miR-34a Silences c-SRC to Attenuate Tumor Growth in Triple-Negative Breast Cancer

Triple-negative breast cancer (TNBC) comprises approximately 20% of breast cancers and accounts for a disproportionate share of morbidity because of its aggressive behavior, increased incidence in younger women, and lack of effective targeted therapies (1–4). This breast cancer subset is defined by what it lacks: estrogen receptor, progesterone receptor, and HER2 amplification, thereby limiting systemic treatment options for these patients to conventional chemotherapy. Transcriptional profiling studies initially suggested TNBC represented a distinct molecular entity that was different from ER-positive luminal-A type cancers (5). It has recently become apparent that TNBC itself is a molecularly heterogeneous disease and is often subdivided into smaller categories including mesenchymal/mesenchymal stem cell–like (M/MSL), luminal androgen receptor type (LAR), and basal-like (BL) with or without an extensive immunomodulatory component (6–8). Despite identifying the molecular pathways dysregulated in TNBC, this extensive transcriptional and genomic heterogeneity has limited the development of broad therapeutic strategies for treating patients with TNBC. Therefore, discovery of a single therapeutic molecule that can function to suppress the multiple pathways supporting TNBC growth would be an important clinical breakthrough.

As miRNAs target multiple pathways and function as signal modulators by fine-tuning gene expression to mediate tumor-suppressive phenotypes, these molecules represent attractive therapeutic options and targets. miRNAs represent a class of small noncoding RNAs that function as global regulators of gene expression and are important master controllers of cell survival and proliferation pathways critically important in cancer development and tumor maintenance (9). These, approximately 22 nucleotide, noncoding RNAs control the expression of protein-coding genes via imperfect binding to the 3′ untranslated region (3′UTR) of the target mRNA, leading to either translational inhibition or degradation (10). Some studies estimate that miRNAs regulate ≥60% of the human protein-coding transcriptome (11). Importantly, single miRNAs can regulate entire cell signaling networks in a cell context–dependent manner (12). As miRNAs can function as either oncogenes or tumor suppressors, and can control the robustness of multiple cell signaling pathways (13, 14), miRNAs that become dysregulated could promote a disease state, such as cancer. Our goal was to identify the key miRNAs dysregulated in TNBC in an effort to elucidate the disease and potentially point to novel miRNA-based therapeutics. We found miR-34a was selectively lost in TNBC cells and harbored tumor-suppressive functions when reintroduced into TNBC cells, with the most robust cell killing occurring in the mesenchymal TNBC subtypes.

While there have been historical challenges associated with the use of RNA in therapy, chemical modifications to the backbone of synthetic RNA molecules in combination with advances in nanoparticle-based delivery methodologies have resulted in significantly increased half-lives in vivo and better therapeutic potential. Lipid- and polymer-based nanoparticle systems can now be used to effectively deliver RNA molecules into tumor cells (15–18). Here we used a polyamine-co-ester terpolymer nanoparticle (tNP) technology, previously shown to actively target plasmid DNA to tumor cells (19), as a safe and effective method for delivering synthetic miR-34a into TNBC tumor cells in vivo. From the combination of our in vitro and in vivo studies, we found that miR-34a ameliorates multiple tumor phenotypes, and acts through the regulation of mRNA targets such as BCL2, NOTCH1, and a novel target SRC. Moreover, this miR-34a effect occurs specifically in the mesenchymal-TNBC subtype, and can sensitize these cells to therapeutic agents such as dasatinib and paclitaxel.

All breast cancer cell lines were purchased directly from ATCC in 2013 and maintained in DMEM supplemented with 10% FBS, 1% penicillin/streptomycin and cultured at 37°C in a 5% CO₂ incubator. All experiments were performed within 6 months to 1 year upon receipt of cells, and characterization by STR analysis was performed by ATCC. Normal breast conditionally reprogramed cell lines (CRC) were maintained without coculturing with feeder cells (20). HMECs (Life Technologies) were maintained as per manufacturer's instructions, while MCF-10A cells were cultured in a defined media (21). For RNA-based functional experiments, mirVana mimics and validated Silencer Select siRNAs (Life Technologies) were transfected into 5 × 10⁴ cells using DharmaFECT 1 (GE Healthcare Dharmacon), as per the manufacturer's instructions. For extended experimental procedures, see Supplementary Materials and Methods.

For qPCR, cells were lysed in TRIzol (Life Technologies), and total RNA was used as the input for subsequent RT-PCR reactions. For miRNA analysis, cDNA synthesis and qRT-PCR were performed according to the miScript II RT Kit (Qiagen) and miScript SYBR Green PCR Kit (Qiagen) protocols, respectively. For mRNA analysis, cDNA synthesis and qRT-PCR were performed according to the RT² First Strand Kit and RT² qPCR Primer assay protocols (Qiagen).

For the assessment of cytotoxicity and development of IC₅₀ curves, sulforhodamine blue (SRB) assays were performed as described previously (22), see Extended Experimental Procedures. For morphologic assessments, cells were stained with 0.2% crystal violet for 10 minutes at room temperature and washed with PBS. Images were acquired using a Zeiss dissection microscope (Zeiss), AxioCam MRc 5 camera (Zeiss) and AxioVision 4.7.1 imaging software (Zeiss). In some instances, stained cells were treated with 10% acetic acid with 0.1% SDS and dye abundance was quantified by measuring absorbance at 600 nm.

For invasion assays, following miRNA or siRNA transfection, cells were cultured in 0.1% serum-containing media overnight, and then seeded into BD BioCoat BD Matrigel Invasion Chambers. For soft agar assays, after miRNA or siRNA transfection, cells were mixed with agarose to a 0.3% concentration in RPMI growth media and cultured for 25 days. At the end of the study, inserts or cells were fixed, stained with crystal violet, and counted.

After miRNA or siRNA transfection, cells were seeded into 12-well plates. After 5 to 6 days, cells were assessed for cellular β-galactosidase activity, using a Senescence β-Galactosidase Staining Kit (SA-β-Gal; Cell Signaling Technology). Five randomly designated areas in each well were then scored for the proportion of cells containing a blue precipitate (cleaved X-gal) as compared with the total number of cells within the same field.

Five- to 6-week-old female CrTac:NCr-Foxn1^nu mice (Taconic) were maintained at Yale University (New Haven, CT) in accordance with Yale Animal Resource Center and Institutional Animal Care and Use Committee guidelines. For tumor formation, miRNA treatments, and tumor analysis see Extended Experimental Procedures.

For miRNA targeting assays, HEK-293T cells were transfected with a pLightSwitch-SRC-3′UTR plasmid or an pLightSwitch-Empty-3′UTR control (SwitchGear Genomics), in combination with miR-34a or miR-Scr control using DharmaFECT Duo (GE Health Dharmacon). After 24 hours, cells were lysed and luciferase activity was measured as per the manufacturer's instructions. Luciferase signal was normalized to the 0 nmol/L miRNA control condition for each UTR construct. Technical quadruplicates were performed and experiments were performed three times.

Unless otherwise stated, a Student t test was used for P value calculations. Values reported are expressed as the mean ± SD, whereas densitometric readings of Western blot analyses are reported as averages. When more than one condition was being compared, such as changes between various treatments and control groups, an ANOVA analysis was performed. Statistical significance was set at P < 0.05.

To identify miRNAs with potential therapeutic value in TNBC, we performed miRNA microarray analysis on >1,000 miRNAs from 19 breast cancer cell lines. Clustering analysis of the top 50 most variant miRNAs indicated the presence of a subset of miRNAs differentially expressed between basal- and luminal-A type cell lines (Supplementary Fig. S1A). qRT-PCR validation experiments were performed on a panel of 13 breast cancer and 9 normal breast cell lines to determine which miRNAs were differentially expressed as compared with normal cells. This included a set of CRC lines established from normal mammoplasty tissue (20). A number of highly expressed TNBC-specific miRNAs were identified, such as miR-221 and miR-155, when compared with luminal-A type cells; however, these miRNAs were also highly expressed in CRC lines (Supplementary Fig. S1B). Furthermore, nontumorigenic MCF-10A and HMEC control lines also harbored high levels of these miRNAs.

Instead, we focused on identifying miRNAs lost in TNBC with the aim of pointing us towards a possible miRNA reintroduction-based therapeutic agent. We identified miR-34a and miR-200 to be aberrantly downregulated in TNBC cell lines as compared with both luminal-A type and normal breast cell lines (Fig. 1A and Supplementary Fig. S1C). These findings support our previous work, where we identified miR-34a to be downregulated in TNBC breast cancer patient samples (23). Furthermore, miR-200 is a known regulator of metastatic processes, and is downregulated in TNBC (24, 25). Given miR-34a is specifically lost in TNBC cell lines as well as in patient samples, and that the therapeutic targets of miR-34a specific to TNBC have not yet been well defined, we selected miR-34a for further investigation.

To determine whether miR-34a harbored prognostic value in TNBC patients, Kaplan–Meier survival plots and Cox regressions were computed from TCGA. High expression of miR-34a in TNBC-patients conferred better overall survival (OS; n = 97, HR = 0.22, P = 0.041) than in those with low miR-34a, an effect not seen in the all-breast cancer patient cohort (Fig. 1B and C). Given the heterogeneity of TNBC, we investigated whether predicted targets of miR-34a were enriched in specific TNBC subtypes. We returned to the original breast cancer cell line dataset where U133A data were available for 22,283 genes across 19 cell lines. We identified a TNBC-overexpressed gene set (264 genes) that demonstrated enrichment for miR-34 target genes (Supplementary Fig. S1D). In addition, target enrichment of another TNBC-related miRNA, miR-200, was also identified. These miRNA targets were not enriched in the TNBC-underexpressed gene set (data not shown).

We then performed hypergeometric analyses of all putative miR-34a targets across the TNBC subtype-specific gene expression profiles reported in Lehmann and colleagues (6). A total of 3,000 predicted miR-34a targets, irrespective of site conservation, were acquired from TargetScan (26), and compared with the top 20% upregulated genes within all TNBCs (∼2,872 genes) and within each TNBC subtype (variable sample size). When the number of miR-34a–TNBC subtype-specific targets were compared with the total number of miR-34a–TNBC targets, hypergeometric analysis indicated an overrepresentation of miR-34a targets present in the MSL- and M-TNBC subtypes, whereas miR-34a targets in the LAR and immunomodulatory TNBC subtypes were substantially underrepresented (Fig. 1D). A stringent enrichment analysis on 760 high confidence miR-34a targets was also performed, which still resulted in significant overrepresentation occurring within the MSL-TNBC subtype (Supplementary Fig. S1E). These data support the notion that the genes important for growth and survival within the mesenchymal–TNBC cell types are highly regulated by miR-34a and that these cells suppress miR-34a to maintain a tumorigenic state.

To determine whether miR-34a promoted tumor-suppressive phenotypes in TNBC, we transfected a panel of breast cancer cell lines with 30 nmol/L miR-34a or scrambled control (miR-Scr), and assayed cell growth using SRB assays (22). BT-549 and MDA-MB-231 TNBC growth was inhibited by miR-34a reintroduction (Fig. 1E), while luminal-A MCF-7 and LAR-TNBC MDA-MB-453 cells were marginally impacted. This TNBC-specific growth inhibition by miR-34a was observed in an extended panel of TNBC lines, while normal CRCs and human fibroblasts (HFF) were not severely impacted (Supplementary Fig. S1F). Lower IC₅₀ values were also obtained in TNBC lines versus luminal-A or LAR-TNBC cell lines (Table 1).

We next investigated additional antitumorigenic phenotypes associated with miR-34a reintroduction. The earliest phenotypic consequence, observed 36 hours after miR-34a transfection, was abrogation of invasion. Using Matrigel-coated Boyden chamber assays, BT-549, MDA-MB-231, and MDA-MB-436 cells transfected with 5 nmol/L miR-34a resulted in approximately 2.1-, 4.0-, and 2.9-fold reductions in the number of invasive cells (Fig. 2A and Supplementary Fig. S2A). This phenotype was present prior to any antigrowth effects mediated by miR-34a, which occurred more than 84 hours after transfection (Fig. 1E), and was not a result of apoptosis (Fig. 2C). These data suggest that miR-34a regulates cellular targets in TNBC cells, important in focal adhesion and invasion, which are the same signaling pathways that define MSL- and M-TNBC subtypes.

We next tested whether miR-34a–transfected TNBC cells could grow in an anchorage-independent fashion. Specifically, BT-549, MDA-MB-436, and MDA-MB-231 cells transfected with 7.5 nmol/L miR-34a resulted in approximately 8.2-, 4.1-, and 3.6-fold reductions in colony formation, respectively, after 20 days of culture in 0.3% soft agar (Fig. 2B and Supplementary Fig. S2B). To determine whether this phenotype was due to a block at a particular stage in cell-cycle progression, we performed propidium iodide (PI) staining assays, which indicated miR-34a–transfected MDA-MB-231 cells resulted in a 5% increase of cells in G₂–M 4 to 6 days after transfection (Fig. 2D). Similar results were obtained with BrdUrd assays (Supplementary Fig. S2C). Despite this marginal G₂–M arrest, miR-34a–transfected TNBC cells underwent striking morphologic changes indicative of senescence—specifically an elongated and flattened cytoplasm (data not shown; ref. 27). Senescent cells can be a result of both G₁ and/or G₂ arrest (27), which may explain the lack of a stage-specific block in cell cycle upon miR-34a transfection. Indeed, SA-β-gal assays performed on miR-34a-transfected TNBC cells confirmed this senescence-promoting effect in a dose-dependent manner (Fig. 2E and Supplementary Fig. S2D and S2E). Conversely, MCF-7 cells, which harbor high endogenous miR-34a levels and higher baseline rates of senescence, had an approximately 2-fold reduction in SA-β-gal activity when treated with a miR-34a sponge (Supplementary Fig. S2F).

To determine whether miR-34a-induced antitumorigenicity was due to differential transfection efficiencies, we performed psiCheck2 luciferase reporter assays, where a miR-34a sensor sequence was inserted downstream of Renilla luciferase. Cotransfection experiments using wild-type or mutant sensors and miR-34a or miR-Scr mimics indicated that 500 pmol/L miR-34a was sufficient to reduce reporter activity by 5- to 10-fold across all lines tested (Supplementary Fig. S2G). Therefore, the antitumorigenic phenotype induced by 5 to 30 nmol/L of miR-34a in MSL- and M-TNBC cells likely reflects a heightened sensitivity to miR-34a reintroduction and not differential transfection efficiency. Overall, these data suggest that miR-34a reintroduction in TNBC shuts down oncogenic signaling pathways that affect invasion and proliferation, eventually resulting in cell death by cytostasis/senescence. Interestingly, some oncogenic mediators could include noncoding RNAs such as the miR-17/92 cluster, which is known to be elevated in TNBC (28–30) and can be downregulated by miR-34a reintroduction (Supplementary Fig. S2H).

Given the in vitro findings, we tested whether miR-34a could reduce tumor growth in vivo. MDA-MB-231 cells were mixed with Matrigel and injected into the flanks of CrTac:NCr-Foxn1^nu mice. Once tumors reached 150 mm³, mice were randomized into a miR-34a or miR-Scr treatment cohort. miRNA mimics were packaged into a neutral-lipid emulsion and injected intratumorally at a 25 μg/tumor dose. After 3 treatments, given every 4 days, miR-34a tumors did not proliferate beyond pretreatment tumor size, whereas miR-Scr tumors grew unperturbed (Fig. 3A, left). At the end of the study, tumors treated with miR-34a harbored an approximately 4.4-fold reduction in tumor weight, as compared with miR-Scr controls (Fig. 3A, right).

To test whether systemic delivery of miR-34a could impact the progression of TNBC, MDA-MB-231 tumors, originally inoculated at the orthotopic mammary site, were treated systemically with tNP-packaged miR-34a or miR-Scr control. After three treatments of a 1.5 mg/kg dose every 4 days, mice given miR-34a had significantly reduced tumor growth (Fig. 3B). Tumors were harvested for further analysis, and while tumor weights were only slightly decreased in the miR-34a arm, a marked reduction in Ki67 staining was observed when compared with controls (Fig. 3E). miR-34a levels were elevated approximately 2.5-fold in the miR-34a–treated tumors (Fig. 3C), while bona fide miR-34a targets such as JAG1 were downregulated (Fig. 3D). Taken together, these data indicate tNP can deliver functionally active miR-34a to tumors located at the mammary orthotopic site, resulting in miRNA-mediated phenotypes in vivo. To assess toxicity associated with systemic delivery of miR-34a, we collected serum from mice given miR-Scr and miR-34a treatments. Liver function tests showed no evidence of hepatic damage (Fig. 3F). Liver tissue architecture was also normal, and we did not find any evidence for TUNEL⁺ cells (Fig. 3G).

We next tested whether miR-34a could sensitize TNBC cells to certain front-line agents for TNBC patients. Using clonogenic assays, we found 10 nmol/L miR-34a induced variable sensitivities to 0.15–4 nmol/L paclitaxel in MDA-MB-231, BT-549, MDA-MB-453, and Hs578T cells (Fig. 4A–C and Supplementary Fig. S3B). Although significant, this was a mild sensitization effect. Instead, given miR-34a has strong antitumorigenic effects in the MSL- and M-TNBC subtypes, and that these subtypes are sensitive to agents such as dasatinib due to heightened expression of focal adhesion genes such as c-SRC (6) (Supplementary Fig. S3A), we tested whether miR-34a could synergize with dasatinib in TNBC cells. Indeed, 10 nmol/L miR-34a sensitized MDA-MB-231, BT-549, Hs578T, and MDA-MB-436 cells to 11–900 nmol/L dasatinib, as compared with miR-Scr control (Fig. 4A and B and Supplementary Fig. S3B and S3C). We did not observe this sensitivity in the LAR-TNBC MDA-MB-453 cells or in HeLa cells (Fig. 4C and Supplementary Fig. S3C), giving credence to the notion that miR-34a synergizes with dasatinib by regulating similar pathways important to the biology of M-TNBC cells.

Given this miR-34a–dasatinib synergy, we wondered whether dasatinib could affect endogenous miR-34a levels in TNBC. MDA-MB-231 cells treated with dasatinib for 60 hours demonstrated a dose-dependent increase in miR-34a levels, with the highest fold-change being 4.4-fold in 100 nmol/L treated cells as compared with the 0 nmol/L control (Fig. 4D). This effect appeared selective given miR-15a levels did not change more than 1.7-fold across any treatment condition. Dasatinib-induced miR-34a was functional as it resulted in reduced miR-34a sensor reporter activity (Fig. 4E), and levels of miR-34a targets such as JAG1 (Supplementary Fig. S3D; ref. 31). These effects are dasatinib-specific, as paclitaxel-treated MDA-MB-231 cells showed no significant increase in miR-34a levels, even at cytotoxic levels ≥1 nmol/L (Supplementary Fig. S3E). While the mechanism(s) by which dasatinib-mediated induction of miR-34a levels occur(s) is unclear, given dasatinib is a c-SRC inhibitor, c-SRC signaling could regulate activity of a transcription factor or an RNA binding protein that controls expression, stability, and turnover of miR-34a. Since we observed no active luciferase expression generated from reporters containing previously described miR-34a promoters (Supplementary Fig. S3F; ref. 32), this indicates that dasatinib-mediated induction of miR-34a was not transcriptional.

If part of the antitumorigenic property of dasatinib is through enhancement of miR-34a activity, then a dasatinib-resistant line (MDA-MB-231-Das^R cells; Fig. 4F), which has an abrogated response to dasatinib, should harbor lower levels of miR-34a and have weaker miR-34a dasatinib-associated cytotoxicity. Indeed, MDA-MB-231-Das^R cells harbored 3.9-fold lower levels of miR-34a in the presence of 200 nmol/L dasatinib (Fig. 4H). In addition, in our clonogenic model, 1 μmol/L dasatinib alone killed approximately75% of the MDA-MB-231-Das^R cells (Fig. 4G), whereas approximately 95% kill is observed in parental cells at this dosage. Furthermore, while addition of miR-34a did sensitize the MDA-MB-231-Das^R cells to dasatinib, the effect was milder, with upwards of 15 nmol/L miR-34a and 1 μmol/L dasatinib required to evoke a 90% kill (Fig. 4G). This value was 98% in parental MDA-MB-231 cells using 10 nmol/L miR-34a + 300 nmol/L dasatinib. Despite these subtleties, miR-34a significantly sensitized both parental TNBC and Das^R cells to dasatinib. Given that clinical trials utilizing dasatinib in combination with front-line chemotherapeutic agents such as paclitaxel are still ongoing for patients with metastatic breast cancer (NCT00820170 and NCT00924352), we conjecture that addition of a miR-34a agent to these experimental treatment regimens would result in significantly higher response rates.

Given that dasatinib is a c-SRC inhibitor, the finding that dasatinib promotes miR-34a levels while concomitantly reducing c-SRC levels (Supplementary Fig. S3D) suggests c-SRC in M-TNBC cell lines may be part of a feedback loop where c-SRC dampens miR-34a expression and miR-34a directly targets and represses c-SRC (Supplementary Fig. S3J). While there was an inverse correlation between miR-34a and c-SRC levels in vitro (Fig. 1A, Supplementary Fig. S3A and S3G), we wanted to determine whether this occurred in tumor samples. We assessed Metabric data, as Illumina mRNA- and Agilent miRNA-microarray data are available on 1,262 patient samples. Spearman rank correlation analysis identified anticorrelation between SRC and miR-34a levels (ρ = −0.064; n = 1262; P < 0.008; Supplementary Fig. S3H). TCGA data were also assessed; however, we found no significant anticorrelation (ρ = −0.01; P = 0.33), which may be due to the smaller sample size (n = 632).

We then looked for miR-34a target sites within the SRC 3′UTR using miRNA prediction algorithms and found three putative target sites (Fig. 5A). We performed 3′UTR luciferase targeting assays whereby the SRC 3′UTR was cloned into pLightSwitch. Cotransfection experiments in HEK-293T cells indicated a dose-dependent reduction in luciferase activity in miR-34a–transfected cells (Fig. 5B). This effect was not observed on an Empty-3′UTR control or in miR-Scr control–transfected cells. Additional mutational analysis of the SRC 3′UTR by Muppala and colleagues (33), indicated that miR-34a–induced repression of SRC 3′UTR luciferase activity was predominately mediated by the middle miR-34a binding site depicted in Fig. 5A, which is also the most conserved out of the three target sites according to TargetScan. Importantly, 72 hours after miR-34a transfection, MDA-MB-231, MDA-MB-436, and BT-549 cells harbored 5.6-, 2.0-, and 4.4-fold reductions in SRC transcript levels, respectively (Fig. 5C). Reductions in c-SRC protein expression, as well as the SRC-regulated gene β-catenin, were also observed in miR-34a–transfected TNBC cells but not in luminal-A MCF-7 cells (Fig. 5D). In addition, miR-34a–treated MDA-MB-231–derived tumors from our orthotopic experiments harbored approximately 1.9-fold reduced SRC levels (P ≤ 0.049), as compared with miR-Scr–treated tumors (Fig. 5E). Taken together, these results indicate miR-34a and c-SRC are part of a regulatory feedback loop in TNBC cells, whereby enhanced activity of c-SRC results in the concomitant repression of miR-34a, and yet does not appear to involve upstream kinases as the pTyr-416/SRC ratio was not altered by miR-34a reintroduction (Supplementary Fig. S3I).

Given these results, loss of c-SRC should result in a phenocopy of the gain of miR-34a in TNBC cells. Indeed, MDA-MB-231 cells transfected with 15 nmol/L si-SRC promoted a strong antigrowth and prosenescent phenotype (Fig. 5F–H), similar to miR-34a–transfected cells. No effects were observed in 15 nmol/L si-Neg control–treated cells. This ≥2.5-fold antigrowth phenotype induced by 15 nmol/L si-SRC was also observed in additional TNBC lines (Fig. 5H and Supplementary Fig. S4A and S4B), and when using a second c-SRC siRNA (Supplementary Fig. S4C). In normal HFF fibroblasts, SRC knockdown resulted in only a 1.3-fold reduction in cell growth as assayed by crystal violet (Supplementary Fig. S4D). si-SRC–transfected TNBC cells also had approximately 2-fold higher miR-34a levels as compared with si-Neg–transfected cells (Fig. 5J and Supplementary Fig. S4F), confirming the negative regulatory actions of c-SRC on miR-34a, as highlighted by the dasatinib experiments. SRC knockdown efficiency experiments confirmed si-SRC–transfected TNBC cells resulted in approximately 5-fold or greater reductions in SRC mRNA levels (Supplementary Fig. S4G), as well as significant reductions in c-SRC protein levels (Fig. 5I and Supplementary Fig. S4C). When senescence assays were performed in si-SRC–transfected MDA-MB-231 cells, we observed a 3.6-fold increase in β-gal–positive cells (Fig. 5G). In an extended panel of si-SRC–transfected TNBC lines, the effects were not as robust (Supplementary Fig. S4E), suggesting loss of a single miR-34a target is not sufficient to induce cytostasis, and that miR-34a regulates multiple proto-oncogene targets in TNBC cells to induce prosenescent phenotypes.

Given that c-SRC is an important proto-oncogene in TNBC regulated by miR-34a, we performed gain-of-function rescue experiments to directly assess the contribution of c-SRC in miR-34a–mediated tumor suppression. We generated stable MDA-MB-231-SRC-ORF-GFP–overexpressing cell lines through lentiviral transduction. Here, c-SRC is expressed from an RNA transcript without its native 3′UTR, and therefore is not under the regulatory control of miR-34a, as we computationally determined potential miR-34a binding sites across the entire SRC transcript via RNAhybrid (34), and found none within the coding region. SRC mRNA levels were more than 40-fold in the SRC-ORF-GFP lines as compared with a control MDA-MB-231-Empty-GFP–overexpressing line, or parental MDA-MB-231 cells (Supplementary Fig. S5A). SRC protein expression was elevated only approximately 3.0-fold in the SRC-ORF-GFP lines as compared with the parental MDA-MB-231 cells (Fig. 6A). This mild overexpression of c-SRC is expected, given persistently high levels of c-SRC results in severe cytotoxicity (35), and is supported by our observation that GFP^high cell populations cannot be maintained in culture. This also indicates the SRC-ORF-GFP line expresses c-SRC at physiologically appropriate levels. This c-SRC is functional, given we observed morphologic changes in these cells including greater frequency of cell rounding accompanied by similar rates of cell growth, as has been reported by others (35). We also observed reduced miR-34a levels in the SRC-ORF-GFP line (Fig. 6E), expected given our findings that c-SRC negatively regulates miR-34a.

We performed several functional tests in the presence of exogenous miR-34a or miR-Scr control, to determine whether SRC-ORF-GFP cells were resistant to miR-34a effects. SRB assays indicated growth of Empty-GFP cells after miR-34a transfection was reduced 3.4-fold, as compared with miR-Scr control treatment (Fig. 6B). However, SRC-ORF-GFP cells only harbored an approximately 1.5-fold reduction in growth post-miR-34a transfection (Fig. 6B and Supplementary Fig. S5B). SRC-ORF-GFP lines were also protected from the effects of miR-34a in soft agar assays. SRC-ORF-GFP cells underwent colony formation with greater efficiency than Empty-GFP lines, as determined by the 1.3-fold increase in colony number (Fig. 6C), and by the increased colony size. While miR-34a transfection resulted in a 3.5-fold decrease in colony formation in the Empty-GFP line, miR-34a induced only an approximately 1.3-fold decrease within the SRC-ORF-GFP line. Therefore we conclude that the overabundance of c-SRC prevents miR-34a–meditated inhibition of anchorage-independent cell growth.

We next tested whether SRC-ORF-GFP lines had a blunted response to the effects of miR-34a on cell invasion (Fig. 6D). Similar to parental MDA-MB-231 cells, Empty-GFP lines transfected with miR-34a resulted in a 2.4-fold inhibition of invasion. At baseline, SRC-ORF-GFP cells had approximately 2.5-fold greater rates of invasion than Empty-GFP cells, as would be expected given the promigratory phenotypes conferred by c-SRC through activation of FAK, RhoA, and PAK signaling (36). However, miR-34a only conferred a 1.2-fold reduction of invasion in the SRC-ORF-GFP cells, indicating gain of c-SRC expression dampens the anti-invasive effects of miR-34a. Overall, these rescue experiments indicate miR-34a regulation of c-SRC partly contributes to the antigrowth effects, but to the majority of the anti-invasive phenotypes mediated by miR-34a reintroduction into TNBC cells.

Given the strong antitumorigenic effects of miR-34a within TNBC cells in vitro and in vivo, we asked whether a miR-34a target gene signature could be used as prognostic indicator for TNBC. MDA-MB-231, BT-549, and MDA-MB-436 cells were transfected with either miR-34a or miR-Scr control for 72 hours, and qPCR breast cancer array experiments were performed on RNA isolates (Fig. 6F). As MDA-MB-436 and BT-549 cells were highly correlated with each other (Supplementary Fig. S5C), they were used in the subsequent K-Means clustering analysis (Supplementary Fig. S5D). This analysis identified 4 gene clusters. Clusters 2 and 3 were most affected by miR-34a reintroduction, as 40% to 56% of these genes were downregulated ≥2-fold on the qPCR array (Supplementary Fig. S5E). KEGG pathway analysis indicated that 16 of the 50 genes in Clusters 2 and 3 were involved in “focal adhesion” (Supplementary Fig. S5F), intriguing given M-TNBC subtypes harbor a gene expression pattern that includes elevated focal adhesion pathways. Furthermore, in the original hypergeometric analysis, putative miR-34a targets were found to be overrepresented in MSL-and M-TNBC subtypes. This clustering and KEGG analysis indicates that miR-34a reintroduction in these cell lines reduces the expression of focal adhesion genes important in M-TNBC biology.

To identify a robust miR-34a gene signature, we asked whether the miR-34a–downregulated genes identified in Clusters 1–4 were also downregulated in a third cell line, MDA-MB-231, upon miR-34a reintroduction. Using a 2-fold cut-off criterion, seven genes emerged BCL2, GRB7, IGF1R, JAG1, KRT18, NOTCH1, and SRC as being consistently downregulated across 3 separate TNBC lines (Fig. 6G). Three of these genes, BCL2, JAG1, and NOTCH1, are known direct targets of miR-34a (37), while the novel miR-34a target c-SRC discovered in this work was part of this signature as well. Data obtained from Metabric indicated all 7 genes were inversely correlated with miR-34a in TNBC patients (n = 203), albeit with some variability in significance (Supplementary Fig. S5G). More important is the finding that this seven-gene signature harbors prognostic value. Using the KM-plotter tool, capable of evaluating survival in 4,142 breast cancer samples (38), we found high expression of these miR-34a targets confers a worse OS (Fig. 6H). This effect was stronger in TNBC patients (n = 245, HR = 3.3, P = 0.00002) compared with those with luminal-A tumors (n = 611, HR = 1.7, P = 0.005). The prognostic impact of this gene signature was further validated using a PROGgeneV2 algorithm on TCGA breast cancer data (Supplementary Fig. S5H). Overall, our data supports the idea that reintroduction of miR-34a in M-TNBC cell lines promotes antitumorigenic phenotypes largely by targeting the proto-oncogene SRC.

In this study, we show miR-34a is dysregulated in TNBC, and when reintroduced can promote tumor-suppressive effects both in vitro and in vivo. Furthermore, we demonstrated that miR-34a replacement could successfully be performed using tNP technology as a delivery method in mouse xenograft models. miR-34a imparts these phenotypic effects by targeting SRC as well as other bona fide targets such as BCL2 and NOTCH1 (39). Moreover, since a double-negative feedback loop exists between c-SRC and miR-34a, TNBC cells can be sensitized to the c-SRC inhibitor dasatinib after miR-34a reintroduction.

miR-34a belongs to the miR-34 family of miRNAs: miR-34a, miR-34b, and miR-34c. MiR-34a is located at 1p36 and is encoded in its own transcript, whereas miR-34b and miR-34c share a primary transcript on 11q23 (40). All of these miRNAs share the same seed sequence, and therefore have similar endogenous mRNA targets. The expression of miR-34b and miR-34c is low across all breast cancers, due to allelic deletions and/or the loss of heterozygosity that frequently occurs at 11q23 (41, 42). In addition, miR-34a can be epigenetically regulated through promoter hypermethylation in some breast cancer cell lines and human tumor specimens (43, 44). Given the small sample size of these studies, it is still difficult to conclude whether aberrant methylation of miR-34a plays a predominate role in TNBC.

miR-34a is well established to be under the transcriptional control of the TP53 tumor-suppressive network (45–47). TP53 directly binds the miR-34a promoter inducing its expression, while miR-34a indirectly promotes p53 expression by targeting and repressing levels of the deacetylase SIRT1, which negatively regulates TP53. Given this positive feed-forward regulatory loop, tumors harboring wild-type TP53 have higher levels of miR-34a than tumors containing a TP53-mutant allele. As many TNBC cell lines and tumors harbor TP53-inactivating mutations, or are TP53 null, it is not surprising we observed a striking reduction in miR-34a levels within these cell lines and patient samples.

We found both dasatinib and siRNA to c-SRC resulted in enhanced miR-34a levels and activity in TNBC cells. We did not observe similar changes when cells were treated with paclitaxel, and found no evidence for SRC-mediated transcriptional regulation of miR-34a. Furthermore, SRC-ORF overexpressing MDA-MB-231 cells harbored approximately 2-fold lower miR-34a levels. Together, these data suggest c-SRC, which is highly expressed in TNBC, results in the post-transcriptional dampening of miR-34a levels. Further investigation will be required to decipher the post-transcriptional mechanism by which c-SRC controls miR-34a turnover or decay in TNBC cells. However, as c-SRC can promote PI3K/AKT signaling, miR-34a levels may be controlled by the PI3K/AKT/MDM2 signaling as suggested by Silber and colleagues (48).

We also found miR-34a directly targets c-SRC within its 3′UTR. Reintroduction of miR-34a into TNBC cells significantly reduced SRC mRNA levels both in vitro and in vivo, as well as c-SRC protein levels in vitro. These data suggest miR-34a and c-SRC form a double negative-feedback regulatory loop (Supplementary Fig. S3J). Numerous miRNA—RNA regulatory loops have been identified, and function as an essential regulatory unit that controls the fine-tuning of gene expression important in proper developmental processes (14). In cancer, these feedback loops can become dysregulated, allowing for a permissive tumorigenic state. In our study, it appears that M-TNBC cells harboring elevated levels of c-SRC exhibit reduced levels and activity of miR-34a. This may explain why miR-34a reintroduction was able to specifically sensitize M-TNBC cells to dasatinib treatment, while it promoted a marginal sensitization to paclitaxel. Furthermore, the impact of miR-34a when reintroduced as a stand-alone agent specifically within the M-TNBC subtype was quite striking. Consistent with this effect, hypergeometric analysis indicated MSL- and M-TNBC cells highly express genes that are putative miR-34a targets, more so than other TNBC subtypes. Overall, these data suggest the pathways aberrantly overexpressed in M-TNBC cells important in supporting a protumorigenic state, can also be suppressed by miR-34a, and is why miR-34a levels are found at an extremely low steady-state in these cancers.

Overall, we observed a robust antigrowth, anti-invasion, and procytostatic phenotype occurring in TNBC cells transfected with miR-34a. This complete shutdown of the protumorigenic network indicates numerous genes essential for tumor survival are being downregulated by miR-34a. For instance, we observed downregulation of canonical targets such as BCL2, NOTCH1, and IGF1R (39, 49). As highlighted in Fig. 6I, many of these miR-34a targets were also embedded within the c-SRC signaling pathway, including SRC itself, which is activated by integrin and focal adhesion kinase (FAK) signaling. Interestingly, GRB7, an adaptor protein important in mediating FAK/SRC signaling was also downregulated by miR-34a in TNBC cells. This effect is predicted to result in a dampening of downstream MAPK and PI3K/AKT signaling pathways important for growth and survival. In support of this, miR-34a–treated cells displayed decreases in AKT-regulated genes such as BCL2 and β-catenin, which are known to be elevated by c-SRC (50). Even the established miR-34a target NOTCH1 can be controlled by c-SRC during an important processing step involving c-SRC–induced Furin-mediated S1 cleavage of NOTCH1 in the trans-Golgi network (51).

Given the multigene targeting occurring in TNBC cells, and from our in vitro and in vivo data, we conclude that miR-34a could be a promising therapeutic agent for treating TNBC patients. Recently, there has been a major effort to develop small RNAs into therapeutic agents (22, 52, 53). Our observations are clinically relevant because an amphoteric liposomal formulation complexed with a synthetic miR-34a mimic (MRX34) is currently being tested in a phase I clinical trial for patients with hepatocellular carcinoma (54). In preclinical models, MRX34 inhibited tumor growth of Hep3B or HuH7 cells orthotopically implanted into the livers of NOD/SCID mice (17). Given miR-34a can target c-SRC and sensitize TNBC cells to dasatinib, our data suggest that the combination of a miR-34a–based therapeutic with dasatinib is worth exploring in future clinical studies of TNBC.

F.J. Slack has ownership interest (including patents) and is a consultant/advisory board member for Mirna Therapeutics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: B.D. Adams, L. Pusztai, W.M. Saltzman, F.J. Slack

Development of methodology: B.D. Adams, C.J. Cheng, W.M. Saltzman, F.J. Slack

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B.D. Adams, V.B. Wali, C.J. Cheng, S. Agarwal, L. Santarpia, L. Pusztai

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B.D. Adams, V.B. Wali, S. Inukai, C.J. Booth, B. Győrffy, L. Santarpia, L. Pusztai, W.M. Saltzman, F.J. Slack

Writing, review, and/or revision of the manuscript: B.D. Adams, V.B. Wali, C.J. Cheng, S. Inukai, C.J. Booth, B. Győrffy, L. Pusztai, W.M. Saltzman

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B.D. Adams, D.L. Rimm

Study supervision: B.D. Adams, L. Pusztai, L. Pusztai, F.J. Slack

Other (interpretation, documentation, and photomicroscopy of the pathologic changes in the tissues): C.J. Booth

The authors thank Dr. Leonard Milstone and Jing Zhou for providing the human fibroblast cell line (HFF), Joanna Hu for providing the CRC lines, Lixia Diao for bioinformatic support on the miRNA array analysis, Geoffrey Lyon, Yale Cell Sorter Core Facility, for assistance with FACS sorting and analysis, Amos Brooks, Research Histology & Yale Department of Pathology for the histology and IHC, the Firefly Frontiers Grant from Firefly BioWorks for providing funds for miRNA profiling experiments, and Shirin Bahmanyar for providing space and making equipment available for these studies. The authors also thank Eleni Anastasiadou and Catherine Adams for the critical reading of this manuscript.

This work was supported by a pilot grant from the Yale Cancer Center and NIH grants (R01 CA157749 and R01 CA131301 to F.J. Slack), P30 CA016359-35S2 (D.L. Rimm), and R01 CA149128 (W.M. Saltzman). L. Pusztai was supported by Breast Cancer Research Foundation Award, B. Győrffy by the OTKA K108655 grant, and L. Santarpia by AIRC grant 6251.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.