Nanoliposomal Delivery of MicroRNA-203 Suppresses Migration of Triple-Negative Breast Cancer through Distinct Target Suppression

Triple-negative breast cancers affect thousands of women in the United States and disproportionately drive mortality from breast cancer. MicroRNAs are small, non-coding RNAs that negatively regulate gene expression post-transcriptionally by inhibiting target mRNA translation or by promoting mRNA degradation. We have identified that miRNA-203, silenced by epithelial–mesenchymal transition (EMT), is a tumor suppressor and can promote differentiation of breast cancer stem cells. In this study, we tested the ability of liposomal delivery of miR-203 to reverse aspects of breast cancer pathogenesis using breast cancer and EMT cell lines. We show that translationally relevant methods for increasing miR-203 abundance within a target tissue affects cellular properties associated with cancer progression. While stable miR-203 expression suppresses LASP1 and survivin, nanoliposomal delivery suppresses BMI1, indicating that suppression of distinct mRNA target profiles can lead to loss of cancer cell migration.


Introduction
Breast cancer is the most commonly diagnosed cancer and the second-largest cause of cancer-related deaths in women in the United States [1]. A majority of breast cancers are carcinomas, and their histological stratification is based primarily on the expression of estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2). Triple negative breast cancers (TNBCs) are defined through their lack of these receptors and are more difficult to manage clinically owing to the lack of protein targets, in contrast to tumors with hormone receptor or HER2 positive breast cancer, which are conventionally targeted with hormone inhibitors such as estrogen-response modulators and aromatase inhibitors or HER2-blocking antibodies, respectively [2]. Most breast cancer patients do not succumb to their primary tumor, but instead to metastases that become apparent after the primary lesion has been removed. For cells to contribute to metastases they must leave the primary site, enter the vasculature, survive in the blood, and then extravasate, and colonize secondary sites. Despite metastasis being a complex, multistep process, clinical and experimental evidence supports a role for epithelial-mesenchymal transition (EMT) in driving metastasis [3][4][5][6].
EMT is an intricate series of molecular and morphological changes that allows epithelial cells to decrease their inter-cellular adhesion and apical-basal polarity, and acquire a mesenchymal morphology and motile phenotype [7]. EMT is known to contribute to development during formation of the mesoderm and drive epithelial wound healing throughout life. When aberrantly activated under pathological conditions, carcinoma cells exploit shuttle protein involved in migration, adhesion, proliferation, and cell cycle progression in a variety of cancers [37]. LASP1 was originally identified from cDNA libraries of metastatic breast cancer, and its corresponding protein is overexpressed in more than 50% of breast cancers [38]. Additional characterized targets include ATM kinase [39], TWIST1 [40], fibroblast growth factor 2 (FGF2) [41], and receptor of Wnt signaling pathway Frizzled (FZD)-2 [42].
Developments in engineered nanomaterials, such as nanocrystalline drugs or nanoenabled drug carriers, are relevant to cancer in the fields of drug delivery, diagnosis, imaging, synthetic vaccine development, and miniature medical devices [43]. Key features including improved circulation and minimal systemic toxicity have shown promising applications for clinical development [43]. Common moieties include liposomes, polymeric micelles, and dendrimers. Based on each unique nanomaterial design, these nanoparticles offer extensive possibilities for cancer treatment.
In addition to the type of nanomaterials, the pharmacokinetics and biological distribution of nanoparticles are also related to the size, shape, and surface properties of nanomaterials [44]. Liposomes are closed spherical vesicles consisting of a lipid bilayer that encapsulates an aqueous phase in which drugs can be isolated, stored, and delivered [45]. The size of spherical lipid vesicles can range from a few nanometers to several micrometers, and nanoliposomes applied to medical use generally range between 50 and 450 nm [46]. Nanoliposomes are considered as an ideal drug delivery system because of their excellent ability to encapsulate many kinds of drugs similar to cell membranes.
Nanoliposomes can be preferentially accumulated in tumors, improve the efficiency of anticancer drugs, and reduce their systemic side effects, depending on the enhanced permeability and retention effect (EPR) [47]. Through the design of very small liposomes (<100 nm), tumor types with EPR can be utilized to realize the organization of the passive intake, thereby increasing drugs in the accumulation of specific targets [48]. Passive diffusion takes advantage of the EPR effect while also reducing toxicity, including inflammatory response, caused by reduced blood circulation to off-target areas of the body [49]. Liposomes that encapsulate microRNAs can provide solutions for major problems in small RNAs' delivery, such as poor cellular uptake and facile enzymatic degradation.
In this study, we confirmed previously established effects of miR-203 constitutive overexpression and extended our analyses to include nanoliposome-mediated delivery. We assayed for effects on EMT status by analyzing migration and sphere formation as well as target transcript and protein expression in multiple cell lines. We find that both endogenous expression and exogenous delivery of a miR-203 synthetic mimic are sufficient to suppress cellular migration, independent of effects on EMT state, yet acting through distinct target mRNAs.

Overexpression of miR-203 through Nanoliposomal Encapsulation
It is well-established that epithelial-mesenchymal transition suppresses miR-203 expression. Likewise, the ability of restored miR-203 expression to suppress migration and sphere formation, two cellular phenotypes that correspond to cancer progression, has been established [31]. Yet, these studies have relied on constitutive overexpression of miR-203, and thus do not mimic what could be achievable in a clinical setting. To determine if more clinically relevant methodologies of microRNA-based therapy are also useful in facilitating the ability of miR-203 to impact cancer cells, we compared stable overexpression to nanoliposomal delivery. Our hypothesis being that miR-203 delivery suppresses migration and sphere formation through target gene downregulation irrespective of the delivery method. These experiments were conducted in two triple-negative breast cancer cell lines, MDA-MB-231 and Hs578t cells, and in our model of experimental EMT, HMLE-Twist cells ( Figure 1A). Constitutive overexpression of miR-203 was accomplished using retrovirus transduction. Nanoliposomes were synthesized as previously described [48] and loaded with synthetic microRNA mimics of either a non-targeted scramble or miR-203. Through transmission electron microscopy and physicochemical characterization, we characterized that the liposomes encapsulated 96% of the RNA and their average particle size was 23.8 nm ( Figure S1).
First, we determined whether the overexpression methods resulted in an increase in miR-203 expression. To do this, we conducted whole RNA extraction on replicates of all cell lines and performed miRNA-specific qRT-PCR to investigate the relative quantity of miR-203. As reference standards, HMLE and MCF7 cells were used as they have high endogenous expression of miR-203 ( Figure 1A). miR-203 was found to be successfully overexpressed in all three cell lines ( Figure 1B,C). To determine the duration of miR-203 transferred using nanoliposomes, we treated cells and then extracted RNA every two days post-treatment. The amount of miR-203 present in cells decreased over 50% between 2 and 4 days post-exposure and was indistinguishable from control cells by 6 days post-exposure ( Figure 1D). (C) miR-203 expression in indicated cell lines exposed to microRNA mimic-loaded nanoliposomes at a concentration of 85 µM for 48 h. Data are presented as the mean and SEM of three independent replicates, two-tailed Student's t-test. (D) miR-203 expression in indicated cell lines for the indicated number of days post-exposure to microRNA-203 mimic-loaded nanoliposomes. Expression is presented relative to control cells exposed to scramble-loaded nanoliposomes. Small nucleolar RNA U6 was used as a normalization control. Data are presented as the mean and SEM of three independent replicates, one-way analysis of variance (ANOVA) with Dunnet's correction for multiple hypothesis testing. * p < 0.05, ** p < 0.01, *** p < 0.001. First, we determined whether the overexpression methods resulted in an increase in miR-203 expression. To do this, we conducted whole RNA extraction on replicates of all cell lines and performed miRNA-specific qRT-PCR to investigate the relative quantity of miR-203. As reference standards, HMLE and MCF7 cells were used as they have high endogenous expression of miR-203 ( Figure 1A). miR-203 was found to be successfully overexpressed in all three cell lines ( Figure 1B,C). To determine the duration of miR-203 transferred using nanoliposomes, we treated cells and then extracted RNA every two days post-treatment. The amount of miR-203 present in cells decreased over 50% between 2 and 4 days post-exposure and was indistinguishable from control cells by 6 days post-exposure ( Figure 1D).

Overexpression and Delivery of miR-203 Suppress Migration
The cell migration assay is a measure of cellular motility, which is enhanced in triplenegative breast cancer cells and following an EMT. We next evaluated whether miR-203 overexpression through either stable overexpression or exogenous delivery was sufficient to suppress the migratory capacity of target cells and whether the methodology of overexpression impacted the degree of suppression. To allow time for microRNA to alter cellular properties through target downregulation, cells were treated with nanoliposomes 48 h prior to plating for the migration assay. At specific time intervals, the distance between the two edges was measured at multiple positions. Stable overexpression of miR-203 was sufficient to suppress migration of MDA-MB-231 and Hs578t cells, consistent with our prior observation of the triple-negative breast cancer cell line SUM159 [31] (Figure 2A). Nanoliposomal delivery of miR-203 into HMLE Twist cells was also sufficient to suppress migration despite constitutive expression of the EMT-TF Twist ( Figure 2B). In general, MDA-MB-231 cells were the most responsive to both methods of overexpressing miR-203. Our data suggest that restoring miR-203 expression levels inhibits migration in triple-negative breast cancer. SEM of three independent replicates, one-way analysis of variance (ANOVA) with Dunnet's correction for multiple hypothesis testing. * p < 0.05, ** p < 0.01, *** p < 0.001.

Overexpression and Delivery of miR-203 Suppress Migration
The cell migration assay is a measure of cellular motility, which is enhanced in triplenegative breast cancer cells and following an EMT. We next evaluated whether miR-203 overexpression through either stable overexpression or exogenous delivery was sufficient to suppress the migratory capacity of target cells and whether the methodology of overexpression impacted the degree of suppression. To allow time for microRNA to alter cellular properties through target downregulation, cells were treated with nanoliposomes 48 h prior to plating for the migration assay. At specific time intervals, the distance between the two edges was measured at multiple positions. Stable overexpression of miR-203 was sufficient to suppress migration of MDA-MB-231 and Hs578t cells, consistent with our prior observation of the triple-negative breast cancer cell line SUM159 [31] (Figure 2A). Nanoliposomal delivery of miR-203 into HMLE Twist cells was also sufficient to suppress migration despite constitutive expression of the EMT-TF Twist ( Figure 2B). In general, MDA-MB-231 cells were the most responsive to both methods of overexpressing miR-203. Our data suggest that restoring miR-203 expression levels inhibits migration in triple-negative breast cancer. Represented as mean ± standard deviation of change in gap length from t = 0. pvalue calculated for final measurement using Student's two-tailed t-test. * p < 0.05, ** p < 0.01, *** p < 0.001. Represented as mean ± standard deviation of change in gap length from t = 0. p-value calculated for final measurement using Student's two-tailed t-test. * p < 0.05, ** p < 0.01, *** p < 0.001.

Delivery of miR-203 Suppresses Sphere Formation in HMLE Twist Cells
Differentiated cells require anchoring to the extracellular matrix in order to proliferate. However, stem-like cells can divide and grow into spheroidal clusters of cells even under low-attachment conditions [50]. We have previously observed that constitutive overexpression of miR-203 was sufficient to suppress sphere formation, an indicator of stem-like properties, in SUM159 and HMLE Twist cells [31]. To test whether distinct methods of overexpressing miR-203 are sufficient to suppress stemness, we measured mammosphere formation. As expected, HMLE Twist cells were responsive to miR-203 constitutive expression; however, breast cancer cell lines were not responsive to miR-203 constitutive expression ( Figure 3). HMLE Twist cells, nevertheless, showed a modest yet significant loss in stemness-associated sphere formation in response to nanoliposomal miR-203 ( Figure 3). Thus, miR-203 delivery was sufficient to affect sphere formation, an in vitro measure of stemness, in an experimental model of EMT.
Non-coding RNA 2021, 7, 45 6 of 17 methods of overexpressing miR-203 are sufficient to suppress stemness, we measured mammosphere formation. As expected, HMLE Twist cells were responsive to miR-203 constitutive expression; however, breast cancer cell lines were not responsive to miR-203 constitutive expression ( Figure 3). HMLE Twist cells, nevertheless, showed a modest yet significant loss in stemness-associated sphere formation in response to nanoliposomal miR-203 ( Figure 3). Thus, miR-203 delivery was sufficient to affect sphere formation, an in vitro measure of stemness, in an experimental model of EMT. Sphere formation in indicated cell lines exposed to microRNA mimic-loaded nanoliposomes at a concentration of 85 μM for 48 h prior to plating for sphere formation. Biological replicates, mean and standard deviation are represented. pvalue calculated for final measurement using Student's two-tailed t-test. ns = not significant; * p < 0.05, **** p < 0.0001.
We noted changes in the expression of BMI1, LASP1, BIRC5 (survivin), and CDH2 in the Nanostring analysis. In order to quantitatively determine the effect of miR-203 over- Sphere formation in indicated cell lines exposed to microRNA mimic-loaded nanoliposomes at a concentration of 85 µM for 48 h prior to plating for sphere formation. Biological replicates, mean and standard deviation are represented. p-value calculated for final measurement using Student's two-tailed t-test. ns = not significant; * p < 0.05, **** p < 0.0001.
We noted changes in the expression of BMI1, LASP1, BIRC5 (survivin), and CDH2 in the Nanostring analysis. In order to quantitatively determine the effect of miR-203 overexpression on individual gene expression, we assessed expression of these genes using qRT-PCR. While most genes failed to show statistically significant differential expression, expression of LASP1 was consistently repressed by constitutive overexpression of miR-203 (Figure 4b), but not by nanoliposomal delivery (Figure 4c). However, nanoliposomal delivery of miR-203 resulted in suppression of BMI1 and CDH2, but only in MDA-MB-231 cells.

Overexpression of miR-203 Suppresses LASP1 and Survivin (BIRC5) Protein Expression, and Delivery of miR-203 Suppresses BMI1 Protein Expression
microRNAs can function through destabilization of their mRNA targets or suppression of protein translation. Therefore, changes in miR-203 target or EMT marker expression may or may not be evident at the transcript level. To determine if key targets were suppressed at the level of protein expression, we conducted western blot analysis.
As expected, constitutive miR-203 overexpression resulted in modest suppression of N-cadherin and vimentin but failed to induce upregulation of the epithelial marker, Ecadherin ( Figure 5). This result is consistent with data from constitutive expression in SUM159 cells [31]. We next examined expression of select target proteins. Expression of LASP1, but not survivin or BMI1, was suppressed in all cell lines with constitutive miR-203 expression ( Figure 5B,C) consistent with qRT-PCR data (Figure 4b). Remarkably,  To determine if key targets were suppressed at the level of protein expression, we conducted western blot analysis.
As expected, constitutive miR-203 overexpression resulted in modest suppression of N-cadherin and vimentin but failed to induce upregulation of the epithelial marker, E-cadherin ( Figure 5). This result is consistent with data from constitutive expression in SUM159 cells [31]. We next examined expression of select target proteins. Expression of LASP1, but not survivin or BMI1, was suppressed in all cell lines with constitutive miR-203 expression ( Figure 5B,C) consistent with qRT-PCR data (Figure 4b). Remarkably, delivery of miR-203 by nanoliposome resulted in no change in expression of LASP1 or survivin, miR-203 targets, nor in N-cadherin or vimentin, EMT markers, but did result in a modest suppression of BMI1 which was also suppressed at the mRNA level-but only in MDA-MB-231 cells. This was surprising as miR-203 delivery by nanoliposome was sufficient to suppress migration in all cell lines tested and to suppress sphere formation in HMLE Twist cells. We thus conclude that miR-203 is capable of suppressing migration either through suppression of LASP1 and survivin or through suppression of BMI1.
Non-coding RNA 2021, 7, 45 9 of 17 delivery of miR-203 by nanoliposome resulted in no change in expression of LASP1 or survivin, miR-203 targets, nor in N-cadherin or vimentin, EMT markers, but did result in a modest suppression of BMI1 which was also suppressed at the mRNA level--but only in MDA-MB-231 cells. This was surprising as miR-203 delivery by nanoliposome was sufficient to suppress migration in all cell lines tested and to suppress sphere formation in HMLE Twist cells. We thus conclude that miR-203 is capable of suppressing migration either through suppression of LASP1 and survivin or through suppression of BMI1.  p-value calculated using Student's two-tailed t-test. * p < 0.05, ** p < 0.01.

Loss of MIR203 Copy Number in TNBC
A growing number of studies demonstrate genetic gain or loss of specific miRNAs in cancer [58]. To determine if miR-203 is gained or lost in specific breast cancer subtypes, we analyzed copy number data collected by The Cancer Genome Atlas database of breast carcinoma patients [59] and available using the UCSC Xena platform [60]. Consistent with the demonstrated lower expression in EMT-positive triple-negative breast cancer cell lines, patient data show a statistically significant decrease in copy number for the MIR203 gene in such cancers whether categorized based on intrinsic subtype, as in basal-like breast cancers, or based on histological analysis of receptor expression, as in triple-negative breast cancers ( Figure 6).

Loss of MIR203 Copy Number in TNBC
A growing number of studies demonstrate genetic gain or loss of specific miRNAs in cancer [58]. To determine if miR-203 is gained or lost in specific breast cancer subtypes, we analyzed copy number data collected by The Cancer Genome Atlas database of breast carcinoma patients [59] and available using the UCSC Xena platform [60]. Consistent with the demonstrated lower expression in EMT-positive triple-negative breast cancer cell lines, patient data show a statistically significant decrease in copy number for the MIR203 gene in such cancers whether categorized based on intrinsic subtype, as in basal-like breast cancers, or based on histological analysis of receptor expression, as in triple-negative breast cancers ( Figure 6). In summary, expression miR-203 is demonstrated to impact cellular phenotypes associated with cancer disease progression including cellular migration and non-adherent cell growth. Moreover, the transient delivery of exogenous microRNA through nanoliposomal encapsulation was capable of mimicking some, though not all, of the effects of constitutive expression.

Discussion
miRNAs have many advantages as a therapeutic approach. Indeed, the pharmacological potential of miRNA mimics or inhibitors, demonstrated in animal disease models, suggests that miRNAs are viable targets or candidates for therapy [61,62]. Mature miRNA sequences are short and are usually completely conserved in multiple vertebrate species. These properties make miRNA relatively easy to target therapeutically and allow for the use of the same miRNA sequences in preclinical efficacy and safety studies as well as in clinical trials. In addition, miRNAs typically have many targets in cellular networks, which in turn can regulate entire pathways in disease states by targeting disease-associated mRNAs. A therapeutic strategy for restoring miRNA activity is to use synthetic double-stranded RNA that contains chemical modifications to improve stability and cellular uptake [63]. Major limitations associated with miRNA delivery are susceptibility to degradation by nucleases, rapid clearance from blood, immunotoxicity, and low tissue permeability [64,65]. miRNA mimics with liposome nanoparticles have been injected In summary, expression miR-203 is demonstrated to impact cellular phenotypes associated with cancer disease progression including cellular migration and non-adherent cell growth. Moreover, the transient delivery of exogenous microRNA through nanoliposomal encapsulation was capable of mimicking some, though not all, of the effects of constitutive expression.

Discussion
miRNAs have many advantages as a therapeutic approach. Indeed, the pharmacological potential of miRNA mimics or inhibitors, demonstrated in animal disease models, suggests that miRNAs are viable targets or candidates for therapy [61,62]. Mature miRNA sequences are short and are usually completely conserved in multiple vertebrate species. These properties make miRNA relatively easy to target therapeutically and allow for the use of the same miRNA sequences in preclinical efficacy and safety studies as well as in clinical trials. In addition, miRNAs typically have many targets in cellular networks, which in turn can regulate entire pathways in disease states by targeting disease-associated mRNAs. A therapeutic strategy for restoring miRNA activity is to use synthetic double-stranded RNA that contains chemical modifications to improve stability and cellular uptake [63]. Major limitations associated with miRNA delivery are susceptibility to degradation by nucleases, rapid clearance from blood, immunotoxicity, and low tissue permeability [64,65]. miRNA mimics with liposome nanoparticles have been injected intravenously and intratumorally [66] in order to restore the function of various tumor-inhibiting miRNAs, such as miR-34a, in mouse cancer models [67]. However, unformulated miRNA mimics degrade rapidly in vivo. As such, an optimized liposome-encapsulated miRNA mimic has been shown to promote endocytosis and protect structures from degradation [68]. Neutral DOPC (1, 2-dioleoyl-tin-glycerol-3-phosphatidylcholine) liposomes were able to deliver miRNA-506 mimics or miRNA-520 in mouse models of ovarian cancer, resulting in significant tumor suppression [69]. Moreover, the recently demonstrated safety of the lipid nanoparticles used to deliver mRNA for vaccination highlights the potential for RNAbased therapeutics [70]. Taken together, these studies suggest that the programmability of small non-coding RNA nanostructures offers great promise for further exploration of miRNA-based therapies.
Several studies have reported a role of miR-203 as a tumor or metastasis suppressor in a variety of cancers including lung cancer [71] and breast cancer [72]. In recent years, a large number of studies have shown that promoter hypermethylation leads to down-regulation of miR-203 in several cancer types. Diao et al. reported that miR-203 hypermethylation was a common event in endometrial cancer and was closely related to microsatellite instability [26]. In addition, Huang et al. found that miR-203 was often downregulated by promoter hypermethylation in rhabdomyosarcoma cell lines and clinical biopsies, and the re-expression of miR-203 through DNA demethylation drugs inhibited the migration and proliferation of cancer cells and promoted final muscle-derived differentiation [73].
We tested the ability of re-introduced miR-203 to affect cellular properties associated with EMT and stemness and to affect gene expression. Our data indicate a consistent effect on the migration of target cells despite little effect on the EMT status. Nevertheless, not all cell lines respond identically to miR-203 restoration. In particular, miR-203 delivery via nanoliposome was sufficient to reduce sphere formation, a measure of stemness, in HMLE Twist cells, but not either breast cancer cell line. Cancer cell lines often display redundant pathway activation, which may contribute to the failure of these cells to respond. Nevertheless, SUM159 cells, another TNBC cell line, do exhibit diminished stemness features in response to stable miR-203 expression [31]. Nanoliposomal delivery of miR-203 significantly impacts cellular migration, yet does not suppress LASP1, which is consistently suppressed by stable expression of miR-203. The most consistent molecular change associated with nanoliposomal miR-203 is suppression of BMI1 protein, in MDA-MB-231 and Hs578t cells, and BMI1 mRNA, in MDA-MB-231 cells. BMI1, a member of the PRC2 complex that regulates chromatin modification, has been linked to EMT, migration, and metastasis [74,75]. Defining the relationship between cell type and molecular response to the recovery of miR-203 will help reveal the specific mechanisms of miR-203 action. This work supports further research on the use of miR-203 or small molecules sufficient to relieve miR-203 repression as a therapeutic for triple-negative breast cancer.

RNA Isolation and Quantitative Reverse-Transcription PCR
Total RNA in cells was isolated using Trizol ® Reagent (Thermo Scientific, Waltham, MA, USA) according to the manufacturer's protocol. The amount and purity of isolation RNA were analyzed by the Nanodrop spectrophotometer. Complementary DNA was synthesized using Reverse Transcriptase. qPCR was performed as quadruplicate using Taqman Master mix and miRNA-specific primers (Applied Biosystems, Foster City, CA, USA; Thermo Scientific). Thermal cycle conditions used for all qPCR reactions were as follows: 10 min at 95 • C, followed by 40 cycles consisting between denaturation for 15 s at 95 • C and annealing for 60 s at 60 • C. PCR reactions were concluded with incubation for 10 min at 72 • C to compete the extension of all synthesized products QuantStudio 5 (Applied Biosystems). The expression of miRNAs was normalized against that of U6, and relative quantification values were calculated using the 2-∆∆Ct method.

Nanostring Analysis
Total RNA isolated as above was subjected to analysis using a PlexSet probe set. A total of 450 ng of RNA in TE buffer was hybridized to probes for AKT2, BIRC5, BMI1, CAV1,  CDH1, CDH2, CLDN7, DNMT3B, FGF2, FN1, FOXC2, FZD2, LASP1, NUAK1, RAB22A,  SNAI1, SNAI2, SOCS3, SPARC, YES1, ZEB1, GAPDH, HPRT1, and PKG1 provided by Nanostring (Nanostring Technologies, Seattle WA, USA). Hybridization was conducted on 96 RNA samples, using a multiplexing arrangement, for 16 h at 65 • according to the manufacturer's protocol. Raw counts were normalized to the geometric mean of the count values for the three endogenous control genes, GAPDH, HPRT1, and PKG1. Expression across samples was computed as a z-score using the formula Z = (x − µ)/σ where x = the observed value, µ = the mean across all samples, and σ = the standard deviation across all samples. Hierarchical clustering is based on Euclidean distance (nSolver 4.0).

Immunoblotting
Cells were lysed in ice-cold radio-immunoprecipitation (RIPA) buffer containing protease inhibitor and phosphatase inhibitor cocktail (Alfa Aesar, Stoughton, MA, USA). Protein was quantified using the Bradford Assay (BioRad, Hercules, CA, USA). Proteins (20-30 µg) were resolved in SDS-polyacrylamide gels (10%) and transferred to PVDF membranes using a Tris-glycine buffer system. Membranes were blocked with 5% non-fat dry milk in 0.1% Tween20 in TBS (TBST) for 1 h at room temperature followed by probing with primary antibodies and gently rocking overnight at 4 • C. Antibodies used for immunoblotting were Mouse anti-E-cadherin (Cell Signaling, Danvers, MA, USA), Rabbit anti-N-cadherin (Cell Signaling), Rabbit anti-vimentin (Protein Technologies, Tucson, AZ, USA), Rabbit anti-survivin (Cell Signaling), Rabbit anti-Lasp1 (Cell Signaling), Rabbit anti-BMI1 (Cell Signaling), and Mouse anti-β-actin (BD Biosciences, San Jose, CA, USA). The next day, the blots were washed with TBST three times, 5 min each time. They were incubated with secondary antibodies (1:2000) at room temperature for 1 h. Chemiluminescent signals were detected with ECL™ prime (Thermo Fisher Scientific, Waltham, MA, USA) using the Biorad ChemiDoc system. If necessary, blots were stripped with ECL Stripping Buffer (Li-Cor, Lincoln, NB, USA) following the manufacturer's protocol. Bands were quantified using ImageJ.

Cell Migration Assay
To perform the cell migration assay, we first drew two parallel horizonal lines upon the bottom of a 24-well plate before seeding cells into it. Then, cells were added to new media up to 0.5 mL/well in a 24-well plate, at a final seeding density as 5 × 10 4 cells per well. The next day, the media was replenished gently, and the well was scratched with a straight line on the cell monolayer across the center of each well with a 200 µL micropipette tip. At regular intervals, we imaged the gaps of each well. Cell migration rates were determined by measuring the distance between the cell frontiers after a specified culture hour. At a given point in time, the distance between the two edges at multiple points was quantified using GraphPad Prism 6.0. The change in gap length is plotted over time.

Mammosphere Assay
Cells were harvested according to standard protocol and resuspended in serumfree mammary epithelial growth medium (MEGM) supplemented with 20 ng/mL bFGF, 10 ng/mL EGF, 4 ug/mL heparin, and 1% Methylcellulose. Then, 1000 cells per well were plated in a 96-well ultra-low attachment plate (Corning, Corning, NY, USA) in 100 µL mammosphere media. Every third day, 100 µL of fresh mammosphere media was added into each well. Spheres formed around 10-14 days. Wells were imaged using 10× magnification on a computer-assisted phase contrast microscope (Nikon, Tokyo, Japan). Rounded spheroids greater than 50 µm were considered as mammospheres.

Patient Analysis
To evaluate the copy number of MIR203 in breast cancer, we acquired the genomic data for 1247 breast tumor samples in TCGA by downloading the log2(tumor/normal) values for copy number, calculated after removal of germline copy number variation. We distinguished the tumors into triple-negative or others using the clinical annotations provided by the TCGA. We compared the copy number levels between groupings within Graphpad Prism 6.0 using a one-way ANOVA accompanied by pair-wise comparisons modified by Dunnet's correction for multiple hypothesis testing.

Statistical Analysis
Data were analyzed using Prism 6 (GraphPad) software. Statistical significance is indicated as * p < 0.05, ** p < 0.01, *** p < 0.001, or ns (not significant), using Student's unpaired two-tailed t-test. Data were normalized in some cases. Either single values per cell line or mean ± standard division are plotted throughout the manuscript. All experiments were performed at least three independent times.