.

Breast cancer stem-like cells (bCSCs) enriched in the CD44hi/CD24−/low subpopulation are proposed to be largely responsible for cancer progression and metastasis (Al-Hajj, 2007; Al-Hajj et al., 2003). They possess stem-like properties including the ability to self-renew and differentiate into CD44−/low/CD24hi progeny, show resistance to standard therapies and increase in numbers after short courses of fractionated irradiation (Phillips et al., 2006). At the molecular level, bCSC gene signatures are associated with decreased overall patient survival, poor metastasis-free survival and tumour recurrence (Liu et al., 2007). Furthermore, bCSCs become more pronounced in tumour tissue following endocrine therapy or chemotherapy, consistent with their selective post-treatment survival (Creighton et al., 2009). Therefore, an improved understanding of bCSCs potential to drive breast cancer progression could inform therapeutic targeting of breast cancer.

The epithelial–mesenchymal transition (EMT) is a crucial embryonic developmental process that is characterized by the losses of E-cadherin, cell polarity, cell–cell and cell–matrix contact as well as gain in motility and fibroblast-like morphology (Bracken et al., 2009; Kalluri and Weinberg, 2009). Recently, EMT has been shown to confer cells with stem-like properties, with migratory and invasive capabilities associated with metastatic competence (Mani et al., 2008). Multiple transcription factors including ZEB1, TWIST, SNAI1, and FOXC1 have been shown to induce EMT (Polyak and Weinberg, 2009; Thiery et al., 2009). An EMT core signature derived from the changes in gene expression shared by upregulation of Goosecoid, SNAI1, TWIST and TGF-β1, and by downregulation of E-cadherin was closely associated with the claudin-low and metaplastic breast cancers subtypes (Taube et al., 2010). Importantly, these two breast cancers subtypes were reported to have a significant similarity to a ‘tumorigenic’ gene signature defined using patient-derived CD44+/CD24 breast tumour-initiating stem-like cells (Hennessy et al., 2009), further supporting the notion that EMT associates with stem-like properties. Cancerous cells that have acquired a more malignant undifferentiated state with worse outcome usually display mesenchymal-like characteristics that are more metastatic and likely to relapse (Thiery et al., 2009). Hence, these studies suggest that the bCSCs and cells with stem-like properties may also be EMT-like.

MicroRNAs (miRNAs) have been shown to regulate gene expression and can control crucial cellular pathways, including stem cell identity and lineage commitment (Gangaraju and Lin, 2009). The acquisition of stem-like and EMT-like properties involves major changes in gene expression, including miRNAs. Different transcription factors and miRNAs can interact within highly interconnected protein–protein, protein–DNA, and protein–non-coding RNA networks (Orkin and Hochedlinger, 2011). These networks can be governed by a milieu of chromatin-remodelling and modifying complexes to regulate chromatin organization and gene expression including DNA methylation and histone modifications (Prezioso and Orlando, 2011; Young, 2011).

The miR-200 family consists of five members (miR-200a, miR-200b, miR-200c, miR-141 and miR-429) which are clustered and expressed as two separate polycistronic pri-miRNA transcripts with the miR-200b-200a-429 cluster at chromosomal location 1p36 and miR-200c-141 cluster at chromosomal location 12p13. A double negative feedback loop between the ZEB transcription factors and expression of miR-200 family members, for example, regulates the induction of EMT and the reverse process, mesenchymal to epithelial transition (MET) (Bracken et al., 2008; Burk et al., 2008; Gregory et al., 2008). While loss of miR-200 family gene expression has been associated with EMT, understanding of the molecular links between the miR-200 family and the stem-like cellular state remains incomplete. Here, we report the use of immortalized human mammary epithelial cells (HMLE) to investigate the function and regulation of the miR-200 during their conversion from a non-stem to a stem-like phenotype and during EMT. We found that HMLE cells spontaneously converted from a non-stem to a stem-like phenotype and this change was accompanied by loss of miR-200 expression. Restoration of miR-200 in stem-like cells partially reprogrammed these cells back to the non-stem phenotype, which was accompanied by distinct phenotypic and molecular changes. Investigation of the epigenetic-based mechanism controlling miR-200 expression revealed both histone modifications and promoter DNA methylation were significantly altered in non-stem and stem-like phenotypes. These results highlight the importance of epigenetic regulation in the maintenance of miR-200 family expression in epithelial cells, and their aberrant silencing in EMT and stem-like cells.


HMLE populations include CD44hi/CD24low cells that spontaneously arise from CD44low/CD24hi cells

To investigate the role of miR-200 in regulating stem-like properties in breast epithelial cells, we studied the immortalized human mammary epithelial cell line designated HMLE. An early passage of this cell line contains a small subpopulation of CD44hi/CD24low stem-like cells which is distinct from the parental CD44low/CD24hi cells (Mani et al., 2008). The natural existence of two definitive populations prompted us to use fluorescence activated cell sorting (FACS) to separate the two populations and assess the role of the miR-200 family in regulating stem-like properties. Upon serial passaging of the parental cells we observed an increase in the proportion of cells that displayed a mesenchymal morphology, which suggested that epithelial cells spontaneously converted to mesenchymal cells during culture (Fig. 1A). This correlated with an increased proportion of the CD44hi/CD24low subpopulation (Fig. 1A). To further explore the nature of these spontaneously derived CD44hi/CD24low/mesenchymal-like HMLE cells, we isolated these cells by FACS.

Fig. 1.

sl-HMLE cells display stem-like and EMT-like properties. (A) Changes in cell morphology and CD44 and CD24 cell surface markers over several passages. In the flow cytometry two-dimensional plots, the y-axis shows CD44 and the x-axis shows CD24 fluorescence intensities. Scale bar: 50 µm. (B) Growth curve of CD44low/CD24hi and CD44hi/CD24low HMLE cells measured by the MTS assay. Data are means ± s.e.m., n = 3. (C) Contour plots from flow cytometry for CD44 (y-axis) and CD24 (x-axis) of sorted CD44hi/CD24low HMLE cells after 4 weeks in culture. (D) Sphere-forming efficiency of first and second generation mammospheres formed from sl-HMLE and nsl-HMLE cells. Only spheres of 150 µm diameter or more were quantified. Data are means ± s.e.m.; n = 4; **P<0.01. (E) Changes in two epithelial lineage markers, CK18 and CK14. mRNAs were measured by real-time PCR. Data are means ± s.e.m.; n = 3; **P<0.01, normalized to GAPDH. (F) Changes in expression of epithelial and mesenchymal markers as measured by real-time RT-PCR. Data are means ± s.e.m.; n = 3; *P<0.05, **P<0.01, normalized to GAPDH and shown relative to the level in HMLE cells (for Twist1 n = 2). sl-HMLE M2 represents sl-HMLE-derived second generation mammospheres.

To determine whether the increase in CD44hi/CD24low HMLE cells was due to different proliferation rates, we measured the cell proliferation rate using the MTS cell proliferation assay. We found no significant difference in growth rate between the CD44hi/CD24low and CD44low/CD24hi sub-populations (Fig. 1B). This result suggested that the increase in proportion of CD44hi/CD24low cells was due to direct conversion from CD44low/CD24hi cells. Next, we determined whether the CD44hi/CD24low HMLE cells could revert back to being CD44low/CD24hi, since normal tissue stem cells are able to give rise to differentiated progeny. Interestingly, the CD44hi/CD24low cells retained their CD44hi/CD24low profile and mesenchymal morphology even after multiple passages (Fig. 1C). These results support the notion that there is a spontaneous conversion of CD44low/CD24hi/epithelial cells to a stable CD44hi/CD24low/mesenchymal-like state over time. Henceforth, we termed these de novo spontaneously derived CD44hi/CD24low/mesenchymal stem-like HMLE cells, ‘sl-HMLE’ and the CD44low/CD24hi/non-stem epithelial cells, ‘nsl-HMLE’.

sl-HMLE cells exhibit stem-like and EMT-like properties

To further explore the stem-like nature of the sl-HMLE cells, we assessed their ability to form mammospheres in culture, which relies on the ability of cells to survive and grow in non-adherent culture conditions (Dontu et al., 2003). These experiments revealed that the frequency of mammosphere formation from sl-HMLE cells was initially similar to that from nsl-HMLE cells (Fig. 1D). However, second generation cultures from sl-HMLE cells produced significantly more mammospheres than second generation cultures from nsl-HMLE cells, indicating sl-HMLE cells possess increased self-renewal potential (Fig. 1D). Stem-like cells have been described as being undifferentiated cells that do not express, or have lower expression of mammary lineage epithelial markers including the luminal epithelial cytokeratin 18 (CK18) and the basal/myoepithelial cytokeratin 14 (CK14) (Dontu et al., 2003; Ponti et al., 2005; Yu et al., 2007). We found that sl-HMLE cells and cells from second generation sl-HMLE mammospheres had lower expression of CK18 and CK14, consistent with a less differentiated, stem-like phenotype (Fig. 1E).

To determine if the sl-HMLE cells have an EMT-like gene signature, as suggested by their elongated mesenchymal morphology, we analysed the expression of E-cadherin, N-cadherin, ZEB1, ZEB2 and Twist1 (Fig. 1F). Consistent with their mesenchymal morphology, the sl-HMLE cells had low expression of E-cadherin, and high expression of the mesenchymal associated genes, N-cadherin, ZEB1, ZEB2 and Twist1 (Fig. 1F). Taken together these data indicate that sl-HMLE cells exhibit stem-like and EMT-like properties in comparison to the nsl-HMLE cells including a less differentiated phenotype and increased self-renewal potential.

The miR-200 family coordinately regulates EMT and stem-like features of mammary epithelial cells

The miR-200 family is an important regulator of EMT and recent studies have demonstrated they also play a role in inhibiting stem-like properties (Iliopoulos et al., 2010; Shimono et al., 2009). To assess whether the miR-200 family can coordinately regulate the stem-like and/or EMT-like properties observed in the sl-HMLE cells, we measured these miRNAs by qPCR. These experiments revealed that miR-200a, miR-200b and miR-200c were downregulated in the sl-HMLE cells and sl-HMLE-derived mammospheres compared to HMLE and nsl-HMLE cells (Fig. 2A). To assess whether the miR-200 levels were similarly reduced in CD44hi/CD24low/− prospective bCSCs, cells were isolated from pleural or ascites effusion samples obtained from breast cancer patients (supplementary material Table S1) and the CD45 subpopulation further sorted for their CD44/CD24 status. We found that the miR-200 family members were consistently downregulated in the CD45/CD44hi/CD24low/− breast cancer cells (prospective bCSCs) compared to the CD45/CD44low/CD24hi and CD45/CD44hi/CD24hi non-CSC subpopulations (Fig. 2B). These findings together with the results obtained from sl-HMLE cells and sl-HMLE-derived mammospheres suggest that the miR-200 family members are involved in cellular pathways that give rise to stem-like properties. To further investigate this role we generated stable sl-HMLE cell lines that overexpress the miR-200b-200a-429 cluster or the miR-200c-141 cluster using pMSCV-puro retroviral vectors (hereafter referred to as sl-HMLE MSCV; sl-HMLE MSCV miR-200b-200a-429; sl-HMLE MSCV miR-200c-141). Enforced expression of miR-200 caused downregulation of ZEB1, ZEB2 and Twist1 with concomitant increase in E-cadherin mRNA and protein (Fig. 3A,B) and rearrangement of the actin cytoskeleton from stress fibre to cortical pattern as well as junctional localization of E-cadherin and ZO-1 (Fig. 3C). Thus, enforced expression of miR-200 caused the sl-HMLE cells to undergo mesenchymal to epithelial transition (MET).

Fig. 2.

Stem-like cells and bCSCs express low levels of the miR-200s. (A) Comparison of miR-200a, miR-200b and miR-200c levels in sl-HMLE cells compared to nsl-HMLE cells and in sl-HMLE-derived second generation mammospheres (M2) compared with HMLE cells as measured by TaqMan real-time PCR. Data are means ± s.e.m.; n = 2; *P<0.05, normalised to snRNA U6. (B) Expression of the miR-200 family members in sorted CD45/CD44hi/CD24low/− prospective bCSCs and CD45/CD44low/CD24hi or CD45/CD44hi/CD24hi non-CSC cells. miRNAs were measured by TaqMan real-time PCR. Data are representative results from three patients, normalized to RNU6B or snRNA U6 and are shown relative to CD44lowCD24hi cells. Error bars represent standard deviation of triplicate PCR assays.

Fig. 3.

Effects of stable overexpression of the miR-200s in sl-HMLE cells. (A) Expression of E-cadherin, ZEB1, ZEB2, Twist1, CK18 and CK14 in the stable pMSCV-puro sl-HMLE cell lines as measured by qPCR, normalised to GAPDH. Data are representative of two independent viral transductions (means ± s.d.) and are expressed relative to sl-HMLE MSCV EV. Error bars represent standard deviation of triplicate PCR assays. (B) Western blots showing E-cadherin, ZEB1 and ZEB2 proteins in the cell lines. Alpha-tubulin was used as loading control. (C) Phase-contrast images (first row) and E-cadherin, ZO-1, F-actin, CK18 and CK14 immunofluorescence staining of sl-HMLE stable cell lines transduced with EV or miR-200b-200a-429 or miR-200c-141 clusters. DAPI staining (blue) was used to detect nuclei and is merged with the indicated specific antibody staining (red) in their respective panels. Scale bar: 50 µm.

Next, we investigated whether restoration of miR-200 in the sl-HMLE cells results in loss of stem cell markers and functional properties. sl-HMLE cells were analysed for CD44 and CD24 marker expression using flow cytometry and in vitro sphere assays were performed to assess stem cell-like function. Expression of miR-200 caused a shift from a predominantly CD44hi/CD24low to the CD44low/CD24hi phenotype (Fig. 4A) and reduced their sphere-forming potential, especially in the growth of second generation spheres (Fig. 4B,C). Expression of miR-200 in the sl-HMLE cells also partially restored CK18 and CK14 expression (Fig. 3A,C), consistent with these cells undergoing differentiation. Expression of miR-200 family members in the sl-HMLE cells had little effect on proliferation rate (Fig. 4D). In summary, these results demonstrate that the miR-200 family can coordinately control EMT and stem-like properties of mammary epithelial cells.

Fig. 4.

Stable overexpression of the miR-200s in sl-HMLE cells results in a change from stem-like to non-stem phenotype. (A) Flow cytometry analysis of CD44 and CD24 expression on sl-HMLE cells and stable pMSCV-puro sl-HMLE cell lines. (B) Formation of first generation and second generation mammospheres from the stable pMSCV-puro sl-HMLE cell lines. Only spheres ≧150 µm in diameter were quantified. The data are means ± s.e.m.; n = 6; *P<0.05, **P<0.01. (C) Phase-contrast images of sl-HMLE cells and stable pMSCV-puro sl-HMLE cells cultured in the sphere assay (first generation). Scale bars: 100 µm. (D) Proliferation rates of the stable pMSCV-puro sl-HMLE cell lines measured using MTS assays. The data are means ± s.e.m.; n = 2.

The miR-200 family exhibit differential DNA methylation states in non-stem and stem-like cells

The sustained reduction in miR-200 gene expression in sl-HMLE cells suggested that epigenetic changes might be involved in the initiation and/or maintenance of the stem-like sl-HMLE subpopulation. We and others have reported that repressed miR-200 genes acquire promoter CpG hypermethylation (Gregory et al., 2011; Neves et al., 2010; Vrba et al., 2010; Wiklund et al., 2011). To investigate the degree of DNA methylation occurring at the miR-200b-200a-429 and miR-200c-141 genes (Fig. 5A) in the different HMLE cell types compared to other breast epithelial and mesenchymal cancer cell lines, we examined CpG promoter methylation using bisulfite PCR melt curve analysis (Smith et al., 2009). The promoters proximal to the transcription start site (TSS) of the miR-200 genes in HMLE and MDA-MB-468 cells were unmethylated which is consistent with the expression of the genes in these cells (Fig. 5B). In the mesenchymal MDA-MB-231 and Hs578T cell lines that no longer express the miR-200 genes, CpG hypermethylation was detected across the promoter. Intriguingly, the CpG methylation profiles of the miR-200b-200a-429 gene in HMLE and sl-HMLE cells were quite similar; the TSS and promoter region was mostly unmethylated except for region B where sl-HMLE cells had a slight increase in DNA methylation (Fig. 5B). In contrast, miR-200c-141 gene silencing and CpG promoter and primary transcript methylation were positively correlated in the two cell types. The sl-HMLE cells had a high degree of methylation in the miR-200c TSS compared to HMLE cells (Fig. 5B). Similar differences in CpG methylation profiles of miR-200b-200a-429 and miR-200c-141 were also observed in HMLE cells treated with TGFβ1 to enrich for the CD44hi/CD24low mesenchymal HMLE population (HMLE+TGFβ1) (Fig. 5B).

Fig. 5.

Methylation profiles of the miR-200 gene promoters. (A) Schematic showing the regions (labelled A,B,C) analysed by PCR melt curve analysis. (B) PCR melt curve analysis of the CpG-rich regions encompassing the miR-200b-200a-429 and </emph>miR-200c-141 transcription start sites in high miR-200 (HMLE, MDA-MB-468) and low miR-200 (sl-HMLE, HMLE+TGFβ1, Hs578T, MDA-MB-231) human cell lines. The breast cancer cell lines were characterized for miR-200 expression by Gregory et al. (Gregory et al., 2008). Methylase-treated DNA from healthy donor lymphocytes (M-ref) were used as a methylation reference/positive control. (C) CpG methylation analysis of the miR-200b-200a-429 and miR-200c-141 loci in HMLE cells following 8, 24 and 46 days of TGFβ1 treatment using the Illumina HM450K methylation array.

Although informative, the bisulphite PCR melt curve analysis is limited in scope, allowing for the analysis of only a small number of CpG sites within a defined genomic region. To increase the resolution of this analysis, we employed the Illumina HumanMethylation450 BeadChip array. This approach enabled us to determine the changes in DNA methylation across the entire miR-200b-200a-429 and miR-200c-141 chromosomal regions in HMLE and HMLE+TGFβ1 cells. The results from the array were consistent with the melt curve analysis of the miR-200 promoter regions. Comparison of beta-values of all probes near the promoter of the miR-200c-141, but not miR-200b-200a-429, revealed a progressive increase in the level of DNA methylation during transition to a mesenchymal/stem-like state (Fig. 5C). Moreover, we found the miR-200b-200a-429 gene body was heavily methylated compared to the miR-200c-141 gene body region (data not shown). This observation was in line with other described polycomb group target genes that typically have a hypermethylated gene body and unmethylated TSS regardless of gene expression or silencing (Deaton and Bird, 2011).

The miR-200b-200a-429 and miR-200c-141 gene promoters exhibit distinct histone modification profiles in non-stem and stem-like cells

The reduced levels of promoter CpG methylation at the miR-200b-200a-429 gene compared to the heavily methylated miR-200c-141 promoter prompted us to consider whether additional epigenetic modifications took place at these promoters in the sl-HMLE cells. Thus, we performed chromatin immunoprecipitation coupled to quantitative PCR analysis (ChIP-qPCR) to detect specific histone modifications near the TSS of the miR-200 genes. Early passage parental HMLE (∼97% CD44low/CD24hi) cells and sl-HMLE cells were analysed for activating (H3K4me3, H3K9/14ac and H3K27ac) and silencing histone modifications (H3K27me3, H3K9me2, and H3K9me3). The unmodified histone H3 control was included for normalization between cell types and we verified the various ChIP assays by performing ChIP analysis of the characterised promoters of the β-ACTIN, GAPDH, MYT1, PAX5 and MYOD genes, which served as positive and negative controls to indicate the expected levels of enrichment of histone modifications and chromatin modifying enzymes for the two cell types (data not shown). Analysis of the miR-200 genes revealed that H3K4me3 and H3K9/14ac occupied the miR-200b-200a-429 and miR-200c-141 gene promoters in HMLE cells that express these miRNAs (Fig. 6A). A higher level of enrichment of the activating histone modifications was detected for miR-200c-141 compared to the miR-200b-200a-429 promoter region, which could reflect increased expression levels of miR-200c-141 compared to miR-200b-200a-429 in HMLE cells (Gregory et al., 2008; Wiklund et al., 2011). Consistent with the miR-200 expression in these cells, the silencing histone modifications, H3K27me3, H3K9me2 and H3K9me3, were not detected across the promoter regions of the miRNAs (Fig. 6A). In contrast to HMLE cells, the activating marks across both miR-200 promoters were erased in sl-HMLE cells. However, whereas the activating marks at the miR-200b-200a-429 TSS promoter were replaced by the polycomb-mediated H3K27me3 mark, this particular silencing histone modification was not detected at miR-200c-141 gene (Fig. 6A). ChIP-qPCR analysis of HMLE+TGFβ1 cells revealed similar correlations between the activating and silencing histone modifications and miR-200 gene expression (Fig. 6A). Furthermore, miR-200b-200a-429 but not miR-200c-141 gene repression was linked with polycomb-mediated H3K27me3 gene silencing. Taken together, these results suggest that dynamic changes in chromatin states across the miR-200 genes between non-stem and stem-like cells are mediated by distinct and complementary epigenetic mechanisms – DNA methylation in the case of miR-200c-141 and PcG-mediated silencing in the case of miR-200b-200a-429.

Fig. 6.

Histone modification profiling of HMLE, sl-HMLE and HMLE+TGFβ1 cells. (A) Association of histone modifications with the promoters of the miR-200b-200a-429 and miR-200c-141 genes in HMLE, sl-HMLE and HMLE+TGFβ1 cells. The x-axis shows the distance from the transcription start site (marked with vertical dashed line) and the y-axis shows the log2 enrichment values over input, normalized to H3. Error bars represent standard deviation from duplicate PCR assays. (B) Expression of SUZ12 and EZH2 in sl-HMLE and nsl-HMLE. Data are means ± s.e.m.; n = 2; *P<0.05 and normalized to GAPDH. (C) Measurement of SUZ12 and EZH2 in the stable pMSCV-puro sl-HMLE cell lines as measured by qPCR, normalised to GAPDH. Data are representative of two independent viral transductions (means ± s.d.) and are expressed relative to sl-HMLE MSCV EV. Error bars represent standard deviation of triplicate PCR assays.

Polycomb group proteins associate with miR-200b-200a-429 but not miR-200c-141 in sl-HMLE and EMT-induced HMLE cells

To confirm that polycomb repressive complex 2 (PRC2) complexes are associated with the miR-200b-200a-429 promoter in sl-HMLE and HMLE+TGFβ1 cells, ChIP analysis was performed using antibodies to the PRC2 components EZH2 and SUZ12. This analysis revealed that EZH2 and SUZ12 occupied the TSS of miR-200b-200a-429 but not miR-200c-141 in the sl-HMLE and HMLE+TGFβ1 cells (Fig. 6A). By contrast, EZH2 and SUZ12 were not detected across these promoters in HMLE cells consistent with miR-200 expression (Fig. 6A). These results indicated that in sl-HMLE and HMLE+TGFβ1 cells, miR-200b-200a-429 but not miR-200c-141 is occupied by PRC2, leading to H3K27me3-mediated gene silencing. Interestingly, the levels of EZH2 and SUZ12 were higher in sl-HMLE compared with nsl-HMLE cells and were repressed after expression of miR-200 in sl-HMLE cells (Fig. 6B,C). This is consistent with reports indicating miR-200b can directly repress SUZ12 (Iliopoulos et al., 2010) and suggest loss of miR-200 in sl-HMLE may enhance PRC2 mediated repression of miR-200b-200a-429.

To investigate whether PgC-mediated silencing of miR-200b-200a-429 gene may occur in breast cancers, we examined breast cancer cell lines. We found that the luminal/epithelial breast cancer cell line, MDA-MB-468, had a similar histone modification profile to HMLE cells, consistent with their expression of the miR-200 genes (Fig. 7). In contrast, in the basal/mesenchymal breast cancer cell line, MDA-MB-231, which no longer expresses the miR-200 family members, the cells had H3K27me3 marks and EZH2 and SUZ12 association indicative of polycomb repression at the miR-200b-200a-429 but not the miR-200c-141 gene (Fig. 7). Similarly, H3K27me3 was found at the miR-200b-200a-429 promoter but not the miR-200c-141 promoter in Hs578T cells (data not shown). Collectively, these results indicate that repression of miR-200b-200a-429 and miR-200c-141 genes is mediated by distinct cell context-dependent epigenetic mechanisms involving specific and dual combinations of histone and DNA methyltransferases.

Fig. 7.

Histone modification profiling of MDA-MB-468 and MDA-MB-231 cells. Association of histone modifications with the promoters of miR-200b-200a-429 and miR-200c-141 clusters in MDA-MB-468 and MDA-MB-231 cells. The x-axis shows the distance from the transcription start site (marked with vertical dashed line) and the y-axis shows the log2 enrichment values over input, normalized to H3. Error bars represent standard deviation from duplicate PCR assays.


The miR-200 family is a well-established mediator of EMT, but their ability to regulate transition between non-stem and stem-like states in a model system has not been well characterised. Here, we show that the miR-200 family can coordinately regulate epithelial cell plasticity and stem-like properties of immortalized human epithelial breast cells (HMLE). During the conversion to a stem-like state, the miR-200 family genes are epigenetically silenced by distinct de novo processes involving DNA methylation (in the case of miR-200c-141) and polycomb protein-associated repression (in the case of miR-200b-200a-429). These findings suggest genetic and epigenetic regulation of miR-200 plays an important role in plasticity between non-stem and stem-like phenotypes.

In this study, we identified a spontaneously generated subpopulation of CD44hi/CD24low/− stem-like cells derived from HMLE which we termed sl-HMLE. The sl-HMLE cells appeared to fit the dynamic EMT-CSC interconversion model more than the conventional stem cell/CSC model (Gupta et al., 2009; Shackleton, 2010). This is an attractive model which permits the study of pathways that contribute to the gain in stem-like properties including miRNA and epigenetic changes. The sl-HMLE cells described here, although similar to EMT-induced HMLE cells (Mani et al., 2008), have several differences: (a) the small subpopulation of CD44hi/CD24low fraction increased over time in culture, (b) sl-HMLE cells are CK14-low, CK18-low, and do not revert back to the HMLE or nsl-HMLE phenotype. While this work was being finalised, Chaffer et al. (Chaffer et al., 2011) reported similar findings in transformed HMEC cultures. They observed that the floating transformed HMECs have enriched CD44hi/CD24low stem-like cells that do not revert back to the parental bulk population which is mainly CD44low/CD24hi. Furthermore, these cells gave rise to ductal structures that stained for both CK14 and pan-cytokeratin when injected orthotopically into the humanized mammary fat pad of nude mice. Thus sl-HMLE behave as stem-like cells, which as an in vitro model system, can be utilized to characterize the molecular and cellular mechanisms regulating conversion between non-stem and stem-like states.

The observation that HMLE cells undergo spontaneous and progressive conversion into a stem-like/mesenchymal counterpart in culture led to the hypothesis that molecular-based switches might be regulating this process. We recently showed that MDCK cells are able to switch and maintain their cell phenotype (epithelial or mesenchymal) in culture due to an autocrine TGF-β/ZEB/miR-200 signalling network (Gregory et al., 2011). Although the balance of miR-200 and ZEB levels via miR-200/ZEB feedback loop (Bracken et al., 2008; Burk et al., 2008) was crucial for the MDCK cells to initiate a mesenchymal phenotype, it required secretion of autocrine TGFβ to drive stable ZEB expression in order to maintain the mesenchymal phenotype in culture after removal of exogenous TGFβ1 treatment (Gregory et al., 2011). Similarly, sl-HMLE cells have increased expression of genes in the TGFβ, IL-6 and HAS-CD44 pathways. These pathways could potentially be involved in the conversion of HMLE cells to de novo sl-HMLE cells in culture and also maintaining the stem-like/EMT-like phenotype of sl-HMLE cells via autocrine and paracrine signalling. Recent studies support the proposal that autocrine and paracrine signalling induce and maintain embryonic stem (ES) cells (Berge et al., 2011) and stem-like/mesenchymal state in the breast (Scheel et al., 2011).

By stable re-expression of each miR-200 genomic cluster in sl-HMLE, we demonstrate here that the miR-200 family functionally represses stem-like properties. In addition we find that bCSCs isolated from metastatic breast cancer patients, like the sl-HMLE cells, have decreased levels of the miR-200 family. These results are consistent with other studies that show that miR-200 family members regulate stemness in normal mammary stem cells (Shimono et al., 2009), bCSCs (Iliopoulos et al., 2010; Iliopoulos et al., 2009; Shimono et al., 2009), pancreatic and colorectal CSCs (Wellner et al., 2009) and with the reversion of EMT observed when miR-200 is expressed in the MDA-MB-231 breast cancer cell line (Park et al., 2008; Burk et al., 2008). Furthermore, we showed that the expression of miR-200 family members are dynamically associated with and at least partially maintained by epigenetic changes. Importantly, we found that the silencing mechanism of the two clusters of miR-200 family members in sl-HMLE and HMLE+TGFβ1 cells were controlled by distinct mechanisms. The miR-200b-200a-429 cluster was occupied by the PcG proteins while miR-200c-141 was silenced by DNA methylation. Intriguingly, this contrasts findings in MDCK cells induced to undergo EMT where DNA methylation occurred at both miR-200 cluster promoters (Davalos et al., 2011; Gregory et al., 2011). During the preparation of this study for publication, other groups demonstrated that combinatorial histone and DNA modifications regulate the miR-200 family members (Au et al., 2012; Cao et al., 2011; Davalos et al., 2011; Vrba et al., 2011). Cao et al. and Davalos et al. showed that PcG proteins can regulate both the miR-200b-200a-429 and miR200c-141 genes (Cao et al., 2011; Davalos et al., 2011). However, data obtained from the sl-HMLE and HMLE+TGFβ1 cells is consistent with the findings from Vrba et al. and Au et al. that PcG mediates the silencing of the miR-200b-200a-429 and not the miR-200c-141 gene (Vrba et al., 2011; Au et al., 2012). This is supported by our findings in the basal breast cancer cell lines, MDA-MB-231 and Hs578T indicating that H3K27me3 occupies the miR-200b-200a-429 cluster and not miR-200c-141 cluster. Together, these studies suggest that different cell lines may acquire distinct silencing mechanisms to regulate miR-200 expression. It will be interesting to determine whether distinct epigenetic mechanisms regulate the miR-200 genes in other contexts where plasticity between non-stem and stem-like states occurs.

PcG proteins are critical regulators of gene silencing important for tissue development, stem cell function, and X chromosome inactivation (Prezioso and Orlando, 2011). While the cause for polycomb-mediated silencing on the miR-200b-200a-429 but not miR-200c-141 promoter remains unclear, it is consistent with the observation that the miR-200b-200a-429 promoter shows higher GC content (>60%) compared to the miR-200c-141 promoter. GC-rich promoters are often targeted by PcG and are commonly protected from CpG methylation (Deaton and Bird, 2011). However, in different types of cancers including breast and prostate, PcG proteins such as EZH2 and SUZ12 are usually found to be elevated (Iliopoulos et al., 2010; Kleer et al., 2003). This may cause aberrant PcG-mediated silencing of non-PcG targets of genes that might contribute to aggressive poorly differentiated cancer. Interestingly, miR-200b was shown to regulate SUZ12 post-transcriptionally in bCSCs (Iliopoulos et al., 2010). This would result in another double-negative feedback loop that adds a further layer of regulation on the miR-200b-200a-429 cluster in addition to the miR-200-ZEB1/2 feedback loop (Bracken et al., 2008). Future experiments are needed to determine if both PRC2 and ZEB1/2 act separately or synergistically as a complex in this setting.

In summary, we have established that the miR-200 gene family plays a functional role in regulating stem-like properties observed in the HMLE cell line model. Moreover, we found that the transcriptional activation and repression of miR-200 family members is maintained by dynamic and distinct epigenetic mechanisms. These findings suggest targeting miR-200 transcripts as a therapeutic means to restore its expression in bCSCs may be of benefit in the treatment of breast cancer.

Materials and Methods

Cell culture

HMLE cells and their derivatives were cultured in HuMEC ready medium (Gibco), and other human breast cancer cell lines were cultured as previously described (Bracken et al., 2008). Cell proliferation was measured using the MTS assay using the CellTiter 96 AQueous Non-radioactive Cell Proliferation Assay (Promega). sl-HMLE MSCV EV, sl-HMLE MSCV miR-200b-200a-429 and sl-HMLE MSCV miR-200c-141 stable cell lines were generated by transducing sl-HMLE with pMSCV-GFP retroviral vectors expressing the primary transcript of miR-200b-200a-429 and miR-200c-141 followed by selection with 1 µg/ml puromycin. Sphere assay was performed as described in (Dontu et al., 2003) with minor modifications. Cells were grown in DMEM∶F12 (1∶1), supplemented with 2% B27 (Invitrogen), 20 ng/ml EGF and 20 ng/ml bFGF (basic fibroblast growth factor; R&D), 4 µg/ml heparin (Sigma), 5 µg/ml insulin, 0.5 µg/ml hydrocortisone. Single cell suspensions were cultured for 10–14 days at a different specified density in 6- or 24-well ultra low attachment plates (Corning, Co-Star) or Polyhema-coated plates (Sigma). Spheres with diameter ≧150 µm were usually counted at day 10–12 to determine the sphere-forming efficiency (SFE). Here, SFE is presented as the number of spheres formed in day 10–14 per number of single cells seeded (spheres/number of cells seeded). Spheres were harvested using 70 µm cell strainers (BD Biosciences) and then dissociated to single cells with trypsin for subsequent passages. Only 20,000 or 10,000 dissociated cells were seeded in subsequent passages. To induce EMT in HMLE cells, the HMLE cells were cultured as described in Mani et al. (Mani et al., 2008) in DMEM∶F12 media (1∶1) supplemented with 10 µg/ml insulin, 20 ng/ml EGF, 0.5 µg/ml hydrocortisone, and 5% fetal calf serum (FCS) and treated with 2.5 ng/ml of TGFβ1 (R&D).

Isolation of cells from human breast cancer pleural and ascites samples

The research using human samples was performed according to the Declaration of Helsinki. Breast cancer pleural or ascites effusion samples were obtained with informed consent from the Royal Adelaide Hospital (RAH Hospital Ethics Committee Protocol No. 081013). Access to patient tumour samples was approved by the appropriate institutional human ethics review boards. Patient details can be obtained at supplementary material Table S1. Patient samples were processed immediately upon received from thoracentesis or paracentesis. Cells were pelleted by centrifuging pleural or ascites fluid mix with acid citrate dextrose (ACD) (5% final concentration). The pellet was washed twice with 1× phosphate-buffered saline (PBS) with 5% ACD and 0.4% human serum albumin (HSA), then incubated in red blood cell lysis buffer (0.15 M ammonium chloride, 10 mM potassium bicarbonate, 0.1 mM EDTA) for 3 min at room temperature. The cells were then washed twice with 1× PBS with 5% ACD and 0.4% HSA before suspension in HuMEC ready medium + 20% FCS. After Ficoll density gradient centrifugation, cells at the interface layer were collected and washed once with 1× PBS with 5% ACD and 0.4% HSA. Cells were counted using white cell fluid dye exclusion. Cells were then cryopreserved at no more than 1×108 cells per vial in 8∶1∶1 ratio of HuMEC ready medium:10% FCS:10% DMSO.

Cell staining and flow cytometry

Cells were stained at a concentration of 1×105 cells per 50 µl or 1×106 cells per 100 µl of 1× PBS with 2% FCS (staining medium). Antibodies were added at appropriate dilution determined from titration experiments. Antibodies included CD45-PeCy5 (1∶10 vol∶vol; 2652U; Beckman Coulter), CD24-PE (1∶5 vol∶vol; 555428; BD Biosciences) or biotin (0.5∶100 vol∶vol; 13-0247; eBiosciences), and CD44-FITC (1∶5 or 10 vol∶vol; 555478; BD Biosciences) or APC (559942; BD Biosciences). Cells were stained for 25 min on ice in the dark and washed with 3–4 ml of staining media each. When biotinylated primary antibodies were used, cells were further stained with streptavidin conjugated fluorophores Pe-Cy7 (0.3125∶100 vol∶vol; 25-4317; eBiosciences) and washed. Cells were analysed using a FACS Aria cell sorter (BD Biosciences) or FC500 Terpsichore (Beckman Coulter) and sorted using a FACSAria cell sorter. Side scatter and forward scatter profiles were used to eliminate debris and cell doublets. Contaminating human CD45 leucocytes cells were eliminated by CD45+ cell exclusion. In cell-sorting experiments, sorted cell populations final purities ranged from 70 to 95%. In some cases, to prevent cell–cell aggregation in single cells suspension, DNase (type IV) (61362; Sigma) was added to the cells in suspension at 50 U/ml. FCS Express 4 software (De Novo) was used to re-analyze the flow cytometry data.

Isolation of RNA and real-time PCR

Total RNA was extracted from cell lines, and real-time PCR was performed by using primers as previously described (Gregory et al., 2008) and as shown in supplementary material Table S2. When small numbers of cells (≤1×105) were obtained especially through cell sorting, RNeasy Plus mini kit (Qiagen) was used to isolate for RNA. MicroRNA PCRs were performed using TaqMan microRNA assays (Applied Biosystems, Foster City, CA). Real-time PCR data for mRNA and microRNA were expressed relative to glyceraldehyde 3-phosphate dehydrogenase (GAPDH) or snRNA U6/RNU6B, respectively.

Western blotting

Western blotting was performed as previously described (Gregory et al., 2008). The following primary antibodies were used: ZEB1 (1∶200 vol∶vol, E20; Santa Cruz), ZEB2 (1∶5000 vol∶vol; Christoffersen et al., 2007), E-cadherin (1∶1000 vol∶vol, 610182; BD Transduction Laboratories), and tubulin (1∶5000 vol∶vol, ab7291; Abcam). Membranes were exposed using enhanced chemiluminescence (GE Healthcare) and imaged using the LAS4000 Luminescent Image Analyzer (Fujifilm).


Cells were seeded on poly-L-lysine and mouse rat tail collagen type I-coated eight-well chamber slides (Nunc) and stained using anti-E-cadherin (1∶500 vol∶vol; as mentioned earlier in the text), ZO-1 (1∶500 vol∶vol; 61-7300 Zymed), CK18 (1∶500 vol∶vol; ab49824 Abcam), CK14 (1∶50 vol∶vol; ab9220 Abcam) or F-actin as previously described (Gregory et al., 2008). Nuclei were visualized by co-staining with DAPI. Cells were examined using an Olympus IX81 microscope, and pictures were taken using a Hamamatsu Orca camera. Images were analysed with Olympus Cell software. All matched samples were photographed (control and test) using identical exposure times.

DNA methylation analysis

Genomic DNA was isolated from cells using either DNeasy Blood and Tissue kit (Qiagen), Trizol, or phenol chloroform ethanol precipitation method (Invitrogen). DNA (0.5–2 µg) was bisulphite modified with the EZ DNA Methylation-Gold Kit according to the manufacturer's protocol (Zymo Research). For melt curve analysis, bisulphite-modified DNA was PCR amplified using primers and conditions that did not discriminate between methylated and unmethylated DNA and did not amplify unmodified DNA (supplementary material Table S3) (Smith et al., 2009). Bisulphite-modified DNA from HMLE (unmethylated reference), M.SssI CpG Methylase treated normal donor lymphocytes (methylated reference), and unmodified DNA from normal donor lymphocytes (negative control) was included in each PCR. The PCR and melt curve was performed using a Rotor-Gene Q (Qiagen) with a 95°C activation step for 15 min; 95°C for 30 s, 55°C for 45 s for 45 cycles; and a final extension step of 72°C for 4 min. The melt of the PCR product was performed from 60 to 90°C, rising in 0.1°C increments, waiting for 30 s at the first step and for 2 s at each step thereafter, and acquiring fluorescence at each temperature increment. The raw melt data were normalized using the HRM Analysis module of the Rotor-Gene 6000 Series Software (Qiagen) as described previously (Smith et al., 2009).

Bisulphite-modified DNA was used for hybridization on Infinium HumanMethylation 450 BeadChip, following the Illumina Infinium HD Methylation protocol, and the BeadChip was scanned using an Illumina HiScan SQ scanner (Illumina, San Diego, CA, USA). The methylation score for each CpG was represented as a β-value according to the fluorescent intensity ratio. β-values may take any value between 0 (non-methylated) and 1 (completely methylated).

ChIP-qPCR assays

ChIP assays using 1×106 cells per reaction were performed as below. Antibodies included anti-histone H3 (ab1791; Abcam), and anti-trimethyl histone H3K4 (ab8580; Abcam), anti-acetyl histone H3K9/14 (06-599; Millipore), anti-acetyl histone H3K27 (07-449; Millipore), anti-trimethyl histone H3K9 (ab8898; Abcam), anti-dimethyl histone H3K9 (ab1220; Abcam), anti-SUZ12 (3737; Cell Signaling), and anti-EZH2 (pAb-039-050; Diagenode). In brief, 6×106 cells were crosslinked in 1% formaldehyde (final concentration) for 10 min at room temperature with gentle rocking or inversion every 2–3 min. Cells were pelleted and washed in ice-cold 1× HBSS (Gibco) containing protease inhibitor mixture (PIC) (Roche). The cells was lysed in 300 µl of lysis buffer (10 mM Tris pH 7.5/1 mM EDTA/1% SDS) containing PIC and incubated on ice for 10 min. After lysis, 900 µl of 1× HBSS containing PIC was added and then aliquoted 200 µl each into six individual tubes. Each 200 µl aliquot was sonicated by using a bioruptor® sonicator (Diagenode), which was empirically determined to give rise to genomic fragments ranging from 200 to 800 bp. The soluble chromatin was collected by 4°C ultracentrifugation (13,000 rpm for 5 min) and pooled into a new 15 ml falcon tube. The supernatant was diluted twofold with 2× RIPA buffer (10 mM Tris-HCl pH 7.5; 1 mM EDTA; 1% Triton X-100; 0.1% SDS; 0.1% sodium deoxycholate; 100 mM NaCl; PIC), 1/10 volume (40 µl) input was removed, and 400 µl of soluble chromatin was distributed to new Eppendorf tubes. Each respective antibody was added at appropriate amount as tested in titration experiments using control promoters. Immunoprecipitations were performed for 2 h at 4°C with rotation, and antibody∶protein∶DNA complexes were then collected with 30 µl of protein A and/or G Dynabeads (Invitrogen) for 2 h of rotation. The beads were washed three times using 200 µl of RIPA buffer and once with TE buffer, then incubated with 200 µl of fresh elution buffer with Proteinase K for 2 h in a thermomixer (1300 rpm, 68°C) to reverse the protein∶DNA cross-links. After incubation, eluates were collected into new Eppendorf tubes. Genomic DNA was recovered by using phenol chloroform extraction and ethanol precipitation. Pellets were washed in 70% ethanol, briefly air-dried, and resuspended in T10E (10 mM Tris pH 7.5; 0.1 mM EDTA) buffer. Quantitation of ChIP DNA (relative enrichment) using the Rotor-Gene 6000 (Corbett Life Science) was carried out in triplicate qRT-PCRs with gene specific primers to assess antibody enrichment (supplementary material Table S4). Enrichment of histone modifications at genomic regions were expressed as the percentage input normalized to respective cell type H3 levels to account for differences in H3 levels in different cell types. [Percentage input was calculated using the formula % (ChIP/Input) = 2[Ct(ChIP)−Ct(Input)×Input dilution factor]×100%, to account for chromatin sample preparation differences].


  • Author contributions

    Y.-Y.L., J.A.W., J.J.A., P.A.G., D.T., A.F.L., P.A.D., Y.K.-G. and G.J.G. conceived and designed experiments. Y.-Y.L., J.A.W., J.J.A., A.G.B. and E.S. performed experiments and analysed the data. Y.-Y.L., J.A.W., J.J.A., P.A.G., Y.K.-G. and G.J.G. wrote the manuscript.

  • Funding

    This work was supported by a fellowship from the National Cancer Foundation Australia [grant number ECF-09-08 to P.A.G.]; a scholarship from the University of Malaya Skim Latihan Akademik IPTA program [to Y.-Y.L.]; and grants from the National Health and Medical Research Council [grant numbers 626918 and 1008440 to G.J.G. and Y.K.-G.]; and the Cancer Council South Australia [grant number 626956 to G.J.G., Y.K.-G. and P.A.G. and 1005078 to P.A.D.].

  • Supplementary material available online at

  • Accepted February 27, 2013.


View Abstract