|Home | About | Journals | Submit | Contact Us | Français|
Chemotherapy resistance frequently drives tumour progression. However, the underlying molecular mechanisms are poorly characterized. Epithelial-to-mesenchymal transition (EMT) has been shown to correlate with therapy resistance, but the functional link and signalling pathways remain to be elucidated. We report here that miR-30c, a human breast tumour prognostic marker, plays a pivotal role in chemo-resistance by a direct targeting of TWF1, which encodes an actin-binding protein and promotes EMT. An IL-6 family member, IL-11 was identified as a secondary target of TWF1 in the miR-30c signalling pathway. Expression of miR-30c inversely correlated with TWF1 and IL-11 levels in primary breast tumours and low IL-11 correlated with relapse-free survival in breast cancer patients. Our study demonstrates that miR-30c is transcriptionally regulated by GATA3 in breast tumours. Identification of a novel miRNA-mediated pathway that regulates chemo-resistance in breast cancer will facilitate the development of novel therapeutic strategies.
Therapy resistance, whether intrinsic or acquired, is one of the most fundamental problems in cancer and contributes to local and distant tumour recurrence and disease progression. A better understanding of the underlying molecular mechanisms will undoubtedly lead to the development of more effective therapies. Non-coding single-stranded microRNAs (miRNAs, miRs) have emerged as important epigenetic regulators of protein coding genes and aberrant miRNA expression has been found in many cancers1. However, the role of miRNAs in chemo-resistance remains to be characterized. In addition, the epithelial-to-mesenchymal transition (EMT) has been connected to breast cancer chemo-resistance2,3 but the regulatory link is unknown. We therefore hypothesized that miRNAs that regulate EMT might be critical for regulating drug resistance. MiRNAs inhibit target genes primarily by degrading mRNA or by repressing translation via seed sequence-mediated imperfect, complementary binding to 3’ untranslated regions (3’UTRs)4,5. By combining microarray analyses, miRNA target prediction algorithms, and experimental validation approaches, we sought to identify miRNA targets that regulate EMT and drug resistance in breast cancer cells.
To identify miRNA biomarkers, we profiled primary breast tumour miRNA expression levels and performed clinicopathologic correlation analyses for breast tumours collected under IRB approved protocols at the University of Chicago Breast SPORE and in independent datasets.
Frozen primary breast tumour tissues (University of Chicago [UC] tumour set, n=46) and adjacent normal tissues (n=5) were first profiled for both gene and miRNA expression. 152 of 757 expressed miRNAs passed the filtering criteria for subsequent unsupervised hierarchical clustering (Fig. 1a; see also patient information in Supplementary Table S1). Unsupervised clustering divided the miRNAs into Cluster A and Cluster B, and these miRNAs displayed distinct expression patterns in the UC breast tumour set. Specifically, Cluster A miRNAs were up regulated in Group One tumours and Cluster B miRNAs were up regulated in Group Two tumours.
The hsa-miR-30 family segregated with the Cluster B miRNAs with high expression in Group Two patient tumours, including many known tumour-suppressor-miRs such as the let-7 family members (Fig. 1a). To determine the role of the miR-30 family members in breast cancer progression, we first examined their clinical relevance for patient survival. In an independent Oxford breast tumour dataset containing a retrospective series of 10-year follow up cases (n=210) with matched miRNA and mRNA expression arrays (GSE 22216 and GSE22220 respectively)6, we identified miR-30c as well as miR-30a, −30a*, and −30e* as individual favourable breast cancer prognostic markers for distant relapse-free survival (DRFS) (Fig. 1b–f). The prognostic role of miR-30c is still significant even when adjusted for age, tumour size, ER status, nodal status, and histological grade (Supplementary Table S2). Furthermore, when breast tumour subtypes are classified based on the PAM50 subtype predictor7 and the 9 Cell Line Claudin-low predictor8, miR-30c expression is associated with breast tumour subtypes in two independent datasets (UC tumour set and Oxford tumour set6) (Supplementary Fig. S1a–c) and its prognostic role is subtype-dependent (Table S3). However, the miR-30c levels were not significantly different between paired primary breast tumours and normal breast tissue adjacent to tumours (Supplementary Table S4).
By univariate analysis of a previous report, both miR-30a* and miR-30c have been shown to correlate with the benefit of tamoxifen treatment and longer progression free survival in oestrogen receptor (ER)-positive breast cancer patients9. However, in that study miR-30c was the only independent miRNA predictor of tamoxifen treatment response by multivariate analysis when adjusted for other clinico-phathologic factors9. We hypothesized that the miR-30 family, especially miR-30c, might functionally regulate therapy response in addition to having an association with treatment benefits. We therefore focused on their role in chemo-resistance regulation in breast cancer.
To test our hypothesis that miR-30c plays a role in breast tumour progression and chemo-resistance, we investigated the relevance of miR-30c to chemotherapy-induced cytotoxicity of tumour cells. Interestingly, for six breast cancer cell lines (T47D, MCF-7, MDA-MB-231, BT-20, HCC-70, and HCC-38), we observed a significant inverse correlation between miR-30c levels and cell survival in response to paclitaxel (0.1–50 nM) and doxorubicin (1–500 nM) treatments, represented as Log2-transformed expression levels and AUC (the Area Under the survival Curve, in a series of 6 treatment doses) (Fig. 2a–b, p=0.04). These correlations are more significant at the highest doses of paclitaxel (50 nM) and doxorubicin (500 nM) (Supplementary Fig. S2a–b and Table S5, p=0.0198 and 0.0191 respectively).
We further characterized the role of miR-30c in regulating chemotherapy resistance by modulating its levels in breast cancer cells. Transient transfection of miR-30c resulted in a significant decrease in survival of chemo-resistant MDA-MB-231 breast cancer cells to paclitaxel (Taxol, 0.5–10 nM) and doxorubicin (10–100 nM) at even low doses (Fig. 2c–e). Growth curve analyses suggested that increased miR-30c levels did not affect cancer cell growth (Fig. 2f). Elevated miR-30c also sensitized the drug response of a second breast cancer cell line BT-20 to paclitaxel and doxorubicin (Fig. 2g). Furthermore, DNA content analysis confirmed that miR-30c significantly increased the apoptotic sub-G1 population of these cancer cells upon paclitaxel and doxorubicin treatments (Fig. 2h).
We also examined the effects of two other individual favourable breast cancer prognostic markers in the miR-30 family, miR-30a and −30a*, on the response of MDA-MB-231 breast cancer cells to chemotherapy drugs. While ectopic miR-30a did not show any effect on doxorubicin treatment at various doses up to 200 nM, it did decrease breast cancer cell survival in response to paclitaxel at very high doses (50 nM) (Supplementary Fig. S3a–b). In addition, miR-30a* sensitized the drug response of breast cancer cells to doxorubicin (200 nM) and paclitaxel (5–50 nM) at relatively high doses (Supplementary Fig. S3a–b).
To examine the molecular mechanisms by which miR-30c regulates chemo-resistance, we investigated its target genes and focused on a few that relate to EMT regulation and cancer progression. Mammalian miRNAs regulate target genes primarily by decreasing mRNA levels10. Multiple computational algorithms, such as GeneSet2miRNA which identifies miRNA activities by at least four of eleven established miRNA target prediction programs11, can predict potential interactions between genes and miRNAs. To strategically identify direct miR-30c target genes, we used global transcriptome analysis (microarray) in combination with GeneSet2miRNA and experimental validation analyses, including immunoblots, luciferase assays, and functional rescue studies.
We first performed microarray analyses of MDA-MB-231 cells that were transiently transfected either with miR-30c oligos or the scrambled control. High miR-30c levels significantly inhibited the expression levels of 291 genes in many networks and pathways, such as anti-apoptosis and cytoskeleton genes (Fig. 3a; see also Supplementary Dataset 1). Using this list, we performed online GeneSet2miRNA11 target gene prediction analysis and narrowed the direct targets of miR-30c down to 77 potential genes, predicted by a minimum of 4 out of 11 established algorithms (Fig. 3a; see also Supplementary Table S6). Among these, the cytoskeleton genes twinfilin 1 (TWF1, also known as PTK9) and vimentin (VIM) were reported to regulate cell motility, drug sensitivity, and cancer progression2,12–15.
The function of TWF1 as a regulator of drug sensitivity was previously revealed by in vivo RNAi screening in a lymphoma model, along with other regulators of actin dynamics, such as Rac212. TWF1 is a conserved actin-binding protein with two actin depolymerizing factor homology (ADF-H) domains16 and belongs to the ADF-H family along with ADF/cofilin, Abp1/drebrin, coactosin, and GMF17. It regulates diverse morphological and motile processes by both sequestering ADP-actin monomers and capping filament barbed ends18,19. We hypothesized that TWF1 is an important target of miR-30c by linking drug resistance to actin-based cytoskeletal organization, such as the stress fiber F-actin, an EMT marker. To address this hypothesis, we performed experimental validation analyses. Inhibition of TWF1 expression by miR-30c was confirmed in MDA-MB-231 cells by real-time PCR and western blots (Fig. 3b–c). In addition, transient transfection of an anti-miR-30c inhibitor increased TWF1 expression in these breast cancer cells (Fig. 3b). We also observed inhibitory effects of miR-30c on the EMT markers VIM and SNAI2 at protein levels compared to the scrambled miRNA control or a mock transfection control of MDA-MB-231 cells (Fig. 3c). Accordingly, elevated miR-30c promoted a round epithelial cell phenotype and reduced the stress fiber F-actin formation in MDA-MB-231 cells (Fig. 3d; large images in Supplementary Fig. S4). Further analyses and quantitative counting of stained vinculin revealed a reduction in focal adhesions, which further validated the effect of miR-30c on F-actin dynamics (Supplementary Fig. S5a–b).
We then investigated whether TWF1 is a direct target of miR-30c. The functional interaction between miR-30 and the 3’UTR of TWF1 was examined in luciferase assays in HEK293T cells. Compared to the scrambled control, co-transfected miR-30c suppressed the expression of the luciferase gene, which was located upstream of the wild type 3’UTR of TWF1 (Fig. 3e; see also Supplementary Fig. S6). Mutation of all three predicted recognition sites within the 3’UTR of TWF1 (Supplementary Fig. S6) abolished the inhibitory effect of miR-30c on luciferase activity (Fig. 3e). These data indicate that the inhibitory function of miR-30c on the target gene requires interaction with the predicted binding sites.
To determine the importance of TWF1 in miR-30c–regulated drug resistance, we performed functional rescue assays by co-transfecting 3’UTR-deficient TWF1 cDNA and miR-30c in MDA-MB-231 cells. While miR-30c expression induced inhibitory effects on chemo-resistance, recombinantly expressed TWF1 (resistant to miR-30 inhibition) reversed miR-30c–mediated sensitivity of the cancer cells to doxorubicin and paclitaxel (Fig. 4a).
Equally important, knockdown of TWF1 by siRNAs copied the phenotype of high miR-30c levels by sensitizing breast tumour cells to chemotherapeutic agents doxorubicin and paclitaxel at 100nM and 10nM treatment doses respectively (Fig. 4b–c).
We further investigated the effect of TWF1 knockdown on F-actin formation in MDA-MB-231 cells. Notably, siRNAs of TWF1 dramatically inhibited F-actin organization, reduced vinculin-stained focal adhesions, and promoted the mesenchymal-to-epithelial transition (MET) with a round cell shape (Fig. 4d; see also Supplementary Fig. S4 and S5a–b).
In addition to TWF1, VIM is another EMT-related cytoskeleton gene that is inhibited by miR-30c, as shown by microarray and western blot analyses (Fig. 3a and 3c). Overexpression of VIM reversed the drug sensitivity caused by miR-30c (Supplementary Fig. S7a), suggesting that VIM is an important target of miR-30c and may partially overlap with the effect of TWF1 in regulating drug resistance. However, knockdown of VIM is not sufficient to mimic the functions of miR-30c or increase drug sensitivity (Supplementary Fig. S7b).
Interleukins (ILs) and cytokines, including IL-6 family members, have been shown to contribute to drug resistance in breast cancer and other cancers20–23. Based on the microarray (SAM) analysis, we observed that miR-30c suppressed expression of IL-11, one of the IL-6 family members, in breast cancer cells (Fig. 3a; see also Supplementary Dataset 1). The specific regulation of IL-11 expression by miR-30c was validated by real-time PCR analyses, demonstrating both an inhibitory effect of miR-30c oligos and a promoting effect of anti-miR-30c on IL-11 expression in MDA-MB-231 cells (Fig. 5a).
To determine the role of IL-11 in the miR-30c pathway, we also performed functional rescue studies by modulating IL-11 expression levels and functions. Overexpression of IL-11 reversed miR-30c–mediated drug sensitivity of MDA-MB-231 cells to doxorubicin and paclitaxel (Fig. 5b). In addition, we examined the effect of a neutralizing antibody to human IL-11 on drug response of MDA-MB-231 cells. The anti-IL-11 IgG added to culture media indeed copied the function of miR-30c by sensitizing cells to the doxorubicin treatment (Fig. 5c), suggesting that the secreted form of IL-11 plays a role in mediating drug resistance. Similarly, siRNA-mediated knockdown of IL-11 sensitized the response of MDA-MB-231 cells to paclitaxel (Fig. 5d).
Based on Geneset2miRNA target gene predictions, IL-11 was not on the list of predicted direct targets of miR-30 (Supplementary Table S6), suggesting that it may be a secondary target of miR-30c. We therefore explored the regulatory link between miR-30c and IL-11. Since actin affects the activity of myocardin-related transcription factors (MRTFs) and serum response factor (SRF), which regulate expression of IL-1124, we hypothesized that TWF1, also an upstream regulator of actin dynamics, regulates IL-11 expression. We therefore examined the effect of modulated TWF1 on IL-11. As anticipated, siRNA-mediated knockdown of TWF1 inhibited IL-11 mRNA and protein expression in MDA-MB-231 cells, similar to the effect of miR-30c on IL-11 (Fig. 5e–f). These data suggest that IL-11 is an important target of TWF1 in the miR-30c signalling pathway, which regulates breast cancer chemo-resistance.
We then examined whether IL-11 also regulated F-actin formation. Indeed, knockdown of IL-11 displayed a MET phenotype very similar to that of miR-30c or knockdown of TWF1, as demonstrated both by F-actin staining and by counting of vinculin-stained focal adhesions (Fig. 5g; see also Supplementary Fig. S4 and S5a–b).
To validate the role of miR-30c as a regulator of breast tumour progression in vivo, we examined the effects of miR-30c precursor expression on chemo-resistance, using triple negative human breast tumour xenograft models generated from breast tumour-initiating cells (BTICs)25. To facilitate the monitoring of miR-expressing cells in vivo, we cloned the miR-30c precursor into a lentiviral vector (pFU-miR-30c–PGK-L2G) that co-expresses the dual-reporter luciferase gene Luc2-eGFP (L2G) in transduced cells. Based on bioluminescence imaging following orthotopic transplantation, miR-30c precursor expression did not alter BTIC-mediated tumour growth in vivo (Supplementary Fig. S8a–b). At a dose of 1mg/kg, doxorubicin did not inhibit tumour growth mediated by BTICs expressing the empty L2G vector control but did significantly inhibit the growth of L2G-30c–expressing tumour cells (n=20, p≤0.00001) (Fig. 6a). This result indicates that expression of the miR-30c precursor reversed the resistance of the human breast tumour model to doxorubicin treatment in vivo.
To examine the functional relevance of downstream signalling pathways in vivo, we measured the expression levels of target genes in sorted breast cancer cells. Compared to vector L2G–transduced cancer cells, inhibition of TWF1 and IL-11 expression was confirmed in sorted miR-30c–expressing human breast tumour cells isolated from tumour models in vivo (Fig. 6b). Furthermore, the reduction of TWF1 expression at the protein level was confirmed in miR-30c–expressing tumours by immunohistochemical staining (Fig. 6c). These studies convincingly demonstrated that miR-30c regulates the downstream TWF1 and IL-11 targets and signalling pathways in vivo.
To investigate mechanisms by which miR-30 is down-regulated in Group One breast tumours, we examined possible regulation by transcription factors. JASPAR online analysis (jaspar.genereg.net)26 identified predicted transcription factor (TF) binding sites in the putative miR-30c promoter regions, including GATA3 binding sites (Supplementary Table S7). TF-miRNA association analyses showed that GATA3 mRNA expression levels correlated with miR-30c expression levels in breast tumours (UC set, n=44, p=0.023) (Fig. 7a); a separate Oxford breast tumour dataset showed a similar correlation6 (n=210, p<0.001) (Fig. 7a). In addition, forced expression of wild-type GATA3 induced a significant increase (8–9 fold) of miR-30c expression in GATA3-deficient MDA-MB-231 breast cancer cells (n=6, p= 1.95E-08) (Fig. 7b). These data indicate that GATA3 regulates miR-30c in breast tumours.
To determine whether GATA3 binds directly to the miR-30c promoters (30c1 and 30c2), we performed chromatin-immunoprecipitation (ChIP) assays with anti-GATA3 antibodies (Santa Cruz) using lysates of MCF-7 cells that express endogenous GATA3 and lysates of MDA-MB-231 cells that contain ectopically expressed GATA3. By ChIP and real-time PCR analyses, we observed an interaction between GATA3 with the 30c2 promoter, which contains two predicted binding elements in both cell lines and an interaction between GATA3 and the 30c1 promoter in MDA-MB-231 cells (Fig. 7c–d). These data collectively demonstrate that GATA3 transcriptionally regulates miR-30c expression in breast tumour cells.
To validate the importance of the miR-30c signalling transduction pathway in clinical tumours, we examined the association of miR-30c with its target genes. Based on the Oxford breast tumour dataset 6 with combined mRNA and miRNA expression profiles, we observed a negative correlation between miR-30c and IL-11 expression (Fig. 8a, n=210, p=0.005). In the recently completed Breast Cancer dataset in The Cancer Genome Atlas (TCGA) 27, we also verified inverse correlations between miR-30c and TWF1 and between miR-30c and IL-11 (Supplementary Table S8, n=521, p=0.0077 and 0.0002, respectively).
In addition, relapse-free analysis with and without adjustment for nodal status, ER, grade, treatment, and tumour stage demonstrated that IL-11 is strongly associated with clinical outcome in three independent breast tumour sets (UNC337, Oxford tumour set, and UC tumour set) for which IL-11 is a poor prognostic marker (Fig. 8b; see also Supplementary Table S9).
The above data suggest a signalling network (Fig. 8c) that links miR-30c with its upstream transcription factor GATA3 and with its downstream targets TWF1 and IL-11 and demonstrates the clinical importance of the miR-30c regulated pathway in human breast cancer.
Our results implicate miR-30c as a phenotypic regulator of chemo-resistance. While our studies focus on the role of miR-30c as a regulator of chemotherapy response in breast tumours, miR-30 expression has been correlated with endocrine therapy resistance, which is often linked to chemotherapy resistance in advanced oestrogen receptor positive breast cancer9.
EMT (VIM expression) has been linked with breast tumour initiating cell resistance to endocrine therapy and chemotherapy2. Our data demonstrate that miR-30c regulates breast cancer chemo-resistance and EMT (F-actin formation) by direct targeting of the cytoskeleton gene TWF1 as well as by indirect targeting of the cytokine IL-11. While TWF1 is known to play a significant role in regulating cancer cell chemo-resistance 12, the underlying mechanisms are not known. We have identified IL-11 as a relevant downstream target of TWF1 and demonstrated novel roles for TWF1 and IL-11 in regulating F-actin formation, an EMT marker. IL-11 belongs to the IL-6 family, which have been associated with drug sensitivity in various cancers and epithelial tissues 21,23,28. Interestingly, TWF1 regulates the expression of IL-11 at both the mRNA and protein levels, although the detailed mechanisms remain to be elucidated. Since TWF1 is an actin-binding protein, it is likely that TWF1 regulates IL-11 through sequestering actin monomers, thereby reducing the free actin monomer pool. A reduced actin monomer pool could release actin-bound myocardin-related transcription factors (MRTFs), especially MRTFA, from trapping, thereby activating MRTFs for nuclear translocation24,29. Activated MRTF binds to and assists serum response factor (SRF) to trans-activate IL-11 transcription, as MRTFA−/− cells display reduced IL-11 expression levels24.
It is of great interest to explore regulatory mechanisms of aberrantly expressed miRNAs in breast cancer. Our studies have identified GATA3 as a transcriptional regulator of miR-30c. GATA3 plays an important role in both normal mammary development and breast tumour differentiation, and its expression correlates with luminal epithelial phenotype and ER status30–32. MiR-30c promoter analyses also show predicted binding sites for other transcription factors (e.g. YY1 and BRCA1) that may contribute to the regulation of miR-30c expression. However, we did not observe significant correlation between YY1 and BRCA1 expression and miR-30c in the UC tumour set (data not shown). In addition to transcriptional regulation, other mechanisms can contribute to deregulated miRNA expression, such as chromatin deletions/amplifications, translocations, promoter silencing, and deficient post-transcriptional processing. In a separate study, we observed copy number variation (CNV) at chromatin regions containing miR-30c1 and miR-30c2 genes33.
In conclusion, we have identified a novel miRNA-mediated pathway that regulates chemo-resistance in breast cancer. The components of this pathway, including GATA3, miR-30c, TWF1, and IL-11, represent attractive targets for developing new therapeutic strategies in the management of breast cancer.
All experiments were performed under the approval of the Institutional Biosafety Committee, Institutional Review Board and the Administrative Panel on Laboratory Animal Care of University of Chicago. All experiments had multiple replicates.
Total RNAs were extracted from frozen tissues and cultured cells for DNA microarrays (Agilent Technologies) as described34. All microarray and patient clinical data are available at UNC Microarray Database (https://genome.unc.edu) and in the Gene Expression Omnibus (GEO) under accession number GSE22049 (clinical tumour/tissues) and GSE32617 (MDA-MB-231 cells). Breast tumours subtype determination was based on gene expression (PAM50 5-subtype predictor7 and claudin-low predictor35) as well as immunohistochemical staining for ER, PR, HER2, EGFR, CK5/6, and CK8/1836.
miRNA arrays, deposited in GEO under accession number GSE39543, were performed for frozen primary breast specimens (n=51) by Exiqon (Woburn) using the miRCURY™ Hy3™/Hy5™ power labelling kit and the miRCURY™ LNA Array (v.10.0; 757 human miRs). The expression values are log2 (Hy3/Hy5) ratios. Unsupervised hierarchical clustering of miRNAs was performed34.
Log2-transformed miRNA levels were used in the miR-phenotype and miR-gene association tests. Expressions of miRNAs across breast tumour subtypes were compared using analysis of variance (ANOVA). P values were calculated using permutation test. For chemo-resistance association analysis, AUC (the area under the survival-concentration curve) was calculated by “”. A linear correlation test with Log 2 transformed values of AUC (or % survival) and miRNA levels were applied in the association test using R software. The correlations between genes and miRNA-30c expression were interrogated in the UC breast tumour set (n=44) and The Cancer Genome Atlas (TCGA) Breast Cancer dataset (n=521 breast samples)27. The normalized gene (“BRCA.exp.547.med.txt”) and miRNA (“BRCA.780.precursor.txt”) expression matrixes were used as provided in the TCGA portal (http://tcga-data.nci.nih.gov/docs/publications/brca_2012/). Pearson’s correlation coefficients were then determined between the expression of each gene and miRNA-30c (“hsa-mir-30c.MIMAT0000244”).
Total RNAs for array analysis were extracted using Trizol (Invitrogen) and RNAeasy mini kit (Qiagen) or RNA was precipitated with isopropanol and glycogen (Invitrogen). RT- PCR for miRNAs/genes used individual miRNA/gene Taqman primers (Applied Biosystems). RNU44 and U6 primers were used for miRNA internal controls and GAPDH for a housekeeping gene control.
A lentiviral gateway vector pFU-attr-PGK-L2G was constructed based on pFU-L2G25 and subcloned with miRNA precursor entry clones using LR clonase II (Invitrogen). Human TWF1 cDNA was subcloned into pDEST40 (Invitrogen) from a donor vector (ORFeome collection, the University of Chicago). Human IL-11 and GATA3 (transcript variant 2) expression vectors were purchased from Origene. Luciferase vector containing the 3’UTR of TWF1 was obtained from Switchgear Genomics. A site directed mutagenesis kit (Stratagene) with appropriate primers (IDT) was used to mutate miR-30-binding sites.
Cells were maintained in DMEM (MDA-MB-231, MCF7, and Hek293T with G418) or EMEM (BT-20) with 10% fetal bovine serum (FBS) + 1% Penicillin-Streptomycin. MiRNAs (Dharmacon, negative control #4), anti-miRNAs (Exiqon, negative control A) and siRNAs (pooled) (Dharmacon) were transfected using Dharmafect (Dharmacon) at 100nM (MDA-MB-231) or 50nM (BT-20, Hek293T) and repeated on the following day. 3’UTR luciferase vectors and cDNA vectors were transfected using Fugene (Roche-Promega). For target gene rescue experiments, cells were transfected twice with cDNA vectors after two six-hour miRNA transfections.
Transfected cells were plated in phenol red free media. Next day calcein AM (Invitrogen) was added to the wells at 4 µg/µl and incubated for an hour at 37°C. Plates were read at 485/535 nM on the Wallac 1420 plate reader (Perkin Elmer).
Cells were harvested 48 hours after transfections, lysed and sonicated in RIPA buffer. 100 µg of protein was loaded onto 4–20% gradient gels for immunoblots with antibodies to TWF1 (Genetex, GTX11439), Vimentin (Abcam, ab61780), SNAI2 (Cell signalling #9585), and β-Actin (Sigma-Aldrich, AC-15). Blots were read on Odyssey (Li-Cor).
Cell culture media were collected 36 hours post transfections. IL protein levels in the supernatants were measured using human cytokine magnetic beads and detection kit (Millipore) and read by Bioplex (BioRad).
Hek293T cells were transfected with miRNAs followed by transfection overnight with 3’UTR luciferase and control renilla vectors (Promega). Cells were lysed 48 hours later and measured using a dual luciferase kit (Promega). The luciferase signal was normalized to renilla.
Paclitaxel and doxorubicin were obtained from either Chemopharmacy at the University of Chicago or Sigma. Breast cancer cells (ATCC) were treated for 72 hours with doxorubicin (0 to 1000 nM) and paclitaxel (0–10 nM). Cytotoxicity was measured using AlamarBlue (Invitrogen). Survival of cells were measured using cell titer blue (Promega) after 72 hours (MDA-MB-231) or 48 hours (BT-20) treatment. Anti-IL-11 neutralizing antibody (R&D systems) was added at 12ug/ml before treatment. One day post transfection, cells were exposed to doxorubicin (100 nM) and paclitaxel (10 nM) for 24 hours and collected for fixation in 70% ethanol and staining with propidium iodide (Sigma). DNA content was analysed using BD LSRII flow cytometer.
Transfected cells were plated on microscope slides (Millipore) for F-actin and focal adhesion staining. Cells were fixed in 4% paraformaldehyde/PBS, washed, and stained with fluorescent phalloidin (Invitrogen F432) and SlowFade Gold antifade reagent with DAPI (Invitrogen) or anti-vinculin (ab18058-abcam). Secondary antibody was anti-mouse Texas Red (T-862-invitrogen). Cell images were taken using Axiovert 100tv (fluorescent), Axiovert 200M (phase contrast) (Zeiss) or Olympus spinning disc confocal microscope. Focal adhesions were quantified using ImageJ software with 50 cells counted per well. Tumour and lung samples were fixed in 10% neutral buffered formalin, embedded, and sectioned. Immunohistochemical staining was performed using GFP (Cell Signaling, #2956) and hTWF1 (GeneTex, GTX11439) antibodies at Histology facility. Images were taken by Pannoramic Scan Whole Slide Scanner (CRi, Cambridge Research and Instrumentation).
GATA3 antibodies (Santa Cruz, cat# 268 for MDA-MB-231 cells, cat#269 for MCF-7 cells) were used for ChIP according to the EZ ChIP™ Chromatin Immunoprecipitation Kit (Millipore Upstate Cell Signaling Solutions Catalog #17–371). DNA was purified using a phenol-chloroform extraction and precipitated by ethanol and glycogen. DNA was quantified using Fast SYBR Green Master Mix (Applied Biosystems). The sequences of 30c1-P1 primers: 5’ TGTGGGCTGCCTGCTGAATACAC and 5’ AGCAAGAGTGAACTGCTGCCTCCT; and 30c2–P2 primers: 5’ TTTGGAAGGGGTGGGCAGTGAAGA and 5’ ACTGACATGGTCATCCCACTTGC.
Xenograft tumours were harvested from mice with 1800 units collagenase III (Worthington) and 100 Kunitz units DNase I (Sigma) as described25. Dissociated tumour cells were stained with CD44-APC, H2Kd-biotin and strepavidin-PE-cy7 (BD), and DAPI in HBSS/2% FBS for further flow analysis or sorting on FacsAriaII (BD).
Second-generation packaging vectors (VPR and VSVG) 25 were used to produce miRNA precursor lentiviruses, and third-generation packaging vectors (pRSV-rev, VSVG, and pMDLg/RRE) for other lentiviruses. Viable CD44+ breast cancer cells were isolated from TN1 xenograft tumours25 and transduced with lentiviruses at 10 or 20 multiplicities of infection (MOI 10 or 20).
Tumour cells were mixed 1:1 with Matrigel (BD) and transplanted into NOD/SCID mouse mammary fad pads (both left 4th and right 4th mammary fat pads)25, with a cohort of 10–20 mice per sample. Doxorubicin was injected intraperitoneally (IP) at 1mg/kg of mouse weight on day 20 and day 39 post implantation. Tumour growth was measured weekly by bioluminescence imaging.
As described25, bioluminescence images were acquired after D-luciferin (Biosynth AG) injection to mice (IP, 100 µl at 30 mg/ml per mouse) or incubation (300 µg/ml) with dissected lungs and tissues, using the IVIS Spectrum® system (Caliper Life Sciences). Data was analysed using LivingImage 4.0 Software (Caliper Life Sciences) and expressed as total photon flux.
If not specified otherwise, a student T test was used to evaluate significance, p values and standard deviations (SD). For animal studies, tumour growth curves were analysed using a linear mixed model in R software37 with the tumour volume (bioluminescence signals) or its fold change as a response variable. Lung metastases were analysed using Wilcoxon rank sum test. Log transformation was taken on both tumour volume and its fold change. For predictors, time, group indicator (vector control versus miRNA overexpression, treated versus untreated), and their interactions were considered initially; then, non-significant ones were dropped later and only significant variables were kept in the final models. As two tumours were planted and observed from each mouse, mouse ID (ear tag number) was included in the model as random effect. To allow autocorrelation among the tumour size measurements from the same mouse, AR (1) (autoregressive model of order 1) was added to the error term.
We thank Drs. Marcus Peter, Suzanne Conzen, and Yves Lussier for suggestions and comments on the project and manuscript. We appreciate the experimental support of several core facilities, including animal facility, optical imaging core facility, flow cytometry facility, integrated microscopy core facility, DNA sequencing facility, functional genomics facility, and IHC core facility at the University of Chicago. We specifically acknowledge Ryan Duggan, James Cao, David Leclerc, Marianne Greene, Terri Li, Xin Jiang, Shirley Bond, Jaejung Kim, Hui Zheng, Andrew Gusev, Dalong Qian, Yohei Shimono, and Ravand Samaeekia for technical help and support.
This study was supported in part by The University of Chicago Women’s Board (J.B.), National Institutes of Health (NIH) T90 Regenerative Medicine Training Program DK070103-05, Department of Defense Breast Cancer Research Program W81XWH-09-1-0331, Paul Calabresi K12 Award 1K12CA139160-02, NCI K99 CA160638-01, Chicago Fellows Program at the University of Chicago, and the University of Chicago Clinical and Translational Science Award (UL1 RR024999) (H.L.), The University of Chicago Cancer Research Center Pilot Fund, BSD Imaging Research Institute Pilot Research Projects Using Animal Imaging, UCMC/Northshore Collaborative Funds, a Segal Grant, and the Virginia and D. K. Ludwig Fund (G.L.G. and H.L.), funds from the Sociedad Española de Oncología Médica (SEOM) (A.P.), the Breast SPORE at University of North Carolina 5-P50-CA58223-17 (A.P. and C.M.P.), NIH R21 CA139278 and Pharmacogenetics of Anticancer Agents Research Group UO1GM61393 (M.E.D), NIH Grants U54 CA 126524 and P01 CA139490 (M.F.C.), the Breast Cancer Research Foundation (M.F.C., O.I.O., and C.M.P.), the Breast SPORE P50CA125183-05, the Doris Duke Charitable Foundation (O.I.O and C.N.), and the University of Chicago Cancer Center Support Grant CA 014599.
AUTHOR CONTRIBUTIONSJ.B., R.D., C.N., S.H. contributed equally to this manuscript. J.B., R.D., C.N., S.H., A.P., K.Y., Y-F.C., Y.W., K.E.S., T.Q., and H.L. designed and performed experiments and analysed data. J.L. provided the miRNA entry clones and the gateway vector backbone. S.Y.P. and D. H. performed biostatistical analyses for animal work and association studies. M.E.D., C.M.P., O.I.O., M.F.C., and G.L.G. designed the project. J.B. and H.L. wrote the manuscript. G.L.G., M.F.C., O.I.O, A.P., D.H., S.H., C.N., and M.E.D. contributed to the editing of the manuscript. G.L.G., M.F.C., O.I.O., and H.L. are equal corresponding authors on this manuscript.