MicroRNA-regulated gene networks during mammary cell differentiation are associated with breast cancer
Eylem Aydoğdu1, Anne Katchy1, Efrosini Tsouko1, Chin-Yo Lin1, Lars-Arne Haldosén2, Luisa Helguero3 and Cecilia Williams1,*1Department of Biology and Biochemistry: Center for Nuclear Receptors and Cell Signaling, University of Houston, Houston, TX 77204, USA, 2Department of Biosciences and Nutrition, Novum, Karolinska Institute, 141 57 Stockholm, Sweden and 3Department of Chemistry, University of Aveiro, Aveiro, Portugal
*To whom correspondence should be addressed. Email: [email protected] Phone: 832-842-8807 Fax: 713-743-0634
MicroRNAs (miRNAs) play pivotal roles in stem cell biology, dif-ferentiation and oncogenesis and are of high interest as potential breast cancer therapeutics. However, their expression and func-tion during normal mammary differentiation and in breast can-cer remain to be elucidated. In order to identify which miRNAs are involved in mammary differentiation, we thoroughly inves-tigated miRNA expression during functional differentiation of undifferentiated, stem cell-like, murine mammary cells using two different large-scale approaches followed by qPCR. Significant changes in expression of 21 miRNAs were observed in repeated rounds of mammary cell differentiation. The majority, including the miR-200 family and known tumor suppressor miRNAs, was upregulated during differentiation. Only four miRNAs, including oncomiR miR-17, were downregulated. Pathway analysis indi-cated complex interactions between regulated miRNA clusters and major pathways involved in differentiation, proliferation and stem cell maintenance. Comparisons with human breast cancer tumors showed the gene profile from the undifferentiated, stem-like stage clustered with that of poor-prognosis breast cancer. A common nominator in these groups was the E2F pathway, which was overrepresented among genes targeted by the differentiation-induced miRNAs. A subset of miRNAs could further discriminate between human non-cancer and breast cancer cell lines, and miR-200a/miR-200b, miR-146b and miR-148a were specifically down-regulated in triple-negative breast cancer cells. We show that miR-200a/miR-200b can inhibit epithelial–mesenchymal transi-tion (EMT)-characteristic morphological changes in undiffer-entiated, non-tumorigenic mammary cells. Our studies propose EphA2 as a novel and important target gene for miR-200a. In con-clusion, we present evidentiary data on how miRNAs are involved in mammary cell differentiation and indicate their related roles in breast cancer.
Introduction
In the clinic, breast cancer is classified into different subtypes depend-ing on the expression of three receptors, the estrogen (ERα), proges-terone (PR) and Her2 receptors. Additional subtypes can be classified using molecular profiling (1), and most triple-negative breast cancers belong to the basal-like subtype. Targeted treatments against ERα or PR or Her2 positive tumors are available. Triple-negative breast tumors, defined by a lack of expression of these receptors, relapse faster than other types of breast cancer and have a higher prevalence among younger women. Despite a high mortality rate, few effective treatment options currently exist for this diagnosis. Current research into breast cancer treatment is aimed at designing disease subtype–
specific treatments. The first critical step is to identify a suitable target. Small non-coding RNAs have emerged as major gene expression reg-ulators. There is an increasing body of evidence showing that microR-NAs (miRNAs) play pivotal roles in stem cell biology, differentiation and oncogenesis (2). Tumor suppressor or oncogenic miRNAs have been described, and dysregulation in their expressions causes tumo-rigenesis (3,4). Their small size and their potential to target multiple pathways simultaneously make them promising candidates for cancer therapeutics (5–7). In order to exploit miRNAs for breast cancer ther-apy, an in-depth understanding of their normal expression and func-tion is essential. How miRNAs impact mammary differentiation or mammary stem cell characteristics have not been thoroughly defined. Reports have compared fluorescence-activated cell sorting-enriched normal mammary stem cells from the mouse mammary gland and human fluorescence-activated cell sorting-enriched breast cancer stem cells (CSC) from clinical tumors with remaining breast cancer cell populations (8). Others have studied enriched CSC populations of MCF-7 breast cancer cells (9) or Sca-1+ populations of the mouse mammary epithelial progenitor cell line, Comma-D (10,11). Some discrepancies exist between the different studies, possibly related to the characteristics of the cell line(s) or cell populations used, and no large-scale correlation of differentiation-regulated miRNAs in mam-mary cells, and their relation to transcriptome changes has been per-formed. To thoroughly characterize the miRNAs that may be involved in maintaining an undifferentiated mammary phenotype or in inducing terminal differentiation, we took advantage of the HC11 non-tumor-igenic murine mammary epithelial cell line. HC11 cells approximate both self-renewal and bipotency and can be cultured for an unlimited number of passages in a proliferating undifferentiating phase. They have the capacity to enter terminal differentiation, indicated by their ability to express milk proteins in vitro (12). Illustrative of their stem cell abilities, HC11 cells can reconstitute the ductal epithelium of a cleared mammary fat pad in vivo with ductal, alveolar and myoepithe-lial cells (12,13). We have shown previously that the undifferentiated stage of HC11 cells share conserved, overlapping transcriptome sig-natures with poor-prognosis and basal-like breast cancer, suggesting relevance for this model in mammary stem cell and breast cancer–related studies (14). We here describe the identification of a set of miRNAs with roles in mammary differentiation that we suggest are therapeutically useful.
Material and methods
Cell cultureHC11 mammary epithelial cells were grown in RPMI 1640 supplemented with 10% fetal bovine serum, l-glutamine, 5 mg/ml insulin, 10 ng/ml epider-mal growth factor (EGF) and 50 mg/ml gentamicin; all cell culture reagents were purchased from Sigma. Proliferating cells were collected under the same conditions. Predifferentiated and fully differentiated HC11 cells were acquired sequentially after induction of differentiation as described previously (14). In short, predifferentiation was induced by the removal of EGF from the medium and lowering of fetal bovine serum to 2% for 48 h. Full differentiation was accomplished through the subsequent addition of 100 nM dexamethasone and 1 µg/ml ovine prolactin for 72 h.
RNA extractionTotal RNA, including miRNA, was extracted using Trizol precipitation and purified using Qiagen miRNeasy kit followed by an on-column DNase1 diges-tion (Qiagen, Valencia, CA). Quantitative and qualitative analysis of RNA was performed on the NanoDrop 1000 spectrophotometer (Thermoscientific) and the Agilent 2100 bioanalyzer (Agilent), respectively.
miRNA profiling analysismiRNA expression profiles of proliferating and differentiated cells were deter-mined in three independent differentiation assays. Two different large-scale profiling methodologies, miRNA microarrays and TaqMan® Low-Density
Array (TLDA), were applied to eliminate shortcomings of each technique. The miRNA microarray contained 700 unique probes and 49 controls with replicates. TLDA enabled the quantification of 468 unique mouse miRNAs (335 in card A and 133 in card B) along with three endogenous and negative controls. miRNA microarray. Mouse miRNA two color dual-sample arrays based on Sanger miRBase release 14.0 (LC-Sciences, Houston, TX) were used to determine the miRNA profile between undifferentiated and differentiated cells. Five milligrams of total RNA per sample were used for each microar-ray. Experiments were performed in duplicate with dye swap design. miRNAs were determined as potentially regulated when P < 0.1, according to the manu-facturer’s recommendation.
TLDA. miRNA expression profiles of two differentiation assays were com-piled using TaqMan® Rodent MiRNA A + B Cards Set v2.0 according to the manufacturer’s protocol without the preamplification step (Applied Biosys-tems, Foster City, CA) using 7900HT Fast Real-Time PCR System (Applied Biosystems, Foster City, CA). In short, 1 µg of total RNA, including the miRNA population, was used to perform the reverse transcription reaction in total volume of 7.5 microliters, and 6 microliters of reverse transcription prod-uct were used to perform real-time PCR. miRNAs were considered differen-tially expressed if in both replicates fold change >1.3 or <0.7, respectively. miRNAs that were determined regulated by both microarray and TLDA were validated using qPCR. In addition, a selection of miRNAs, whose probes were present only in one of the large-scale platforms and showed regulation in both replicates, were added to the qPCR analysis (Supplementary Data 1–4 and Supplementary Tables 1–2 are available at Carcinogenesis Online).
cDNA synthesis and real-time PCR for miRNA and mRNA analysisFor miRNA analysis, 1 mg of isolated total RNA per sample was polyade-nylated using Ncode miRNA First-Strand cDNA Synthesis Kit (Invitrogen). Subsequent first strand cDNA synthesis was performed immediately using Superscript III and universal reverse transcription primers according to man-ufacturer’s protocol (Invitrogen). SYBR-Green qPCR SuperMix (Applied Biosystems, Foster City, CA) was used for real-time qPCR with 16 ng of cDNA, 2 pmol of universal qPCR primer and 2 pmol of miRNA-specific forward primer per 10 µl of final reaction volume. All SYBR-Green amplifi-cations were checked with melting curve analysis. Only amplifications with one defined melting curve were used for further analysis. TaqMan miRNA assays were used for miRNAs that could not be specifically amplified using SYBR-Green technology. The two-step assay was carried out according to manufacturer’s protocol (Applied Biosystems, Foster City, CA) and 10 ng total RNA was used per cDNA synthesis. For mRNA analysis, cDNA was synthesized using Superscript III (Invitrogen) and random hexamer prim-ers as described previously (14). All performed SYBR-Green qPCR ampli-fications were checked with melting curve analysis. Primer sequences are provided on request. Both SYBR-Green and TaqMan miRNA assays were performed using ABI PRISM 7500 (Applied Biosystems, Foster City, CA) and fold changes were calculated using the ∆∆CT formula. U6 snRNA was used as the reference gene for all miRNAs, whereas glyceraldehyde 3-phosphate dehydrogenase and 18S were used for mRNA qPCRs. Unpaired two-tailed t-tests were used to compare differences between two parallel treatment groups of identical origin. Significance is presented as *P < 0.05, **P < 0.01, or ***P < 0.001.
Indirect immunofluorescence and image acquisitionImmunostaining was carried out on cells grown on permanox two-well cham-ber slides (Nunc) and transfected as described below. Following 24-h treat-ment, cells were fixed in 4% buffered formalin for 20 min and permeabilized with 0.5% Triton X-100 for 10 min. After several washes with phosphate-buff-ered saline, unspecific binding was blocked with 15% fetal bovine serum in
0.1% Tween 20-phosphate-buffered saline and E-cadherin staining was car-ried out over night with a mouse anti-E-cadherin antibody (BD Transduction Labs), followed by antimouse Alexa 488 antibody (Invitrogen). The slides were mounted in ProGold anti-fade medium (Molecular Bioprobes). Images were acquired using the same settings and when edited with Adobe Photoshop 6.0, the same adjustments were applied to all images.
BioinformaticsAriadne Pathway Studio was used for analysis of networks and pathways. Fis-cher’s exact test is applied to define enriched Gene Ontology functional groups among the differentially expressed genes. Results were considered statistically significant for P ≤ 0.05. The Sub-Network Enrichment Analysis algorithm was used to define overrepresented subnetworks. Results were considered statisti-cally significant for P ≤ 0.05. Values are expressed as means with 95% con-fidence intervals. TargetScan was employed to predict targets of differentially expressed miRNAs, and the regulation of target genes via cell differentiation is obtained by comparison with the genome-wide mRNA analysis.
Survival cluster analysisTo analyze the effect of undifferentiated HC11 gene signature in breast cancer patient samples, the list of significant murine genes from the HC11 cell stud-ies were matched to their human orthologs. Expression profiles of these genes were then used to group 258 clinical, untreated, primary breast cancer samples from a patient cohort from Uppsala, Sweden (15), by hierarchical clustering (Eisen Cluster and TreeView). Disease-free survivals of the two patient groups were graphed using the Kaplan–Meier plot function in the SigmaPlot software, and associations with disease parameters (ERα/PR status, lymph node positiv-ity, tumor grade, recurrence, distant metastasis, death) were calculated using the Fisher’s exact test.
miRNA mimic transfectionsHC11 cells were transfected with miRIDIAN miRNA Mimic (Dharmacon, Lafayette, CO). A Dy547-labeled transfection control was used to determine optimal conditions and visualize transfection efficiency. Images were taken 24 h after transfection initiation. Cell density and transfection agent (Dharma-FECT 1) concentrations were optimized for HC11 cells: 5000 cells/well and 300 000 cells/well in 96 and 6 well plates, respectively. All mimic and control transfections were then performed using 25 nM final concentration of mimic and 0.1% DharmaFECT 1. qPCR analysis of direct and downstream target genes was performed 24 h posttransfection.
Results
Molecular mechanisms involved in stem cell maintenance or differ-entiation frequently play important roles in cancer development. In this study, we aimed to identify the miRNA subsets that are selec-tively expressed in undifferentiated mammary epithelial cells com-pared with their functionally differentiated counterparts. The model we used, the HC11 cell line, is a non-tumorigenic, undifferentiated mammary epithelial cell line that originates from BALB/c mice in mid-pregnancy (12). Mid-pregnancy is a period of mammary stem cell expansion (16). These cells also possess a mutation in the p53 gene, potentially further increasing the replicative potential of stem cells (17). In addition, the p53 mutation confers Li Fraumeini char-acteristics and a predisposition to mammary carcinoma. The stage that comprises undifferentiated cells exhibits stem cell characteris-tics and shares a gene signature with stem cells (14), we therefore refer to this stage as ‘stem cell-like’ (SC-like) in this study. We have previously defined the exact gene expression profiles of HC11 cells for each differentiation stage (14). In order to investigate the expression and roles of miRNAs in the corresponding differentia-tion stages, we used the approach presented in Figure 1A.
HC11 murine mammary epithelial cells showed complete and robust differentiation patternTo identify changes in miRNA expression, we prepared three repli-cates of HC11 cell differentiation experiments. Each replicate’s appro-priate differentiation stage was confirmed using previously defined markers (14) as exemplified in Figure 1B–D. Expression indicative of the undifferentiated stage of these cells including a high level of Melk (Figure 1A and B) that is also highly expressed in breast CSCs (18), and a high level of Birc5 (Figure 1C), an inhibitor of apoptosis. The fully differentiated cells showed high expression of, for example,
Table I. Tumors with HC11 SC-like gene signature are more metastatic
Disease parameters and outcome of breast tumors-HC11 SC-like gene profile cluster analysis. Tumors clustering with HC11 SC-like gene profile, cluster 2 (in Figure 5), have a significantly higher rate of high-grade tumors, ERα negativity, lymph node positivity, recurrence, distant metastasis and death.v
epithelium-specific Ets-domain-containing transcription factor, Ehf (Figure 1D). In addition, beta-casein protein is expressed in the fully differentiated stage (Figure 1A). These results clearly demonstrate the robust and reproducible gene regulation during the induced differen-tiation of HC11 cells.
miRNAome was regulated during differentiationThe distinct transcriptional changes observed during the differen-tiation of HC11 cells made it plausible to assume that the same approach could detect corresponding changes in miRNA levels. The short length (~22 nt) of mature miRNAs and the additional
presence of pri-/pre-miRNAs make miRNA analysis less straight-forward than conventional mRNA analysis. To identify changes in the miRNAome, we used two complementary large-scale screening techniques: miRNA microarrays and TLDA. We investigated the changes in miRNA expression between the undifferentiated SC-like stage and the fully differentiated stage (Supplementary Tables 1–2 are available at Carcinogenesis Online). Indicated changes were fur-ther analyzed using qPCR in the three different stages of differen-tiation: SC-like, predifferentiated and fully differentiated, in three replicated differentiation experiments. We identified and confirmed significant regulation of 21 miRNAs during the differentiation
Fig. 1. Experimental setup for analysis of miRNA expression during differentiation of HC11 cells. (A) HC11 cell differentiation model: HC11 cells grown in the undifferentiated stage were induced to a committed, predifferentiated stage by withdrawal of EGF and reduction of serum to 2% for 48 h. These cells were subsequently induced to complete differentiation by addition of prolactin and dexamethasone for 72 h. The differentiation process was carried out in triplicates using different cell passages. Each stage of differentiation was confirmed by immunofluorescence analysis of protein expression levels of the stem cell marker Melk and the milk protein beta-casein. (B, C, D) Expression of marker genes throughout differentiation of HC11 cells: mRNA levels of Melk, Birc5 and Ehf analyzed with qPCR at each differentiation stage. Melk and Birc5 are examples of SC-like stage markers and Ehf of the differentiated stage. Black, light gray and gray represent three differentiation replicates, respectively. ***P < 0.001. Part of A (schematic model of HC11 cell differentiation in middle panel) inspired by (40).
process of mammary SC-like cells, qPCR data are shown in Fig-ure 2. Four miRNAs were downregulated during the differentia-tion process (Figure 2A), whereas the remaining 17 miRNAs were upregulated (Figure 2B). We note that the miR-200a/miR-200b, miR-148a/miR-152 and miR-30b/d families, together with miR-146b and miR-21, were strongly upregulated during initiation of differentiation, prior to terminal differentiation. Further, miR-200a was the most regulated miRNA, exhibiting a 160-fold induction (Figure 2B). miR-200a belongs to the same family as miR-200b (10-fold induced), which together with miR-205 (6-fold induced) are known to play an important role in the regulation of epithelial–mesenchymal transition (EMT) (19,20) and in mammosphere for-mation (8,9,20). Identification of miRNAs with known functions in mammary stem cells (miR-200 family) provided evidence for the reliability of the analysis and model. Alongside previously reported miRNAs, we identified a number of miRNAs previously unassoci-ated with mammary differentiation.
miR-200a/miR-200b downregulated Zeb1, Zeb2 and Suz12 and induced E-cadherin in HC11 SC-like cellsmiR-200a and miR-200b have been shown to play a number of critical functions in mammary stem cells (8,9,20). They have been shown to inhibit EMT, which is important during development, progression and metastasis (21). A low expression of E-cadherin is indicative of EMT. The miR-200a/miR-200b have been shown to increase E-cadherin expression through downregulation of its transcriptional repressors Zeb1, Zeb2 and the polycomb complex Suz12 (19,20,22), as illus-trated in Figure 3A. In the HC11 differentiation model, the upregula-tion of miR-200a/miR-200b during differentiation correlated with an increase in E-cadherin mRNA levels. In order to ascertain whether the observed upregulation was a consequence of increased miR-200a and/or miR-200b levels, we independently transfected miR-200a and miR-200b mimics into the SC-like, undifferentiated stage. We observed that both the miR-200a and the miR-200b mimic medi-ated downregulation of Zeb1, Zeb2 and Suz12 and thereby induced E- cadherin transcription (Figure 3B). Immunoflourescent assay illus-trates that miR-200b mimic transfection increased E-cadherin stain-ing and E-cadherin membrane localization in HC11 cells, whereas cells transfected with negative control mimic showed less intercellular E-cadherin, a cytoplasmic dotted pattern of staining and more cell spreading (Figure 3E). The latter are all characteristic features of cells undergoing EMT. Here, we show for the first time that introduction of miR-200a or miR-200b to undifferentiated HC11 cells increase levels of E-cadherin mRNA, thus partly mimicking the normal dif-ferentiation pattern (Figure 3B). Introduction of miR-200b in the undifferentiated HC11 cells also resulted in morphological changes, supporting a corresponding inhibition of EMT, as illustrated in
Figure 3E. miR-200b–transfected cells showed more epithelial mor-phology, the cells tend to group together forming more tightly packed clusters and show a clearer polyhedral morphology compared with negative control. The finding that miR-200a and miR-200b can be used to inhibit EMT characteristics in undifferentiated, highly migra-tory HC11 cells is of significant therapeutic importance.
EphA2 mRNA is downregulated by miR-200a expressionmiR-200a was found to be the most highly regulated miRNA in our study. It has a known role in targeting Zeb1, Zeb2 and Suz12 and thereby affecting EMT in mammary cells. However, miRNAs usu-ally target multiple genes and it is possible that miR-200a has addi-tional roles. When miRNAs form duplexes with the 3ʹ untranslated region end of their target mRNAs, they inhibit the translation of these genes and frequently also degrade the targeted mRNA (23). Given this, we used our previously defined HC11 transcriptome data (14) in combination with bioinformatic predictions to characterize putative miRNA:mRNA functional pairs (Supplementary Table 3, available at Carcinogenesis). This led to the prediction that miR-200a targets the ephrin receptor of the protein–tyrosine kinase family (EphA2). EphA2 has been shown to regulate mammary gland branching (24) and to promote mammary tumorigenesis and metastatic progression in mice (25,26). During differentiation of the HC11 cells, the mRNA level of EphA2 was downregulated concomitant with the upregulation of miR200a (Figure 3C and Figure 2B). Using miR-200 mimics, we showed that miR-200a, but not miR-200b, expression downregulated EphA2 in transfected HC11 cells (Figure 3D). That EphA2 is down-regulated as a consequence of miR-200a expression is a novel finding with major clinical implications.
Stem cell and oncogenic networks are targeted by regulated miRNAsThe low miR-200a/miR-200b expression in the SC-like stage correlates with reports of the miR-200 family being downregulated in both normal and breast CSCs (8,9). Their overexpression has been shown to block formation of mammospheres in both human and murine cells (20) and to suppress expression of stem cell factors (27), indicative of an ability to inhibit parallel stem cell mechanisms. We assumed that these and the remaining identified miRNAs play additional roles during differentiation and used bioinformatic approaches to generate an overview of such potential interactions. This allowed the construction of an integrated network of genes and miRNAs regulated during HC11 cell differentiation, and their known functions as reported in literature. These networks were then screened for overrepresented biological pathways and interactions. Figure 4 and Supplementary Figure 1 (available at Carcinogenesis Online) illustrate identified overrepresented pathways, the Wnt, Stat and E2F-pathways.
Fig. 2. miRNA levels during HC11 cell differentiation transitions. qPCR confirmation of differentially regulated miRNAs identified by large-scale profiling. Black, light gray and gray represent SC-like, predifferentiated and fully differentiated stages, respectively. U6 snRNA was used as reference gene. (A) miRNAs highly expressed in the SC-like stage of HC11 cells. (B) miRNAs highly expressed in the differentiated stage of HC11 cells. P values reflect comparisons between the SC-like stage and fully differentiated stage. *P < 0.05, **P < 0.01 and *** P < 0.001.
We note that many of the differentiation-induced miRNAs are considered tumor suppressors (28). These includes the let-7 family, which is downregulated in cancer cells and further downregulated in CSCs (5,20), and the 80-fold upregulated miR-146b (Figure 2B), which has been reported to suppress breast cancer metastasis (29). Our network analysis indicated that several of the differentiation stage miRNAs (e.g. miR-148a, miR-200b, miR-27b, miR-205 and miR-146a/b) target different oncogenes (Met, Ran, Maff and Dek). In addition, miR-200b targets several members of pathways with essential roles in stem cell self-renewal. We predicted targets in the Wnt, Hedgehog and Notch signaling pathways (Prkca, Met and Lfng, all subsequently downregulated during differentiation, Supplementary Table 3, available at Carcinogenesis Online). Figure 4A illustrates an intricate network of Wnt and stem cell–related genes, oncogenes and miRNA interactions inferred from the gene expression analysis and bioinformatic predictions. Members of the Wnt pathway are targeted by miR-30b, miR-200b and miR-148a. Myc, which interacts with Wnt, is also affected through multiple pathways involving miRNAs
miR-30b/d, miR-205 and miR-200b. Further, a downstream product of Wnt signaling activation, Serpine1, is targeted by both miR-148a and miR-181a, and numerous genes affecting Serpine1 expression are simultaneously targeted by several miRNAs (including miR-148a, miR-27b, miR-200b, miR-205, miR-26a and miR-181a). Serpine1 is a member of the serine proteinase inhibitor family, which are reported stem cell markers (30), and was found to be highly expressed in the SC-like stage of the HC11 cells (Figure 4C). This illustrates how multiple differentiation-induced miRNAs can work together in targeting connected pathways implicated in breast cancer.
We also note known oncomiRs among the four miRNAs upregulated in the SC-like stage. miR-17 is a potential oncomiR in several cancers (31) and miR-206 has been shown to contribute to abrogation of estro-genic responses in breast cancer cells and to contribute to a basal-like phenotypic switch (32). We predict miR-17 to target the tumor sup-pressor genes Dab2 and Celsr2. These genes were upregulated during differentiation (14) concomitant with miR-17 downregulation. Dab2 is a putative tumor suppressor that inhibits proliferation in ovarian
Fig. 3. EMT genes regulated by miR-200a and miR-200b in HC11 cells. (A) The E-cadherin pathway targeted by miR-200 family. (B) qPCR analysis of Zeb1, Zeb2, Suz12 and E-cadherin transcript levels in miR-200a– or miR-200b–transfected HC11 cells. (C) mRNA expression level of EphA2 throughout HC11 cell differentiation process analyzed by qPCR. (D) qPCR analysis of EphA2 transcript levels in miR-200a– or miR-200b–transfected HC11 cells. (E) Immunoflouresence staining of E-cadherin in miR-200b–transfected HC11 cells. miR-200b expression led to a more epithelial morphology and where cells tended to group together forming more tightly packed clusters, it led to a clearer polyhedral morphology compared with negative control. *P < 0.05, **P < 0.01 and ***P < 0.001.
cancer (33), negatively regulates Wnt signaling (34), and is needed for embryonic stem cell differentiation (35). This gene has not previously been reported as having a role in mammary differentiation or breast cancer. In addition, we predict that miR-17 targets both Stat3 and the Stat3-regulated gene Pik3r1 (Figure 4B). The Stat3 pathway has been observed to be transcriptionally downregulated in the SC-like stage (14). Although activation of Stat3 through phosphorylation is required for self-renewal of embryonic stem cells, we note that its transcript is downregulated (Figure 4D) along with its downstream target genes Pik3r1, Igfbp5, Mxd4 and Cdkn1b in the SC-like stage (as illustrated in Figure 4B). In addition, miR-206 appears to work together with miR-17 in affecting the Stat3 pathway, targeting the Stat3 downstream fac-tor Mxd4 (Figure 4B). Dnmt1, a repressor of Stat3 and Stat1, is highly expressed in the SC-like stage, but appears to be targeted by miR-152 and miR-148a upon differentiation, thus allowing the upregulation of Stat transcripts. We predict that the expression of miR-17 and miR-206,
and the lack of expression of miR-152 and miR-148a, are important in maintaining the low transcript levels of the Stat pathway–associated genes in undifferentiated HC11 cells. These novel correlations can help to define future functional studies that will need to be performed for miRNAs identified in this study.
A miRNA targeted 152-gene profile delineated good- and poor-prognosis patientsDeregulation of developmental and stem cell critical pathways, as exemplified above, is implicated in cancer development. We have previously observed a correlation of HC11 SC-like gene expression and that of basal-like and poor-prognosis breast tumors (14). To explore the relation between the set of miRNAs identified and poor-prognosis breast cancer, we first used unsupervised clustering between the HC11 SC-like gene expression profile and that of 258 patients’ primary and untreated breast cancer tumors (Figure 5A).
Fig. 4. miRNAs target stem cell signaling pathways. (A) miRNAs target WNT, E2F pathways and oncogenes. Pathways were assembled based on bioinformatics and literature, combined with biological interpretation of the microarray data, enriched Gene Ontology functional groups and subnetworks among the differentially expressed genes. Red: entities highly expressed in HC11 SC-like stage. Blue: Entities highly expressed in HC11 fully differentiated stage. The darker the shade, the stronger the regulation. (B) The Stat pathway is downregulated in SC-like stage of HC11 cells and targeted by miRNAs highly expressed in SC-like stage. Data assembled and illustrated as in (A). (C, D) Transcript levels of Serpine1 and Stat3 genes, respectively, in HC11 SC-like predifferentiated and fully differentiated cells, as determined using qPCR.
Tumors clustering with the HC11 SC-like gene expression profile showed significantly lower survival (P 5 0.009) as shown in Figure 5B. These tumors also showed a significantly higher rate of distant metastases (P 5 0.007) and of higher grade tumors (P 5 7 10−14), Table I. This further supports the correlation between this model of SC-like gene expression and that of poor-prognosis breast cancer. We focused on the 152-gene profile that delineated good- and poor-prognosis patient clusters (yellow frame in Figure 5A) and evaluated this profile in its biological context. Here, the Hedgehog pathway (P 5 3 10−6), and the subnetworks of Melk (P 5 0.00007), FoxM1 (P 5 3 10−5) and the E2F pathway (P 5 4 10−29) were significantly overrepresented (Supplementary Table 4, available at Carcinogenesis Online). These pathways are all implicated in both stem cell and breast cancer biology. We further investigated this 152-gene profile for miRNA regulation and found that 11 genes were predicted targets of miRNAs implicated in the HC11 SC-like
differentiation (Supplementary Table 5, available at Carcinogenesis Online). Five (E2F1, E2F7, Cdc7, Cdc25A, Dnmt1) out of the eleven genes (45%) were associated with the E2F pathway. The importance of the E2F pathway in breast cancer is supported by reports of elevated E2F activity in basal-like breast tumors (36). Our findings suggested that the E2F pathway is the primary pathway shared between undifferentiated HC11 cells and poor-prognosis tumors, and a main target for differentiation-induced miRNAs in mammary cells. E2F1 is known as an oncogenic transcription factor that activates genes necessary for tumor cell proliferation (37). We also note that the stem cell–related factor Serpine1 is transcriptionally regulated by three E2F-regulated transcription factors (Myc, Etv4, Nr4a1/Nur77) (Figure 4A). Thus, we have identified that the E2F pathway is a common characteristic between the undifferentiated HC11 cells and that of poor-prognosis tumors and propose that this pathway is targeted by miRNA during mammary differentiation. These data imply that miRNAs targeting the E2F pathway may be indicative of a good prognosis in breast cancer patients and may be investigated for future biomarker and therapeutic applications.
An identified set of miRNAs can differentiate between human non-tumor and breast cancer cellsSince the identified miRNAs are implicated in breast cancer, we proceeded to analyze their levels in human breast cancer cell lines of different tumor subtypes. We found that 6 of the 17 differentiation-induced miRNAs could differentiate between cancer and non-cancerous cell lines, and an additional four miRNAs were specifically downregulated in triple-negative breast cancer cells. let-7c, miR-27b, miR-22, miR-143, miR-30b and miR-30d showed significantly decreased expression levels in all breast cancer cell lines (ERα-positive MCF-7 and T47D and triple-negative MDA-MB-231) compared with non-tumorigenic epithelial MCF-10A cells (Figure 6A). Low let-7 expression has been previously associated with cancer (38,39) and let-7 miRNA has been defined as a tumor suppressor (39). Several of the remaining miRNAs have as yet unknown functions in breast cancer. The shared miRNA signature presented here implies conserved and related functions of these six miRNAs in murine mammary development and human breast cancer. A second group of miRNAs (miR-200a, miR-200b, miR-146b and miR-148a) was significantly downregulated in the triple-negative MDA-MB-231 cells, compared with both the non-tumorigenic MCF10A and the ERα-positive breast cancer cell lines (Figure 6B). Interestingly, these include two of the three miRNAs predicted in targeting the E2F pathway that is active in poor-prognosis and metastatic breast cancer, miR-200b and miR-148a. Our findings suggest that the absence of miR-200a, miR-200b, miR-146b and miR-148a specifically correlates with triple-negative breast cancer markers and hence their therapeutic delivery may constitute a future approach.
Discussion
In this study, we have identified a set of miRNAs with potential roles in mammary differentiation, stem cell maintenance, breast carcino-genesis, as well as for clinicopathological parameters and breast can-cer survival. We show that their expression patterns can distinguish between cell lines of different breast cancer subtypes. Furthermore, we show that miR-200a and miR-200b can inhibit EMT characteris-tics in the HC11 undifferentiated cells, and that miR-200a mediates downregulation of the important breast oncogenic receptor, EphA2. Our data sheds light on the roles of miRNAs and their networks in the maintenance of an undifferentiated phenotype, during induction of differentiation, and in breast carcinogenesis.
Our primary aim was to study the involvement of miRNA in the mammary differentiation program, and we found many correlations to known stem cell mechanisms in the process. The exact identity of mammary stem cells is not fully known, and no markers exist that can distinguish between mammary stem cells and the common progenitor cells (40). Our study highlights miRNAs involved in
Fig. 5. Clustering of breast cancer patients using the expression profile of HC11 SC-like genes. Transcription profiles obtained from 258 primary and untreated breast tumors clustered with HC11 SC-like gene profile. (A) The tumors could be clustered into two groups based on the HC11 expression profile. The yellow frame indicates the 152 genes responsible for the cluster division. (B) Kaplan–Meier plots of disease recurrence for clusters 1 (red) and 2 (green) patients. Patients in cluster 2 showed a significantly lower survival (P 5 0.009).
the differentiation process, several of which may also participate in maintaining an undifferentiated state. As HC11 cells stop proliferating during functional differentiation, these miRNAs may also be involved in the regulation of proliferation. However, we identified the majority of miRNA regulations at the transition from an undifferentiated state into a committed, and still proliferating, predifferentiated stage. We could also identify characteristic expressions of these miRNAs in different human cell lines, all of which are proliferating. Therefore, we do not expect that the primary role of these miRNAs is to affect proliferation.
We found miR-200a to be the most strongly upregulated miRNA during differentiation, and also that a second member of the same family, miR-200b, was strongly induced. A third member, miR-200c, that is known as a potential molecular link between breast CSCs and normal mammary stem cells (8), was, however, not differentially expressed in our study. miR-200c shares seed sequence and target genes with miR-200b, and it is possible that its functions is carried out solely by miR-200b in the HC11 cells. The HC11 cells undergo EMT in the SC-like stage, as evidenced by their morphology, vimen-tin expression, and lower E-cadherin levels compared with differenti-ated cells (14). These EMT characteristics disappear when the cells enter the predifferentiated stage, coinciding with the upregulation of miR-200a/miR-200b. By expressing miR-200a/miR-200b mimics, we confirmed their EMT inhibiting mechanism of downregulating Zeb1, Zeb2 and Suz12 and upregulating E-cadherin in the SC-like stage of HC11 cells. We also showed that delivery of miR-200b into cells in the SC-like stage consequently changed their morphology. In addition, a novel function of miR-200a in mammary cells were identified: its expression downregulates the mammary oncogenic receptor EphA2.
Mechanisms that are conserved between species are likely to play pivotal roles in the healthy and diseased states. The correla-tions between this murine SC-like model and human breast can-cer prompted us to analyze our differentially expressed miRNAs in human breast cancer cell lines. Our findings detail two different miRNA expression profiles in breast cancer cell lines. Four miRNAs, including miR-200a/miR-200b, show unique downregulation in the triple-negative MDA-MB-231 cells, but not in ERα-positive MCF-7 and T47D breast cancer cells (Figure 6B). Different forms of breast cancer have been speculated to develop from different ‘cells of ori-gin’ (40,41). For example, there are indications that basal-like breast cancers with BRCA1 mutations originate from mammary progeni-tor cells (42). Several studies have also shown that basal-like tumors share many features with undifferentiated mammary cells (14,43). It is thus possible that the differential miRNA expressions between the cell lines are reflective of such differences in cell origin. They may also reflect different levels of differentiation of the cell lines. It is interesting to note that three out of these four miRNAs specifically downregulated in the triple-negative cell line (miR-200a, miR-146b
and miR-148a) showed the most significant upregulation (>50-fold) during HC11 differentiation (Figure 2B) and that two were further implicated in targeting the poor-prognosis E2F pathway. A second group of miRNAs was downregulated in all three breast cancer cell lines compared with non-tumorigenic mammary epithelial cells (Figure 6A), and we speculate that these might be useful candidates for breast cancer marker evaluation.
Our data suggest that the E2F activity is imperative for the similarities between the undifferentiated SC-like HC11 cells and that of poor-prognosis breast cancer. We have previously proposed that HC11 cells resemble the precursor cells for basal-like breast cancer and may provide a suitable model for understanding basal-like breast cancer and its potential relation and conversion from normal stem or progenitor cells. A related observation was made in other studies that found a correlation between gene expression of enriched human stem cells from mammospheres with undifferentiated breast cancer tumors (43). In the present study, we observed lower survival rates for breast cancer patients whose tumors exhibited an SC-like profile, and that the E2F pathway was implicated. This pathway was also specifically targeted by differentiation-induced miRNAs. We predict that these miRNAs will be downregulated in poor-prognosis, basal-like tumors. Indeed, we could show that the miRNAs targeting the E2F pathway, miR-200b and miR-148a, were selectively downregulated in a basal-like, metastatic breast cancer cell line. The E2F pathway is one of several EGF-induced signal transduction cascades. HC11 cells require EGF treatment to maintain their undifferentiated SC-like phenotype. EGF has also been shown to convert neurogenic precursors in the adult brain into multipotent stem cells, indicative of an involvement in stem cell maintenance (44). We note that the major change in miRNA expression occurred in the transition between the SC-like stage and the committed predifferentiated stage (Figure 2A, B), not between committed and fully differentiated cells. The predifferentiated stage is induced by EGF withdrawal (Figure 1A). Consequently, we speculate that EGF withdrawal may induce the cascades regulating these miRNAs. The differential expression of miRNAs was further enhanced during the second transition, from predifferentiated to fully differentiated cells (Figure 2A, B), without any additional changes in EGF levels. As mentioned above, the miRNAs that may be induced as a result of EGF withdrawal, in turn downregulate the EGF–E2F pathway. Thus, it is plausible that EGF withdrawal induces regulations of miRNAs, which attenuate the remaining EGF cascade activity in a feed-forward loop. Most basal-like tumors have high EGF signaling (45). EGF signaling and its downstream pathways may be the common denominator between the HC11 SC-like cells and basal-like breast tumors, and potentially, between mammary stem or progenitor cells and breast CSCs.
Fig. 6. Expression levels of miRNAs in breast cancer and non-tumor breast cell lines. qPCR analysis of miRNAs in ERα-positive MCF-7 and T47D, and triple-negative MDA-MB-231 breast cancer cells, relative to their levels in non-tumor MCF-10A cell line. U6 snRNA was used for normalization of expression. (A) miRNAs with decreased levels of expression in all three breast cancer cell lines. (B) miRNAs with decreased levels of expression in the triple-negative breast cancer cell line, MDA-MB-231.
This study identifies distinct changes in miRNA expression during dif-ferentiation of mammary SC-like cells. By assembling transcriptome analysis and miRNA expression data with predicted targets and path-way information, an intricate gene regulation network emerged. We demonstrate that the miRNA targeted pathways are conserved in poor-prognosis breast cancers and show that these miRNAs are also differen-tially expressed in human breast cancer cells. We show that miR-200a and miR-200b inhibit EMT characteristics in the undifferentiated, non-tumorigenic HC11 cells and that miR-200a mediates repression of EphA2. Candidate miRNAs are identified, which we propose should be further studied to define their clinical utilization. We anticipate that elu-cidating their roles and pathways in mammary stem cells and in breast cancer will help drive the development of new biomarkers and new therapeutic miRNA approaches for breast cancer treatment.
Acknowledgements
E.A. carried out the miRNA expression studies, qPCR analysis, bioinformatic analysis, cultured cells, performed miRNA mimic experiments and drafted the manuscript. A.K. participated in miRNA analysis, bioinformatics and revised the manuscript. E.T. participated in qPCR analysis and revised the manuscript. C.-Y.L. performed clus-tering and survival analysis of clinical data and critically revised the manuscript. L.-A.H. participated in the design of the study and criti-cally revised the manuscript. L.H. cultured cells, performed the immu-noflourescent staining experiments and participated in the design of the study and critically revised the manuscript. C.W. conceived and coordinated the study, participated in its design and helped to draft the manuscript. All authors read and approved the final manuscript. We thank Preethi Gunaratne and Xiaolian Gao, University of Houston, and Melissa Landis, The Methodist Research Hospital Institute, for helpful advice and discussions.
Supplementary material
Supplementary Tables 1–5, Figures 1–4, and Data 1–4 can be found at http://carcin.oxfordjournals.org/
Funding
Texas Emerging Technology Fund (Agreement no. 300-9-1958).
Conflict of Interest Statement: The authors declare no conflict of interests.
References
1. Sorlie,T. et al. (2001) Gene expression patterns of breast carcinomas dis-tinguish tumor subclasses with clinical implications. Proc. Natl. Acad. Sci. USA, 98, 10869–10874.
2. Johnson,S.M. et al. (2005) RAS is regulated by the let-7 microRNA family. Cell, 120, 635–647.
3. Esquela-Kerscher, A. et al. (2006) Oncomirs—microRNAs with a role in cancer. Nat. Rev. Cancer, 6, 259–269.
4. Calin,G.A. et al. (2006) MicroRNA signatures in human cancers. Nat. Rev. Cancer, 6, 857–866.
5. Yu,F. et al. (2007) let-7 regulates self renewal and tumorigenicity of breast cancer cells. Cell, 131, 1109–1123.
6. Kota,J. et al. (2009) Therapeutic microRNA delivery suppresses tumori-genesis in a murine liver cancer model. Cell, 137, 1005–1017.
7. Wiggins,J.F. et al. (2010) Development of a lung cancer therapeutic based on the tumor suppressor microRNA-34. Cancer Res., 70, 5923–5930.
8. Shimono,Y. et al. (2009) Downregulation of miRNA-200c links breast can-cer stem cells with normal stem cells. Cell, 138, 592–603.
9. Sun,J.G. et al. (2010) Microarray-based Analysis of MicroRNA Expres-sion in Breast Cancer Stem Cells. J. Exp. Clin. Cancer Res., 29, 174.
10. Greene,S.B. et al. (2010) A putative role for microRNA-205 in mammary epithelial cell progenitors. J. Cell Sci., 123, 606–618.
11. Ibarra,I. et al. (2007) A role for microRNAs in maintenance of mouse mammary epithelial progenitor cells. Genes Dev., 21, 3238–3243.
12. Ball,R.K. et al. (1988) Prolactin regulation of beta-casein gene expression and of a cytosolic 120-kd protein in a cloned mouse mammary epithelial cell line. EMBO J, 7, 2089–2095.
13. Humphreys,R.C. et al. (1997) Stably transfected HC11 cells provide an in vitro and in vivo model system for studying Wnt gene function. Cell Growth Differ., 8, 839–849.
14. Williams,C. et al. (2009) Gene expression in murine mammary epithelial stem cell-like cells shows similarities to human breast cancer gene expres-sion. Breast Cancer Res., 11, R26.
15. Miller,L.D. et al. (2005) An expression signature for p53 status in human breast cancer predicts mutation status, transcriptional effects, and patient survival. Proc. Natl. Acad. Sci. USA, 102, 13550–13555.
16. Tiede,B.J. et al. (2009) A novel mouse model for non-invasive single marker tracking of mammary stem cells in vivo reveals stem cell dynamics throughout pregnancy. PLoS One, 4, e8035.
17. Cicalese,A. et al. (2009) The tumor suppressor p53 regulates polarity of self-renewing divisions in mammary stem cells. Cell, 138, 1083–1095.
18. Hebbard,L.W. et al. (2010) Maternal embryonic leucine zipper kinase is upregulated and required in mammary tumor-initiating cells in vivo. Can-cer Res., 70, 8863–8873.
19. Gregory,P.A. et al. (2008) The miR-200 family and miR-205 regulate epi-thelial to mesenchymal transition by targeting ZEB1 and SIP1. Nat. Cell Biol., 10, 593–601.
20. Iliopoulos,D. et al. (2010) Loss of miR-200 inhibition of Suz12 leads to polycomb-mediated repression required for the formation and maintenance of cancer stem cells. Mol. cell, 39, 761–772.
21. Wyatt,L. et al. (2007) The protein tyrosine phosphatase Pez regulates TGF-beta, epithelial-mesenchymal transition, and organ development. J. Cell Biol., 178, 1223–1235.
22. Iliopoulos,D. et al. (2009) MicroRNAs differentially regulated by Akt iso-forms control EMT and stem cell renewal in cancer cells. Sci. Signal, 2, ra62.
23. Baek,D. et al. (2008) The impact of microRNAs on protein output. Nature, 455, 64–71.
24. Vaught,D. et al. (2009) Regulation of mammary gland branching morpho-genesis by EphA2 receptor tyrosine kinase. Mol. Biol. Cell, 20, 2572–2581.
25. Brantley-Sieders,D.M. et al. (2008) The receptor tyrosine kinase EphA2 promotes mammary adenocarcinoma tumorigenesis and metastatic pro-gression in mice by amplifying ErbB2 signaling. J. Clin. Invest., 118, 64–78.
26. Pedersen,K. et al. (2009) A naturally occurring HER2 carboxy-terminal fragment promotes mammary tumor growth and metastasis. Mol. Cell Biol., 29, 3319–3331.
28. Esquela-Kerscher,A. et al. (2006) Oncomirs—microRNAs with a role in cancer. Nat. Rev. Cancer, 6, 259–269.
29. Bhaumik,D. et al. (2008)Expression of microRNA-146 suppresses NF-kappaB activity with reduction of metastatic potential in breast cancer cells. Oncogene, 27, 5643–5647.
30. Krzyzanowski,P.M. et al. (2007) Identification of novel stem cell markers using gap analysis of gene expression data. Genome Biol., 8, R193.
31. He,L. et al. (2005) A microRNA polycistron as a potential human onco-gene. Nature, 435, 828–833.
32. Adams,B.D. et al. (2009) The role of miR-206 in the epidermal growth fac-tor (EGF) induced repression of estrogen receptor-alpha (ERalpha) signal-ing and a luminal phenotype in MCF-7 breast cancer cells. Mol. Endocrin., 23, 1215–1230.
33. Fazili,Z. et al. (1999) Disabled-2 inactivation is an early step in ovarian tumorigenicity. Oncogene, 18, 3104–3113.
34. Jiang,Y. et al. (2007) The inhibitory effects of Disabled-2 (Dab2) on Wnt signaling are mediated through Axin. Oncogene, 27, 1865–1875.
35. Huang,C.L. et al. (2010) Disabled-2 is required for mesoderm dif-ferentiation of murine embryonic stem cells. J. Cell. Physiol., 225, 92–105.
36. Alles,M.C. et al. (2009) Meta-analysis and gene set enrichment relative to ER status reveal elevated activity of MYC and E2F in the “basal” breast cancer subgroup. PLoS One, 4, e4710.
37. Rogoff,H.A. et al. (2004) Life, death and E2F: linking proliferation control and DNA damage signaling via E2F1. Cell Cycle, 3, 845–846.
38. Iliopoulos,D. et al. (2009) An epigenetic switch involving NF-kappaB, Lin28, Let-7 MicroRNA, and IL6 links inflammation to cell transforma-tion. Cell, 139, 693–706.
39. Johnson,C.D. et al. (2007) The let-7 microRNA represses cell proliferation pathways in human cells. Cancer Res., 67, 7713–7722.
40. Visvader,J.E. (2009) Keeping abreast of the mammary epithelial hierarchy and breast tumorigenesis. Genes Dev., 23, 2563–2577.
41. Dimri,G. et al. (2005) Mammary epithelial cell transformation: insights from cell culture and mouse models. Breast Cancer Res., 7, 171–179.
42. Molyneux,G. et al. (2010) BRCA1 basal-like breast cancers originate from luminal epithelial progenitors and not from basal stem cells. Cell Stem Cell, 7, 403–417.
43. Pece,S. et al. (2010) Biological and molecular heterogeneity of breast cancers correlates with their cancer stem cell content. Cell, 140, 62–73.
44. Doetsch,F. et al. (2002) EGF converts transit-amplifying neurogenic precursors in the adult brain into multipotent stem cells. Neuron, 36, 1021–1034.
45. Milanezi,F. et al. (2008) EGFR/HER2 in breast cancer: a biological approach for molecular diagnosis and therapy. Exp. Rev. Mol. Diag., 8, 417–434.
Received November 14, 2012; revised April 14, 2012; accepted April 30, 2012
