Ethics Statement

This study was performed following standard ethical procedures of the Spanish regulation (Ley de Investigación Orgánica Biomédica, 14 July 2007) and was approved by the ethics committee of the Hospital Virgen del Rocío de Sevilla and the Fundación Pública Andaluza para la Gestión de la Investigación en Salud de Sevilla (FISEVI), Spain. Written informed consent has been obtained and all clinical investigation has been conducted according to the principles expressed in the Declaration of Helsinki.

Human Tumors

A series of 70 formalin-fixed paraffin-embedded (FFPE) human breast tumors were obtained from the archives of the Department of Pathology of the Hospital Universitario Virgen del Rocío, Sevilla, Spain. The series included 16 luminal breast carcinomas (ER+/PR+/HER2−), 21 HER2 carcinomas (HER2+/ER−/PR−), 26 TN tumors (HER2−/ER−/PR−), and 7 MBCs (which are triple negative as well). Clinical data of the tumor sample are provided in Table S1. Carefully examination of the H&E-stained sections was carried out by pathologists (JP, MAL, BV) confirming the diagnosis on the basis of histological criteria. Immunohistochemistry was carried out on tissue microarray sections using the Envision method (Dako, CA, USA) and primary antibodies for E-cadherin (Zymed, CA, US) and Vimentin (Novocastra, UK). Immunohistochemical staining was evaluated by pathologists (JP, MAL, BV).

Cell Lines

Six breast cancer cell lines with different phenotypes - MCF7, T47D, BT-474, SKBR3, MDA-MB-231 and BT549 - were obtained from the American Tissue Culture Collection (ATCC) and cultured according to ATCC’s recommendation. The non-tumorigenic basal cell lines MCF12A (from ATCC) and Myo1089 [37](kindly provided by Dr Michael Allen and Dr Louise Jones, Queen Mary’s University London) were grown in DMEM:F12 medium supplemented with 5% horse serum (Invitrogen, Life Technologies, Paisley, UK), 20 ng/ml EGF (PreproTech, London, UK), 500 ng/ml hydrocortisone (Sigma, Poole, UK), 100 ng/ml cholera toxin (Sigma), and 10 µg/ml insulin (Sigma). These non-tumorigenic basal cell lines were utilized as models for spontaneous EMT as previously reported [38]. The epithelial and mesenchymal-like subpopulations within these cell lines were purified by FACS sorting using EpCAM and CD49f surface markers as described before [38].

Nucleic Acid Isolation

Total RNA was extracted in triplicate from 90% confluent cells using mirVana miRNA isolation kit (Ambion, Austin, TX, USA). For FFPE tumors samples, RNA was isolated from selected areas containing >70% tumor cells. After deparaffinization, the tissue was homogenized for 2 minutes in lysis buffer, using an Ultra-turrax (T10 Basic, IKA, Germany) according to the manufacturer’s protocol (Recover All™ total nucleic acid isolation kit, Ambion). DNA from FFPE tumors samples and cell lines were prepared using QIAGEN QIAamp ™ DNA FFPE Tissue Kit and QIAamp® DNA Mini Kit respectively, according to the manufacturer’s instructions (QIAGEN). The quantity and quality of RNA and DNA were measured with the NanoDrop ND-100 Spectrophotometer (Thermo Scientific, USA).

Real-Time qRT-PCR

miRNA expression was measured by qRT-PCR using customized Taqman miRNA Arrays (Applied Biosystems). In brief, 700 ng total RNA was subjected to reverse transcription using the TaqMan MicroRNA Reverse Transcription Kit (Applied Biosystems) according to the manufacturer’s instructions. Megaplex™ RT primers (Human Pool A) were used for reverse transcription and thermal-cycling was performed over 40 cycles (16°C, 2 min; 42°C 1 min; 50°C 1 min). qRT-PCR reactions were performed on an Applied Biosystems 7900HT Fast Real Time PCR System (95°C 10 min, 40 cycles of 95°C 15 sec and 60°C 60 sec). RNU48 was selected as endogenous control for normalization due to its consistent expression among all samples.

For gene expression analysis we used 800 ng of total RNA for the reverse transcription using the High Capacity RNA to cDNA kit (Applied Biosystems) according to the manufacturer’s protocol. qRT-PCR reactions were performed using Taqman Array Cards (Applied Biosystems) including probes for: SNAI1, SNAI2, ZEB1 and ZEB2. 18S was used as endogenous control to normalize variations in the quantities of input cDNA. Applied Biosystems 7900HT Fast Real Time PCR System was used with the following PCR conditions: 50°C 2 min; 94.5°C 10 min; 40 cycles of melting 97°C 30 sec and annealing 59.7°C 1 min.

Relative mRNA and miRNA expression levels were calculated by the formula 2^−ΔΔCt using SDS software (Applied Biosystems).

DNA Methylation Analysis

1 µg of genomic DNA was used for sodium bisulfite conversion using EpiTect® Plus DNA Bisulfite kit (QIAGEN) following the supplied protocols. Primers for PCR amplification of the converted DNA were designed using EpiDesigner® software (Sequenom, San Diego, CA, USA). Two overlapping amplicons were analyzed in the promoter region of each genomic locus (miR-200b-a-429 and miR-200c-141). Primer sequences are shown in Table S2. Amplicons were analyzed using the MassARRAY MALDI-TOF mass spectrometry platform (Sequenom). Data were analyzed using EpiTyper software (Sequenom) and the methylation level of the promoter sequences was calculated by averaging the methylation level of the individual CpG units.

Statistical Analysis

Statistical significance of relative changes in miRNA and mRNA expression between different groups of breast tumors was determined by the Wilcoxon rank-sum test, correcting p-values with the Benjamini-Hochberg algorithm (p<0.05). A moderate t-test was used when analyzing statistical differences in cell line expression data. These analyses were conducted on Integromics RealTime StatMiner™ 4.2 package (Integromics S.L., Granada, Spain). SPSS (SPSS inc., Chicago, IL) and JMP statistical software (SAS Institute Inc., Cary, NC, US) were used to perform the correlation analyses. Correlation of the normalized expression data and %CpG methylation data was determined using the Pearson coefficient.

miR-200f Expression Decreases in Breast Tumors with EMT Features

We investigated miR-200f expression pattern in 70 human breast cancers comprising ER+, HER2+, and TN tumor subtypes, including a subset of MBCs with sarcomatous (homologous or heterologous) differentiation (Figure 1A). No MBCs with squamous differentiation were included. We found a significant decrease of all miR-200f members in MBCs compared to the rest of tumor types. Consistent with this, basal B breast cancer cell lines, which have been proposed to represent in vitro models of MBC and claudin-low tumors [5], exhibited significantly reduced levels of miR-200f expression compared to luminal cell lines (Figure S1). Moreover, compared with ER+ breast tumors, TN cancers also exhibited decreased expression of the whole mir-200f, while HER2+ cancers only showed significant downregulation of miR-200a and miR-141 (Figure 1A). Hence, miR-200f expression was higher in ER+ tumors, whereas TN and HER2+ tumors exhibited intermediate levels, and MBCs showed strong reduction of miR-200f compared with the rest of groups. Reduced levels of miR-200f in MBCs and claudin-low tumors relative to other tumor phenotypes has been previously reported, but detailed analysis of miR-200f in a variety of human breast cancer phenotypes were not carried out in these studies [39], [40]. Bockmeyer et al [41] analyzed miRNA expression in a profiling experiment comparing luminal A, luminal B, basal-like and malignant myoepithelioma, and observed low level miR-200c and mir-429 expression restricted to malignant myoepithelioma. However, our larger series of distinct breast cancer phenotypes allowed us to detect a significant reduction in all miR-200f members in TN tumors, in addition to the marked reduction observed in MBCs. As expected, by immunohistochemical analysis we observed that TN and MBC tumors were positively associated to the mesenchymal marker vimentin, and ER+ tumors were associated to the epithelial marker E-cadherin (Figure S2). Therefore, given that basal-like tumors are suggested to be particularly prone to acquire EMT-like characteristics [17] and MBCs are considered as examples of complete EMT [9], [10], [11], our data indicate that the expression of miR-200f could be an ideal biomarker to assess the degree of occurrence of EMT in breast cancer.

In order to assess further the association between miR-200f and EMT transcriptional regulation in breast cancer we analyzed the expression of diverse EMT-transcriptional inducers including ZEB1 and ZEB2, which have been reported to directly inhibit transcriptional activation of miR-200f expression through conserved E-box elements [27], [28]. We found a strong and significant increase in the expression of SNAI1 and SNAI2 and a moderate but not statistically significant increase of ZEB1 and ZEB2 in MBCs compared with the rest of tumor types (Figure 1B). There were no additional differences in the rest of tumor groups for any of the analyzed genes. In agreement with our observations in tumor samples, we detected higher levels of EMT-transcriptional inducers in basal B cell lines, which exhibited upregulation of SNAI2, ZEB1 and ZEB2 (Figure S1). These results indicate that EMT-transcriptional inducers could account for mir-200f downregulation in our tumor series. Accordingly, we observed significant negative correlations between SNAI1/SNAI2 (but not ZEB1 or ZEB2) and all miR-200f members’ expression levels in our tumor series (Figure 1C and Figure S3). Since SNAI1/SNAI2 are E-cadherin repressors affecting promoter transcriptional activity through binding to E-box elements, it is conceivable that these EMT-transcriptional inducers could affect transcriptional activation of miR-200f through the conserved E-box elements that have been functionally defined in previous studies [27], [42]. Consistent with this idea, we demonstrated that overexpression of SNAI2 in MDCK cells had a strong repressive effect in the transcriptional activation of miR-200b-a-429 promoter, and this effect was strictly dependent on E-box motifs (Figure S4). This observation supports the notion that other EMT-transcriptional inducers, distinct from ZEB1 and ZEB2, could directly regulate miR-200f expression. In fact, during the preparation of this manuscript a study has been published describing a mutually inhibitory regulatory loop for SNAI2 and miR-200b/miR-1 in a murine model of aggressive prostate cancer. Transcriptional repression of miR-200b-a-429 and miR-1 by SNAI2 occurs by direct interaction through E-box elements; and conversely, SNAI2 was functionally validated as target gene for miR-1 and miR-200b [43]. As miR-1 and miR-200 target sequences identified in SNAI2 3′-UTR are well conserved, mutually repressive regulation for SNAI2 and miR-1/miR-200b-a-429 is likely to be a common regulatory mechanism operating in human and other species. SNAI2 has been described as the EMT-transcriptional inducer responsible for conferring a basal-like phenotype in breast cancer [44], [45], so it would be possible that this new regulatory axis plays an important contribution to the EMT-like features observed in these tumors.

Hypermethylation of miR-200c-141 Promoter Sequences in MBCs

We next studied the possible mechanisms by which miR-200f expression could be regulated in the different breast cancer subtypes. Previous studies have shown that mir-200f is subjected to epigenetic regulation in cancer and normal tissues [32], [33], [34], [35], [36], [42]. The promoter sequences driving transcriptional activation of both miR-200f clusters are located within CpG islands, and DNA methylation in these CpG islands is associated to inactivation of miR-200f in cancer [46]. We therefore evaluated the CpG methylation rates of the respective miR-200f promoter regions in our tumor series using the Sequenom MassArray® MALDI-TOF platform. The promoter sequences analyzed for miR-200b-a-429 cluster span 415 bp encompassing the transcription start site (TSS) functionally defined by Bracken and colleagues [27]. For miR-200c-141 cluster we studied a region of 426 bp surrounding the TSS that has also been previously determined [42] (Figure 2A).

For these analyses we focused our study in ER+, TN and MBC tumors and observed that DNA methylation levels of miR-200c-141 locus were significantly higher in MBCs compared with either ER+ or TN tumors (Figure 2B). Although the methylation levels of miR-200b-a-429 locus were slightly increased in MBCs, there were no significant differences among the different tumor types. Interestingly, the expression of miR-200c-141 and its DNA methylation levels showed a significant inverse correlation in breast tumors, but no such association was observed for miR-200b-a-429 cluster (Figure 2C). In agreement with these data in breast tumors, basal B cancer cell lines also exhibited elevated DNA methylation rates at miR-200c-141 promoter, reaching levels >80%, as observed in MBCs. Additionally, we also detected elevated methylation at miR-200b-a-429 locus in basal B cell lines (>50%), that correlates with downregulation of miR-200b, miR-200a and miR-429 (Figures S5 and S1). Taken together, these results indicate that DNA methylation is a major epigenetic mechanism regulating transcriptional activation of miR-200c-141 cluster in MBCs/basal B cancer cells and suggest that this epigenetic event is associated with the acquisition of EMT features in these tumors.

A recent study [47] has functionally defined an independent promoter region for miR-200b-a-429 cluster, capable of directing transcriptional activity at the same level as the promoter sequences formerly described [27] and studied here. Remarkably, the two promoter sequences were differentially methylated in breast cancer clinical specimens suggesting different regulatory roles [47]. Similarly, Vrba et al [33] described that normal human mammary cells exhibit differential methylation of the proximal and distal parts of the CpG island (Figure 2A), where the respective miR-200b-a-429 promoter regions are located. Wee et al [47] have also shown that methylation of both promoter regions is inversely associated with miR-200b expression levels in breast cancer cell lines as well as in clinical specimens. However, we did not observe such correlation in our tumor series for the analyzed promoter region (Figure 2C). This discrepancy could be due to differences in sample size. The authors also reported that hypermethylation of the newly identified promoter, but not of the formerly defined one, was associated with loss of ER+ or PR+ expression, and gain of HER2+ phenotype. In agreement with this we did not observe differences in DNA methylation levels of the former promoter region in TN or MBCs (which lack expression of ER and PR), compared with ER+ tumors (Figure 2B).

It has recently been reported that the two clusters comprising the miR-200f display different but complementary epigenetic mechanisms to regulate their expression in human mammary cells [33]. The authors describe that miR-200b-a-429 cluster is epigenetically repressed in mammary fibroblasts by the repressive histone mark H3K27me3, while DNA methylation is the major epigenetic mechanism repressing the transcriptional activation of miR-200c-141 cluster. Our results are in accordance with these observations, since we only observed expression/DNA methylation correlation for miR-200c-141 cluster (Figure 2C). Nevertheless, it would be necessary to determine other epigenetic marks in order to establish the actual epigenetic state of miR-200f loci. Indeed, in hepatocellular carcinoma, miR-200b-a-429 promoter sequences show decreased histone H3 acetylation, a permissive epigenetic mark that is in turn modulated by miR-200a through direct targeting of the histone deacetylase 4 (HDAC4) [48].

Spontaneous EMT in Normal Basal Breast Cell Lines is Associated to Hypermethylation of miR-200c-141 Locus

To study further the association between modulation of miR-200f and induction of EMT in breast cancer and to validate the epigenetic regulation of miR200 during this process, we utilized our recently described model systems for spontaneous EMT in mammary cells [38]. We have demonstrated that the immortalized non-tumorigenic cell lines MCF12A and Myo1089 contain two functionally and phenotypically distinct subpopulations: EpCAM^pos/CD49f^high cells (referred as to ‘EpCAM+’) show the typical cobblestone epithelial morphology and phenotypically resemble basal A breast cancers, while EpCAM^neg/CD49f^med/low cells (‘Fibros’) show fibroblast-like spindle morphology and exhibit a gene expression pattern resembling stromal fibroblast and Basal-B/claudin-low cancers [38]. Importantly, the Fibros subpopulation arises from EpCAM+ cells through an EMT-like process.

Here, we sorted these two subpopulations from MCF12A and Myo1089 cells and studied the expression of the miR-200f members and their DNA methylation status as well as expression of EMT-transcriptional inducers. As expected, the expression levels of all of miR-200f members showed a significant decrease in Fibros compared to EpCAM+ subpopulation in both cell lines (Figure 3A). Moreover, downregulation of miR-200f in the Fibros subpopulation is accompanied by an increase in the expression levels of ZEB1 and ZEB2 (Figure 3B and [38]), while differential expression of SNAI1 and SNAI2 was only observed in MCF12A and Myo1089 cell lines, respectively. Importantly, Fibros subpopulation from Myo1089 cells showed increased methylation level of miR-200c-141 locus but not of miR-200b-a-429 locus (Figure 3C), which parallels the result observed in MBCs (Figure 2B). No differential DNA methylation levels were found in the MCF12A cell line, but this could be due to the fact that EMT in this cell line is more transient than in Myo1089 cells [38]. Overall, these data reproduced what we observed in MBCs, which exhibited strong reduction of miR-200f with hypermethylation at miR-200c-141 locus, and increased levels of EMT-transcriptional inducers.

Conclusion

We have observed that TN and specially MBCs breast cancers exhibit lower level expression of miR-200f compared to ER+ tumors. Using an in vitro model of EMT and the analysis in human breast cancer samples we have shown that the strong decrease of miR-200f expression levels is accompanied by upregulation of EMT-transcriptional inducers and hypermethylation of miR-200c-141 locus, thus supporting the functional association between EMT and miR-200f regulation in breast cancer. Therefore, our data indicate that expression and methylation status of miR-200f could be used as a hypothetical biomarker to assess the occurrence of EMT in breast cancer both in vitro and in vivo.