Research ArticleCancer Biology

TRPS1 Targeting by miR-221/222 Promotes the Epithelial-to-Mesenchymal Transition in Breast Cancer

See allHide authors and affiliations

Science Signaling  14 Jun 2011:
Vol. 4, Issue 177, pp. ra41
DOI: 10.1126/scisignal.2001538

Abstract

The basal-like subtype of breast cancer has an aggressive clinical behavior compared to that of the luminal subtype. We identified the microRNAs (miRNAs) miR-221 and miR-222 (miR-221/222) as basal-like subtype–specific miRNAs and showed that expression of miR-221/222 decreased expression of epithelial-specific genes and increased expression of mesenchymal-specific genes, and increased cell migration and invasion in a manner characteristic of the epithelial-to-mesenchymal transition (EMT). The transcription factor FOSL1 (also known as Fra-1), which is found in basal-like breast cancers but not in the luminal subtype, stimulated the transcription of miR-221/222, and the abundance of these miRNAs decreased with inhibition of the epidermal growth factor receptor (EGFR) or MEK (mitogen-activated or extracellular signal–regulated protein kinase kinase), placing miR-221/222 downstream of the RAS pathway. Furthermore, miR-221/222–mediated reduction in E-cadherin abundance depended on their targeting the 3′ untranslated region of the GATA family transcriptional repressor TRPS1 (tricho-rhino-phalangeal syndrome type 1), which inhibited EMT by decreasing ZEB2 (zinc finger E-box–binding homeobox2) expression. We conclude that by promoting EMT, miR-221/222 may contribute to the more aggressive clinical behavior of basal-like breast cancers.

Introduction

Breast cancer is a heterogeneous disease that includes distinct morphological and molecular subtypes (1). Gene expression studies have identified four major breast cancer subtypes by mRNA profiling: luminal A, luminal B, HER2-positive, and basal-like (2, 3); the basal-like subtype, which overlaps substantially with the triple-negative subtype, is associated with aggressive clinical behavior and poor prognosis despite treatment with chemotherapy. The basal-like subtype, relative to the luminal subtype, expresses higher levels of molecular markers that are consistent with cells that have undergone an epithelial-to-mesenchymal transition (EMT) and morphologically resemble mesenchymal cells. These molecular markers include the epithelial marker E-cadherin and mesenchymal marker vimentin, which are important in permitting cells to becoming less adherent and gaining the ability to migrate and invade through the basement membrane as part of the initial steps that promote metastasis (4). Intriguingly, a thorough analysis of the basal-like subtype resulted in the delineation of the claudin-low subtype, in which cells typically express the breast stem cell markers CD44Hi/CD24Lo and have an increased abundance of mesenchymal markers and a poor clinical outcome (57). These cells with stem cell markers are typically refractory to chemotherapy and have an enhanced capability to initiate tumors in mice; therefore, these data present a compelling case for a relationship between the molecular subtype and the aggressive clinical course.

MicroRNAs (miRNAs) are small noncoding RNAs that negatively regulate mRNA stability (8) by binding to the 3′ untranslated region (3′UTR) of a protein-coding gene harboring a “seed” sequence, which is typically complementary to two to eight nucleotides of the miRNA, and recruiting the RNA-induced silencing complex (RISC) to facilitate mRNA degradation. miRNAs enable the identification of the tissue of origin of various cancers (9), can have oncogenic or tumor suppressor properties (10, 11), and contribute to various developmental programs (12). miRNAs have also been identified that enhance several aspects of breast cancer pathogenesis including metastasis (13), invasion (14), and self-renewal (15). The luminal and basal-like subtypes of breast cancer show distinct miRNA expression profiles (16); it has been unclear, however, whether miRNAs specifically expressed in basal-like breast cancers contribute to their enhanced ability for migration, invasion, and molecular features associated with the EMT.

Results

miR-221/222 is abundant in basal-like breast cancer and promotes the EMT

We applied differential expression analysis with moderated t statistics (17) to a panel of breast cancer cell lines profiled on a miRNA microarray and identified miR-221 and miR-222 (miR-221/222) as the two miRNAs with expression patterns that most clearly discriminated between the basal-like and the luminal subtypes of breast cancer (Fig. 1A and fig. S1A). Both of these miRNAs tended to be abundant in the basal-like subtype of breast cancer but not in the luminal subtype. Analyzing the genomic location of miR-221/222 revealed that they are located within 1 kb of each other in the same orientation on chromosome X and are thus likely co-regulated (18). Consistent with previous observations (19), we found that the abundance of miR-200c was lower in basal-like breast cancer cell lines than in luminal breast cancer lines (fig. S1). We confirmed the microarray observations for miR-221/222 with quantitative reverse transcription–polymerase chain reaction (qRT-PCR) (Fig. 1B). Classification of the basal-like breast cancer cell lines into basal A or basal B subtypes, where the basal B subtype expresses higher levels of mesenchymal markers and has a greater invasion capability (20, 21), suggested that miR-221/222 was not specific to either subtype, although a trend for increased expression of miR-221/222 was observed in the basal B subtype compared to basal A. We also found that miR-221/222 was more abundant in triple-negative tumors than in estrogen receptor/progesterone receptor (ER/PR)–positive tumors (fig. S2).

Fig. 1

miR-221 and miR-222 are present in basal-like breast cancer cell lines and induce EMT. (A) Top 20 miRNAs showing significant differential expression between luminal and basal-like breast cancer cell lines (table S1). (B) miR-221/222 abundance determined by qRT-PCR across the breast cancer cell lines. Data are presented as means ± SD. (C) Selected gene sets altered by treatment with miR-221/222. Direction indicates change elicited by miR-221/222 treatment. (D) qRT-PCR on MCF10As 3 days after transfection with either scramble control mimic (miR-Control) or miR-221/222 mimic. Data represent means of triplicates ± SD. (E) Images of MCF10A cells transfected with scramble control mimic or miR-221/222 mimic. A representative image is shown from three independent experiments for each panel. (F) Inhibition of miR-221/222 in basal-like cell lines. Cells were transfected and assayed for E-cadherin and vimentin abundance after 48 hours by qRT-PCR. (G) Migration of MCF10A and MDA-MB-231 cells. Migration was assessed with a Boyden chamber assay 5 days after transfection. Data represent means of triplicates ± SD and are displayed as relative fluorescent units (RFU). (H) Invasion by MCF10A and MDA-MB-231 cells. Invasion was assessed by a Boyden chamber invasion assay 3 days after transfection. Data represent means of triplicates ± SD.

miRNAs have the potential to regulate multiple targets; given their differential abundance in the basal-like and luminal subtypes, we wished to determine in an unbiased fashion the possible roles that miR-221/222 might play in breast cancer. We transfected an immortalized, nontransformed human mammary epithelial cell line, MCF10A, with synthetic oligo mimics for miR-221/222 or with a scrambled miR mimic to function as a negative control (miR-Control) and then analyzed gene expression with microarray and Gene Set Enrichment Analysis (GSEA) (Fig. 1C). Intriguingly, we found that gene sets involved in the EMT, transforming growth factor–β (TGFβ) pathway, and RAS pathway were modulated in cells transfected with the miR-221/222 mimics. Because EMT contributes to the pathogenesis of breast cancer, and basal-like breast cancers show a more aggressive clinical behavior than luminal subtypes do, we investigated whether the miR-221/222 mimic affected the expression of the epithelial marker E-cadherin or the mesenchymal marker vimentin. The abundance of the mRNA encoding vimentin was increased by the miR-221/222 mimics but not the scrambled control, whereas that of the mRNA encoding E-cadherin was decreased (Fig. 1D). Furthermore, we were able to detect vimentin by immunofluorescence in cells treated with the miR-221/222 mimic, as well as a decrease in the abundance of E-cadherin, relative to that in cells treated with the scrambled control mimic (Fig. 1E and fig. S3). Moreover, MCF10A cells transfected with the miR-221/222 mimics adopted a more elongated cellular morphology consistent with a mesenchymal phenotype compared to cells treated with the scrambled miR (Fig. 1E and fig. S3). To determine whether inhibition of miR-221/222 would promote the opposite phenotype [mesenchyme-to-epithelial transition (MET)], we treated the basal-like breast cancer cell line MDA-MB-231 with an inhibitor of miR-221/222. This led to an increase in the abundance of E-cadherin and a decrease in that of vimentin (Fig. 1F, left panel). Similar results were obtained with another basal-like cell line, CAL85-1 (Fig. 1F, right panel), suggesting that inhibition of miR-221/222 promotes the MET.

We also determined whether cell migration and invasion were enhanced by miR-221/222 in MCF10A cells and, conversely, whether synthetic oligo inhibitors to miR-221/222 could attenuate MDA-MB-231 cell migration and invasion through a basement membrane matrix. We observed an increase in the migration toward medium containing 10% serum of MCF10A cells transfected with the miR-221/222 mimic compared to the scrambled miR mimic. Furthermore, there was a decrease in cell migration of MDA-MB-231 cells transfected with the miR-221/222 inhibitors relative to those transfected with a scrambled miR inhibitor (Fig. 1G). The capacity of MCF10A cells to invade through the basement membrane matrix was enhanced by the mimics of miR-221/222, whereas the number of invasive MDA-MB-231 cells was attenuated by inhibition of miR-221/222 (Fig. 1H). These data further suggest that miR-221/222 promotes the EMT in breast cancer cells.

The ability of miR-221/222 to promote the EMT phenotype in breast cancer cell lines led us to investigate whether miR-221/222 abundance correlated with the EMT phenotype in primary breast cancers. Because a high abundance of vimentin correlates with breast cancer metastasis and disease recurrence (22) and is associated with the poor-prognosis claudin-low subtype (7), we determined whether miR-221/222 abundance correlated with that of vimentin mRNA in primary breast cancer samples (fig. S4) and found a positive correlation (P = 0.02). In addition, there was a negative correlation between the abundance of miR-221/222 and that of the mRNA encoding E-cadherin (fig. S5) (P = 0.05). Together, these data are consistent with a role for miR-221/222 in promoting the EMT and indicate that miR-221/222 abundance correlates with EMT markers and phenotypes.

FOSL1 drives the transcription of miR-221/222

miR-221 and miR-222 do not reside within an intron of any known gene, suggesting that they require their own transcriptional regulation. To identify potential transcriptional regulators, we searched upstream of the miR-221/222 promoter for conserved binding sites for transcription factors whose mRNA abundance correlated with that of miR-221/222 in all 134 cell lines for which we had matching miRNA and mRNA data (Fig. 2A, PCC = 0.795, P < 0.0001). This led to the identification of a sole candidate transcription factor, FOSL1 (also known as Fra-1), which is a member of the Fos family of transcription factors. Fos-family transcription factors heterodimerize with members of the Jun family to form the AP-1 complex (23); moreover, FOSL1 is thought to promote breast cancer invasiveness and morphological transformation (24, 25). We identified an AP-1 binding site 12 kb upstream of miR-221/222 (Fig. 2B) and subcloned a genomic fragment from this region into a minimal promoter-containing luciferase construct to create a reporter gene (FOSL1-WT-BS) that could be used to determine whether FOSL1 acts through this element to stimulate or repress transcription. As a negative control, we created a reporter construct in which key base pairs required for AP-1 binding were mutated in the putative AP-1 binding site (FOSL1-Mut-BS), and for a positive control, we used a commercially available AP-1 reporter. Transfection of the plasmid encoding FOSL1 into MCF7 breast cancer cells (a luminal subtype) increased the luciferase activity of the AP-1–Luc reporter about fourfold relative to that in cells lacking FOSL1 and also enhanced luciferase activity from the FOSL1-WT-BS construct about sixfold, but had no notable effect on the luciferase activity of the FOSL1-Mut-BS construct (stimulating it ~1.5-fold) (Fig. 2C). We then used chromatin immunoprecipitation (ChIP) analysis to determine whether endogenous FOSL1 binds to the putative FOSL1 binding site on the miR-221/222 promoter. Antibodies to FOSL1 and an isotype control were used for the ChIP assay, and the resultant bound DNA was determined by qRT-PCR using specific oligos for the predicted FOSL1 binding site on the miR-221/222 promoter and a region 5 kb downstream of the promoter to serve as a negative control. The ChIP assay revealed substantial association of FOSL1 with the predicted miR-221/222 FOSL1 binding site region in MDA-MB-231 cells, in which FOSL1 and miR-221/222 are abundant, but not in MCF7 cells, which have little FOSL1 or miR-221/222. In contrast, no enrichment was apparent for the negative control promoter region in either cell line (Fig. 2D). Transcription factors can reside on promoters without mediating transactivation and can repress nearby transcriptional modules; we therefore wished to determine whether the region on the miR-221/222 promoter to which FOSL1 was bound was transcriptionally active. ChIP analysis using antibodies directed against acetylated histone 3 (anti-AcH3) to immunoprecipitate transcriptionally active promoter regions (26) uncovered a substantial enrichment for the predicted FOSL1 binding region on the miR-221/222 promoter in MDA-MB-231 cells but not in MCF7 cells (Fig. 2E). Moreover, small interfering RNA (siRNA) directed against FOSL1, which decreased the abundance of the mRNA encoding FOSL1 (Fig. 2F) and of FOSL1 protein (Fig. 2G), also decreased the abundance of miR-221/222 but not that of the negative controls miR-155 or miR-638. Together, these data are consistent with the hypothesis that FOSL1 drives miR-221/222 transcription.

Fig. 2

FOSL1 positively regulates transcription of the miR-221/222 gene cluster. (A) FOSL1 is positively correlated with miR-221/222 abundance. miR-221/222 and FOSL1 mRNA expression was determined in 134 cell lines by microarrays. Red/green dots represent luminal/basal-like subtype. (B) Cartoon depicting the human genomic architecture and promoter region of miR-221/222 and the highly conserved AP-1 binding site. (C) FOSL1 stimulates transcription from a heterologous promoter containing the predicted binding site from the miR-221/222 promoter. MCF7 cells were transfected as indicated and luciferase activity was determined 18 hours after transfection. Data represent means of triplicates ± SD. (D) FOSL1 binds to the miR-221/222 promoter region. qPCR of the predicted FOSL1 binding site on the miR-221/222 promoter or negative control after ChIP as indicated. Data represent means of triplicates ± SD. (E) FOSL1 binds to a transcriptionally active promoter region of miR-221/222. qPCR of the miR-221/222 FOSL1 promoter binding site or negative control after ChIP as indicated. Data represent means of triplicates ± SD. (F) FOSL1 depletion by siRNA reduces miR-221/222 abundance. MDA-MB-231 cells were transfected as indicated. qRT-PCR carried out 3 days after transfection. Data represent means of triplicates ± SD. (G) Western blotting (WB) of FOSL1 and actin following FOSL1 siRNA.

miR-221/222 is downstream of RAS

Previous studies have suggested an exquisite addiction of basal-like breast cancer cell lines to the RAS to RAF to MEK (mitogen-activated or extracellular signal–regulated protein kinase kinase) (RAS-RAF-MEK) pathway (27, 28). Furthermore, oncogenic RAS promotes EMT in certain cellular contexts and enhances occupancy of AP-1 sites by Fra-1 and JunD (29). Having identified genes downstream of the RAS pathway being modulated in the miR-221/222 mimics experiment (Fig. 1C), we tested the hypothesis that the RAS-RAF-MEK pathway regulates the expression of miR-221/222 in basal-like breast cancer cell lines. We treated the basal-like breast cancer cell lines MDA-MB-231 and CAL85-1 with the MEK inhibitor PD0325901 (herein referred to as MEKi) and determined the abundance of miR-221/222 by qRT-PCR (Fig. 3A). In both cell lines, MEKi treatment led to a decrease in phosphorylation of the MEK substrates ERK1/2 (extracellular signal–regulated kinase 1 and 2). The abundance of miR-221/222 decreased in both cell lines upon MEKi treatment, indicating that the RAS-RAF-MEK pathway contributes to miR-221/222 production. Consistent with this hypothesis, gene expression profiling of MCF10A cells 24 hours after infection with an adenovirus containing an active form of MEK or of the RAS isoform HRAS revealed that 47% (75 of 158) of genes whose transcripts decreased in expression in response to MEK signaling and 46% (118 of 258) of those that decreased with RAS signaling also decreased with miR-221/222 (Fig. 3B).

Fig. 3

The EGFR-RAS-RAF-MEK pathway increases miR-221/222 abundance. (A) MEKi decreases miR-221/222 abundance. qRT-PCR was performed after treatment with MEKi for 3 days. Data represent means of triplicates ± SD with Western blots performed as indicated. (B) Venn diagram representing overlapping miR-221/222 targets decreased in expression by active forms of HRAS or MEK in MCF10A cells. Gene expression was assessed by microarray. (C) Erlotinib decreases miR-221/222 abundance. qRT-PCR after treatment with erlotinib or dimethyl sulfoxide (DMSO) for 3 days as indicated. Data represent means of triplicates ± SD. Western blots were performed as indicated. (D) EGFR siRNA decreases miR-221/222 abundance. qRT-PCR was performed 3 days after transfection. Data represent means of duplicates ± SD. (E and F) MEKi treatment increases E-cadherin mRNA and decreases vimentin mRNA. qRT-PCR was performed after treatment with MEKi for 3 days. Data represent means of triplicates ± SD. (G) miR-221/222 attenuates MEKi-mediated up-regulation of E-cadherin. MDA-MB-231 cells were transfected with miR-221/222 mimic or scrambled control 24 hours before MEKi treatment. E-cadherin mRNA abundance was assessed 24 hours after MEKi treatment by qRT-PCR. Data represent means of triplicates ± SD. (H) The RAS-RAF-MEK pathway increases miR-221/222 abundance via FOSL1. qRT-PCR was performed as indicated. Data represent means of triplicates ± SD.

Although MDA-MB-231 cells contain a constitutively active mutant form of RAS, CAL85-1 cells contain wild-type RAS, leading us to explore the identity of receptor tyrosine kinases that could stimulate RAS in these cells and thereby downstream expression of miR-221/222. Basal-like breast cancers have an epidermal growth factor receptor (EGFR)–associated gene signature (30), and EGFR is present in up to 66% of these cancers (1); therefore, we explored the hypothesis that EGFR might drive the abundance of miR-221/222 through the RAS-RAF-MEK-FOSL1 cascade. We treated CAL85-1 cells, which have abundant EGFR, with the EGFR inhibitor erlotinib and assessed ERK1/2 phosphorylation and miR-221/222 abundance (Fig. 3C). We found that erlotinib reduced ERK1/2 phosphorylation and miR-221/222 abundance to a degree similar to that seen with MEKi (Fig. 3A). Consistent with the erlotinib and MEKi data, transfection of CAL85-1 cells with a siRNA targeting EGFR, which decreased the abundance of EGFR mRNA ~80%, also decreased the abundance of miR-221/222 (Fig. 3D). To investigate the possibility that EGFR might signal to miR-221/222 independently of RAS, we assessed the effects of EGFR siRNA on miR-221/222 abundance in MDA-MB-231 cells, which carry a constitutively active RAS mutant (Fig. 3D). We observed no significant decrease (P = 0.47) in miR-221/222 abundance in MDA-MB-231 cells treated with EGFR siRNA. These data indicate that EGFR signaling increases miR-221/222 abundance more prominently in the genetic context of RAS wild-type status and suggest that RAS activation can increase miR-221/222 abundance independent of upstream signaling through EGFR.

Treatment of MDA-MB-231 or CAL85-1 cells with MEKi increased the abundance of the epithelial marker E-cadherin and decreased that of vimentin, consistent with their becoming more epithelial and less mesenchymal (Fig. 3, E and F). To determine the extent to which the MET mediated by MEKi depended on miR-221/222, we added a mimic of miR-221/222 or scrambled mimic to MDA-MB-231 cells, treated them with different concentrations of MEKi, and then measured E-cadherin mRNA abundance (Fig. 3G). The miR-221/222 mimic attenuated the increase in E-cadherin abundance at every concentration of the MEKi tested, suggesting that the MET phenotype induced by MEKi partially depends on decreased miR-221/222 abundance.

Because FOSL1 stimulates the transcription of the miR-221/222 gene cluster (Fig. 2) and cells treated with MEK or EGFR inhibitors had decreased abundance of miR-221/222, we hypothesized that inhibition of RAS-RAF-MEK might decrease FOSL1 abundance. Indeed, we found that MEKi treatment decreased FOSL1 mRNA abundance in both MDA-MB-231 and CAL85-1 cells (Fig. 3H). Together, these data suggest that miR-221/222 is positively regulated by the EGFR-RAS axis.

The EMT inhibitor TRPS1 is targeted by miR-221/222

We implemented two strategies to identify critical targets responsible for the phenotypic changes induced by miR-221/222. Prediction algorithms identify hundreds of potential miR-221/222 targets; therefore, our first strategy to filter and identify candidate genes was based on their meeting two key criteria: (i) Their mRNA was more abundant in luminal than in basal-like breast cancer cell lines, and (ii) their mRNA abundance was decreased by miR-221/222 mimics in MCF10A cells (Fig. 4A). Using this strategy, we were able to eliminate all but 21 potential targets (table S2). The second strategy involved the implementation of a scoring mechanism we called the cumulative regulatory score (CRS) to rank the potential cumulative regulatory effect multiple miRNAs may have on any one target gene using both mRNA and miRNA expression data. The CRS is essentially a sum of the product of the miRNA and mRNA t statistics (see Supplementary Materials). Because miR-221/222 has the strongest basal-specific abundance, we calculated the CRS for predicted miR-221/222 targets and used precision-recall analysis, which is a statistical method to assess the positive predictive value and true positive rate, of these predicted targets to identify high-ranking targets for miR-221/222 (table S3). We used a precision of ~70% at ~13.5% recall when sorting the scores by absolute values. Two genes were identified by both our initial strategy and CRS: tricho-rhino-phalangeal syndrome type 1 (TRPS1) and iroquois homeobox 5 (IRX5). Whereas knockdown of either TRPS1 or IRX5 decreased the abundance of the mRNA encoding E-cadherin, only TRPS1 knockdown increased the abundance of the mRNA encoding vimentin (Fig. 4B). Consequently, we focused on TRPS1 in our subsequent experiments.

Fig. 4

miR-221/222 promotes the EMT by targeting TRPS1-mediated inhibition of ZEB2. (A) Identification of candidate miR-221/222 targets. Plot represents change in gene expression of MCF10A cells treated with miR-221/222 mimic relative to scrambled miR control and change in baseline gene expression between basal and luminal cell lines. (B) Knockdown of candidate miR-221/222 targets by siRNA. Changes in mRNA abundance were determined by qRT-PCR. (C) TRPS1 knockdown increases vimentin abundance. Immunofluorescence staining of MCF10A cells as indicated. A representative image is shown from three independent experiments for each panel. (D) miR-221/222 reduces the abundance of TRPS1. Western blotting was carried out as indicated. (E) TRPS1 3′UTR predicted site for miR-221/222. (F) TRPS1 3′UTR is a target for miR-221/222. MCF10A cells were transfected as indicated. (G) miR-221/222 increases ZEB2 mRNA abundance. MCF10A cells were transfected as indicated, and ZEB2 mRNA abundance was assessed by qRT-PCR. (H) ZEB2 transcript levels are decreased by TRPS1. MCF10A cells were treated as indicated and ZEB2 mRNA abundance was determined. (I) Overexpression of TRPS1 blunts miR-221/222–mediated up-regulation of ZEB2. MCF10A cells were transfected as indicated and abundance of ZEB2 mRNA was assessed by qRT-PCR 48 hours after transfection. (J) Endogenous TRPS1 binds to the ZEB2 promoter. ChIP assays were performed with antibodies against TRPS1 or isotype control. (K) ZEB2 is required for miR-221/222 negative regulation of E-cadherin. MCF10A cells were treated as indicated and abundance of E-cadherin was assessed by qRT-PCR. All treatment experiments in this figure were carried out in triplicate and are displayed as means ± SD where charts are displayed.

TRPS1 is a member of the GATA family transcriptional factors, which bind specifically to “GATA” DNA sequences. Mutation or deletion of TRPS1 causes tricho-rhino-phalangeal syndrome type 1, a condition characterized by unique facial features and skeletal abnormalities such as a bulbous nose, sparse hair, and mild growth retardation (31). TRPS1 knockout mice develop similar defects in bone and hair development (32), and TRPS1 has also been implicated in kidney development through its promotion of MET (33). Moreover, decreased abundance of TRPS1 is a marker of poor clinical outcome in breast cancers (34). Consistent with its effects on mRNA abundance, TRPS1 knockdown increased vimentin immunofluorescence and decreased E-cadherin immunofluorescence (Fig. 4C). In addition, TRPS1 abundance was decreased by miR-221/222 and TRPS1 siRNA, consistent with their reduction of its mRNA abundance (Fig. 4D).

To confirm that TRPS1 is a direct target of miR-221/222, we cloned the 3′UTR of TRPS1 into a luciferase reporter construct and also created a luciferase reporter containing a mutant version generated by mutating nucleotides 2 and 3 of the predicted seed sequence (Fig. 4E). Relative to the scrambled miR mimic, miR-221/222 decreased luciferase expression of the wild-type construct but not that of the mutant 3′UTR construct (Fig. 4F). Consistent with these data, TRPS1 mRNA abundance in primary breast cancers was negatively correlated with that of miR-221/222 (fig. S6, P = 0.01). Collectively, these data indicate that TRPS1 is a target of miR-221/222 and that its down-regulation by miR-221/222 promotes the EMT phenotype.

TRPS1 repression of ZEB2 is necessary for EMT

Because TRPS1 can repress transcription from GATA-containing binding sites, we searched for conserved GATA sites upstream of genes that encode proteins implicated in the EMT. Zinc finger E-box–binding homeobox2 (ZEB2) emerged as the sole candidate for regulation by TRPS1, on the basis of the presence of GATA sites nearby its promoter and its increased mRNA abundance in response to miR-221/222 (Fig. 4G). ZEB2 is positively associated with EMT in development. ZEB2 is a transcriptional repressor that decreases E-cadherin expression and that of other components of intercellular junction complexes (35). Consistent with TRPS1 being a negative regulator of ZEB2, we observed an increase in the abundance of the mRNA encoding ZEB2 after TRPS1 knockdown (Fig. 4H). Furthermore, addition of an expression vector containing the coding sequence for human TRPS1 without its native 3′UTR decreased ZEB2 mRNA abundance and attenuated the increase in ZEB2 mRNA mediated by miR-221/222 (Fig. 4I). Because transcription factors can influence gene expression directly or indirectly, we wished to determine whether endogenous TRPS1 was present on the ZEB2 promoter. We postulated that TRPS1 was likely to bind to the same sites as those identified for the TRPS1 family member, GATA2 (36). ChIP of the ZEB2 promoter from MCF10A cell lysates using an antibody to TRPS1 revealed enrichment of the predicted binding sites, but not to a region of the ZEB2 promoter lacking a GATA family consensus site (Fig. 4J).

ZEB2 knockdown with siRNA attenuated the decrease in E-cadherin mRNA produced by miR-221/222 mimic or knockdown of TRPS1 (Fig. 4K, top and bottom panels, respectively). Together, these data suggest that negative regulation of E-cadherin mRNA by miR-221/222 is a result of miR-221/222’s targeting TRPS1 and thereby relieving its inhibition of ZEB2.

Discussion

The EMT is thought to occur through multiple signaling pathways that are influenced by the tumor microenvironment (37). More specifically, receptor tyrosine kinase signaling and the RAS pathway play key roles in mediating this process (38). ERK2 acts as a critical positive regulator of EMT downstream of RAS (39), and our results identify miR-221/222 as a downstream effector of this axis that promotes the EMT by targeting TRPS1 and relieving its inhibition of ZEB2 (Fig. 5). TRPS1 is required for the normal formation of nephrons through its promotion of the MET and ureteric bud branching during the early stages of murine renal development (33). Furthermore, TRPS1 haploinsufficiency in the mouse promotes renal fibrosis and TGFβ-induced EMT (40). These findings support our conclusion that TRPS1 acts to inhibit the EMT. Consistent with this, lower abundance of TRPS1 is a poor prognostic marker in breast cancer (34). The defects associated with TRPS1 mutation (31) are indicative of a crucial role in modulating facial and hair morphogenesis and skeletal development and suggest that mutations in TRPS1 may affect the timing of the EMT during neural crest development.

Fig. 5

Proposed model of miR-221/222 promoting EMT. miR-221/222 abundance is positively regulated by the EGFR-RAS-RAF-MEK-ERK2-FOSL1 axis and promotes EMT by targeting TRPS1, which directly represses the transcription of ZEB2. As a consequence of increased miR-221/222 abundance, ZEB2 abundance increases, permitting the repression of E-cadherin and up-regulation of vimentin to promote EMT. Red or blue color indicates increased abundance in basal-like or luminal subtypes of breast cancer, respectively.

We identified ZEB2 as one of the downstream genes responsible for TRPS1-mediated repression of EMT. However, it is conceivable that TRPS1 may negatively regulate other positive regulators of EMT that we have not identified. ZEB2 directly represses E-cadherin transcription (35) and has been suggested to directly increase vimentin transcription (41), although the mechanism of the latter remains unclear. Up-regulation of ZEB2 expression also occurs by TGFβ signaling or by a decrease in the abundance of miR-200 and miR-205 that directly target the 3′UTR of ZEB2 (19). Furthermore, ZEB2 has been previously implicated downstream of RAS and ERK2 (39). Therefore, it is possible that during EMT in breast cancer, miR-221/222 targets TRPS1 to relieve its repression of ZEB2 above a certain threshold that would then permit activation of a feedback loop that leads to transcriptional repression of the miR-200 family (19) to engage and commit cells to an EMT program.

miR-221/222 targets the 3′UTR of various genes, including those encoding estrogen receptor (ESR1) (42), p27 (43), and Kit (44). Of these, ESR1 has been reported to promote metastasis-associated family member 3 (MTA3)–dependent repression of snail homolog 1 (SNAI1), which is a zinc finger transcriptional repressor that directly represses transcription of the gene encoding E-cadherin (45). This provides a potential mechanistic link to the EMT and could contribute to our observations of EMT induced by miR-221/222. However, we did not detect any change in SNAIL abundance in response to miR-221/222.

It is believed that tumor cells undergo EMT to escape the primary site and then undergo MET to colonize distant sites, which ultimately leads to a poor clinical outcome (4). Tumor cells surviving after conventional chemotherapy regimens in breast cancer patients harbor molecular properties consistent with a mesenchymal and breast stem cell phenotype (5), and exposure of luminal-type breast cancer cells to chemotherapy can result in traits that are similar to those characteristic of basal-like cells (46). Furthermore, ER-positive tumors that are resistant to antiestrogen therapy have been found to have an active RAS-RAF-MEK pathway (47), and a cell line model of acquired resistance to antiestrogen therapy expresses abundant miR-221/222 (48, 49).

Our observations that treatment of basal-like breast cancer cell lines with a MEKi resulted in the MET could conceivably provide a strategy to sensitize resistant cells to chemotherapy and blunt metastasis. Thus, a MEKi in combination with chemotherapy might provide a feasible strategy to test in patients with basal-like breast cancer.

Materials and Methods

Cell lines, tumor samples, and reagents

All cell lines were procured from the American Type Culture Collection or DSMZ (Deutsche Sammlung von Mikroorganismen und Zellkulturen) and maintained in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% fetal bovine serum (FBS) supplemented with l-glutamine with the exception of MCF10A cell lines, which were maintained in DMEM/F12 media containing 5% horse serum, EGF (20 ng/ml), hydrocortisone (0.5 μg/ml), cholera toxin (100 ng/ml), and insulin (10 μg/ml). Breast tumor samples were obtained from ILSbio, SeraCare Life Sciences, and Asterand. ER/PR and HER2 status was determined by standard immunohistochemistry techniques. Chemically synthesized miRNA inhibitors, mimics, and scrambled controls were obtained from Exiqon or Dharmacon. siRNA SMART pools were purchased from Dharmacon. Cells were seeded at 100,000 to 200,000 cells/ml and transfected by Lipofectamine 2000 or RNAiMAX (Invitrogen) at indicated concentrations. MCF10A cells were transfected using a Nucleofactor (Amaxa) with program T-020 and Kit L. We inhibited MEK with PD0325901 (Pfizer) and inhibited EGFR with erlotinib (Tarceva, Genentech Inc.). We used 4 μM MEKi and 0.13 μM erlotinib, or as indicated in figure legends. For Western blot analysis, anti-FOSL1/FRA1 antibody from Santa Cruz (SC605) and anti-TRPS1 from R&D Systems (AF1838) were used at 1 μg/ml. Details of vector cloning are available in the Supplementary Materials.

Microarray and quantitative real-time PCR

RNA was purified using RNeasy (Qiagen) or miRVANA (Ambion) kits with DNase digestions on RNeasy columns. For mRNA microarray, RNA was run on Affymetrix HGU133P2 chips, and small RNAs were run on Ambion miRVana Bioarray V2 (Asuragen). We profiled miRNA and mRNA abundance in 134 cancer cell lines including 50 lung, 27 breast, 26 colorectal, 16 melanoma, 9 ovary, and 6 non-Hodgkin’s lymphoma. The gene expression arrays were normalized with GC Robust Multi-array Average (gcRMA). A 1:1 relationship between probe sets and genes was established by assigning the most variable probe set to a gene. The miRNA arrays were normalized with the VSN algorithm and converted to log base 2. Details of CRS are available in the Supplementary Materials. For qPCR, complementary DNA (cDNA) was generated with the High-Capacity cDNA Reverse Transcription kit (Roche). cDNA was diluted 1:20 and used for 20-μl qPCR reactions on 7900HT Fast Real Time (Applied Biosystems). The following primers and probes were used: E-cadherin: forward 5′-CCCGGGACAACGTTTATTAC-3′, reverse 5′-GCTGGCTCAAGTCAAAGTCC-3′, universal probe 35 (Roche); vimentin: forward 5′-AGCCTCAGAGAGGTCAGCAA-3′, reverse 5′-AAAGTGTGGCTGCCAAGAAC-3′, universal probe 16; and normalization genes ACTB: forward 5′-ATTGGCAATGAGCGGTTC-3′, reverse 5′-GGATGCCACAGGACTCCAT-3′, universal probe 11; GUS: forward 5′-CGCCCTGCCTATCTGTATTC-3′, reverse: 5′-TCCCCACAGGGAGTGTGTAG-3′, universal probe 57; G6PD: forward 5′-CTGGTGGCCATGGAGAAG-3′, reverse 5′-GCATTTCAACACCTTGACCTT-3′, universal probe 22; GAPDH: forward 5′-AGCCACATCGCTCAGACAC-3′, reverse 5′-GCCCAATACGACCAAATCC-3′, universal probe 60. Assays on Demand Taqman primers and probes (Applied Biosystems) were used for FOSL1, ZEB2, and TRPS1. Reactions were performed in duplicate and delta-delta-Cycle Threshold (ddCt) values were calculated on the basis of the average of the normalization genes. For miRNA, RNA was purified with the miRVANA miRNA Isolation Kit (Ambion). Taqman miRNA Assays (Applied Biosystems) were used for miR-221/222 and normalized to an average of RNU44, RNU48, and RNU6B.

Chromatin immunoprecipitation

For FOSL1, chromatin preps from MDA-MB-231 and MCF7 cells were made with the ChIP-IT Express Enzymatic kit from Active Motif according to the manufacturer’s instructions. Sheared chromatin was precipitated with 3 μg of anti-FOSL1 (Fra-1) antibody (Santa Cruz, SC183) or control rabbit antibody (Chemicon International). qPCR was used to determine levels of DNA precipitated with the following primers and probes: FRA-1 miR-221/222 binding site: forward 5′-GATGTGGTCAGCAGAGTCATTT-3′, reverse 5′-CGGTTCGTAGCTCAATTGTTAC-3′; probe: 5′-CCAGACATCTAAAAGGAAACCCACATCAA-3′; negative control [5411 base pairs (bp) 3′ to the putative FRA-1 binding site, outside of any transcriptional regulatory elements according to the UCSC Genome Browser]: forward 5′-CCTTACAACCTGTTGAAGAACTGA-3′, reverse 5′-TCCAGTATTTGCTCCCAGAGT-3′; probe: 5′-CACACGTCCACTGTGTGGCAGC-3′. For TRPS1, chromatin preps from MCF10A cells were made with the ChIP-IT Express Enzymatic kit from Active Motif according to the manufacturer’s instructions. Sheared chromatin was precipitated with 3 μg of anti-TRPS1 antibodies (G-20 and N-18, Santa Cruz) or control rabbit antibody (Chemicon International). qPCR was used to determine levels of DNA precipitated with the following primers: Prediction #1 (chr2:144,994,080 to 144,994,379): forward 5′-CCAAAAAGGGATAAAAAGAGAGA-3′, #1-1 reverse 5′-TTCATGCTTTTTCTTCTCACCA-3′; Prediction #2 (chr2:144,992,910-144,993,149): forward 5′-AATTGTTAGGGGAAATCCTGAAA-3′, reverse 5′-TTTATATGCAGTCGTGTGTCTCA-3′.

Immunofluorescence

MCF10A cells were transfected with miR-221/222 mimic or control at 100 nM with RNAiMax for 4 days. For phalloidin staining, cells were transfected in a 12-well plate and fixed in 4% paraformaldehyde for 20 min before permeabilization in 0.5% saponin and staining with rhodamine phalloidin (Molecular Probes) for 15 min. Cells were washed and nuclei were stained with Hoechst (Sigma). For vimentin staining, cells were transfected as above in an eight-chamber glass slide, fixed in ice-cold methanol for 5 min at −20°C, and permeabilized in acetone for 3 min at −20°C. Cells were then stained with antibody directed against vimentin (anti-vimentin) (Sigma-Aldrich) at room temperature for 1 hour followed by a secondary anti–mouse-Cy3 (Jackson Laboratory) and Hoechst 33258. Cells were imaged and nuclei were quantitated on the Cellomics VTI ArrayScan. For E-cadherin staining, MCF10A cells were seeded in glass-bottomed 96-well plates and transfected with 100 nM of a mixture of miR-221/222 mimic (Dharmacon, C-300578-05 and C-300579-07) or 100 nM TRPS1 siRNA (Exiqon, 4392421). On day 5, cells were fixed with 3.7% formaldehyde, permeabilized with 0.5% Triton X-100, and blocked with 1% bovine serum albumin diluted in tris-buffered saline for 30 min at 37°C. Cells were then incubated with Ms anti–E-cadherin (Abcam, ab1416) for 45 min at 37°C followed by Cy3-conjugated anti-Ms antibody for another 45 min at 37°C. The nuclei were stained with DAPI (4′,6-diamidino-2-phenylindole) Fluoromount G (Southern Biotech, 0100-20). Images were visualized with an inverted microscope (Zeiss Observer.Z1) equipped with a CoolSNAP HQ camera. A 20×/0.4 LD Plan Neofluar objective was used.

Invasion and migration assays

Transwell migration assays were performed in a 24-multiwell insert system with a porous polycarbonate membrane (8-μm pore size) according to the manufacturer’s instructions (BD Biosciences). After transfection of 100 nM miR-221/222, miR-221/222 inhibitor (Dharmacon), or NTC, the cells (5 × 104 per well) were seeded on the upper side of the filter. Horse serum (10%) (MCF10A) and 10% FBS (MDA-MB-231) were used as chemoattractants in the lower chambers. The cells that migrated to the lower surface of the membrane were stained with DiIC12 fluorescent dye and fluorescence was measured at 530/560 nm (excitation/emission) with a bottom reading plate reader, SpectraMax M5 (Molecular Devices). For the invasion assay, we used BD BioCoat Matrigel Invasion Chambers (BD Biosciences). After 3 days of transfection, cells on the upper side of the filters were mechanically removed. Cells that had crossed to the lower side of the filter were fixed and stained with Diff-Quik kit (Siemens). The filters were photographed and the cells were counted.

Adenovirus infection

Stocks of recombinant adenoviruses expressing green fluorescent protein (GFP), HRAS (G12V), and MEK1 (S217E, S221E) transgenes as well as negative control vectors were purchased and propagated in human embryonic kidney (HEK) 293 cells according to the vendor-supplied protocol (Cell Biolabs). Optimal multiplicity of infection (MOI) was determined for MCF10A cells by GFP transfection. Cells were lysed 24 hours after infection with collection of total RNA and protein. Isolated RNA was reverse-transcribed to cDNA and then run on Affymetrix HGU133P 2.0 microarrays.

Supplementary Materials

www.sciencesignaling.org/cgi/content/full/4/177/ra41/DC1

Materials and Methods

References

Fig. S1. Plot of miR-221, miR-222, and miR-200c expression across luminal and basal-like cell lines.

Fig. S2. Triple-negative breast cancers have more abundant miR-221/222 than ER/PR breast cancers.

Fig. S3. Phase contrast and immunofluorescence images of MCF10A cells transfected with miR-221/222.

Fig. S4. miR-221/222 abundance correlates with that of the mRNA encoding vimentin.

Fig. S5. miR-221/222 abundance inversely correlates with that of the mRNA encoding E-cadherin.

Fig. S6. miR-221/222 abundance inversely correlates with that of the mRNA encoding TRPS1.

Table S1. Luminal-basal specific differential gene expression.

Table S2. miR-221/222 gene targets down-regulated in basal-like subtype and in MCF10A cells overexpressing miR-221/222.

Table S3. Filtered candidate target genes for miR-221/222 based on CRS.

References and Notes

  1. Acknowledgments: We thank R. Neve, J. Fridlyand, M. Birkner, Z. Zhang, J. Settleman, and E. Thompson for their input and thought-provoking discussions. Author contributions: D.D. designed the study and wrote the manuscript. A.T.A., R.-F.Y., P.Y., and R.B. performed all microarray and statistical analyses. All other authors contributed the experimental data. Competing interests: All authors are employees of Genentech Inc., except A.T.A. Accession numbers: Microarray data are available from Gene Expression Omnibus with accession numbers GSE10843, GSE12790, and GSE29327.
View Abstract

Stay Connected to Science Signaling

Navigate This Article