miRNA-520f Reverses Epithelial-to-Mesenchymal Transition by Targeting ADAM9 and TGFBR2 Free

Cancer is a leading cause of death across the world, with the most common type of cancer being carcinoma, which originates from epithelial tissues. Despite major advances in our understanding of the molecular and genetic basis of cancer, metastasis still causes 90% of all cancer-related deaths, and remains one of most complex and challenging problems of contemporary oncology. Thus, there is an urgent need for the development of new therapeutic approaches to treat and prevent metastatic disease. A potential strategy to achieve this goal is to target epithelial-to-mesenchymal transition (EMT).

EMT plays a key role in tumor cell invasion and metastasis (1). Within the normal epithelium, cells interact with neighboring cells and the basement membrane to provide tissue integrity, a paracellular barrier, and cell polarity. During carcinoma progression, cells lose their epithelial characteristics and gain a more mesenchymal phenotype. This so-called EMT is associated with the ability of cells to migrate and invade into the surrounding tissue. These processes are the first steps toward metastatic spread (2). In addition, cells that have undergone EMT share many characteristics with stem cells, such as increased drug resistance, complicating systemic therapy for metastatic disease (3, 4).

EMT in tumor cells is the result of transcriptional reprogramming in the cell (5). In particular, transcriptional repression of CDH1 (encoding the cell-cell adhesion protein E-cadherin) has been shown to trigger EMT. Several transcription factors such as ZEB1, ZEB2, SNAI1, and SNAI2 are known to repress CDH1. EMT is a reversible process, and silencing the CDH1 repressors results in mesenchymal-to-epithelial transition (MET), restoring the non-invasive epithelial phenotype (6–9). Reversing EMT may, therefore, be an effective approach to combat advanced tumors in various types of cancer.

miRNA are small noncoding RNA molecules that posttranscriptionally regulate gene expression by controlling the translation and stability of mRNAs (10). Individual miRNAs can regulate up to hundreds of mRNAs and a single mRNA may be targeted by several miRNAs (11). Although initially discovered for their role in differentiation, they have now been shown to play critical roles in a wide range of cellular processes, such as proliferation and apoptosis (12). Furthermore, deregulation of miRNAs plays an important role in cancer (13). Over expression of oncogenic miRNAs leads to repression of tumor-suppressor genes, and the loss of tumor-suppressive miRNAs enhances the expression of oncogenes. Because miRNAs function as master regulators of gene expression, miRNA-based therapy is an attractive approach for cancer therapy.

Recently, it was found that downregulation of miR-200 family members (including miR-141 and miR-200c) results in EMT (14). Ectopic expression of these miRNAs in mesenchymal tumor cells restored their epithelial phenotype. This transition was characterized by reactivation of E-cadherin expression, restoration of cell–cell adhesion, inhibition of tumor cell invasion, and increased sensitivity to chemotherapeutic agents. Transcription factors ZEB1 and ZEB2 were identified as the major targets of EMT-regulating miR-200 family members. The other repressors of CDH1, however, were not targeted by the miRNAs of the miR-200 family. Identification of novel miRNAs that target the whole or a broader repertoire of CDH1 repressors and other EMT regulators is critical to formulating a universal miRNA-based anti-EMT or anti-metastatic therapy.

In this study, we aimed to identify novel miRNAs that are able to reverse EMT and inhibit metastasis. Because repression of CDH1 is a key molecular event in EMT, we screened miRNAs for their ability to re-activate CDH1 transcription. Positive miRNAs were further validated by analyzing their effect on tumor cell invasion and metastasis. In addition, we identified potential target genes of the most promising miRNA, miR-520f, to elucidate the mechanism of action.

The PANC-1 cell line was purchased from the ATCC and T24 cells were received from the University of Colorado Health Sciences Center (Department of Pathology) in 1998. Both cell lines were authenticated in 2016 using the PowerPlex 21 system (Promega) by Eurofins Genomics (Germany). The tumor cell lines and culture conditions are described in the Supplementary Materials and Methods and in Supplementary Table S1. Cell lines were frequently tested for Mycoplasma infection, using a Mycoplasma-specific PCR, and cells were propagated for no more than 6 months or 30 passages after resuscitation from stocks.

Human miRNAs were selected from both the public miRNA repository (www.mirbase.org) and proprietary small RNA deep-sequencing data (15). The miRNA sequences were amplified from human genomic DNA, with the amplicons containing the full-length pre-miRNA hairpin and a flanking sequence on both sides. The primers for the amplicons were complemented with a 5′ GCGC overhang and a restriction site for directional cloning into the pCDH-CMV-MCS-EF1-puro vector (SBI, System Biosciences). Correct cloning was verified by DNA sequence analysis (see Supplementary Materials and Methods and Supplementary Table S2). VSV-G pseudotyped lentiviral particles were packaged, amplified, and provided by SBI (SBI, System Biosciences). Viral particles were stored at −80°C.

T24-pEcad-luc/Rluc (Supplementary Table S1) cells were infected with lentivirus (1.0 μL of undiluted lentiviral particles, average multiplicity of infection = 64) in complete medium containing 2 μg/mL polybrene (Sigma) in a 96-well format, and selected with 1 μg/mL puromycin (Sigma). Six days after infection, firefly and Renilla luciferase activities were measured using the dual-luciferase reporter assay system (Promega), on a Victor3 multilabel counter (PerkinElmer). The Fluc/Rluc ratio was used as a measure for reporter activation.

The miR-520f precursor sequence was amplified using DNA from pCDH-miR-520f infected T24 cells as a template and pCDH-specific primers (Supplementary Table S3). The miR-520f precursor was cloned into the NheI/EagI site of the pmRi-ZsGreen1–inducible Vector (Clontech, Supplementary Fig. S1A). T24 cells stably transfected with the pTet-On Advanced Vector (Clontech, Supplementary Fig. S1A) were obtained by clonal selection in G418-containing medium (600 μg/mL, Gibco). Inducible reverse tetracyclin-responsive transcriptional activator (rtTA) expression was confirmed by a luciferase reporter assay in T24-Tet-On clones that were transiently transfected with the pTRE-Tight-luc vector (Supplementary Fig. S1B). A stable T24-Tet-On clone was then transfected with the pmRi-ZsGreen1-miR-520f vector, and clonally selected in medium containing puromycin (1 μg/mL, Sigma). Transfected clones displayed doxycycline-inducible miRNA expression at 1 μg/mL doxycycline (Supplementary Fig. S1C and S1D). Transfections were performed using X-tremeGENE 9 transfection reagent, according to the manufacturer's instructions (Roche).

miRNA mimics and siRNAs (Supplementary Table S4) were transfected using Lipofectamine RNAiMAX Reagent (Invitrogen) according to the manufacturer's protocol.

Total RNA was isolated using TRIzol, according to the manufacturer's instructions (Invitrogen). Concentration and purity of the RNA was determined on a Nanodrop-1000 spectrophotometer (Thermo Scientific). For microarray analysis, RNA was further purified using an RNeasy micro kit (QIAGEN). RNA integrity was determined on an Agilent Bioanalyzer 2100 (Agilent technologies).

For gene expression analysis, 2 μg DNase-I–treated total RNA was used to generate cDNA, using random hexamer primers (Roche) and SuperScript II Reverse Transcriptase (Invitrogen). For miRNA analysis, a stem loop (SL) RT-PCR was performed; for this, 100 ng total RNA was reverse transcribed using 0.375 pmol miR–specific SL-RT primer. Real-time-PCR analysis was performed using LightCycler 480 SYBR Green I Master mix (Roche). RNA not subjected to reverse transcriptase was used as a control for nonspecific PCR amplification. All primers and amplification conditions are listed in Supplementary Table S5. (SL) RT-PCR was performed on a LightCycler LC480 instrument (Roche). Expression levels of B2M and GAPDH (mRNA) or RNU6-1 (U6; miRNA) were used for normalization. Relative gene expression levels were calculated according to the mathematical model for relative quantification in real-time PCR described by Pfaffl (16).

To assess protein levels of E-cadherin and ADAM9, we performed SDS-PAGE and Western blot analysis on T24 and PANC-1 protein extracts. A more detailed description can be found in the Supplementary Materials and Methods. Mouse anti–E-cadherin (HECD-1, Takara) and rabbit anti-ADAM9 (#2099, Cell Signaling Technologies) were used to detect E-cadherin and ADAM9 respectively. Mouse anti–β-actin (AC-15, Sigma) staining was used for normalization.

Invasion assays were performed using BioCoat Matrigel Invasion Chambers (Corning), according to the manufacturer's instructions. Before the invasion assay, T24-imiR-520f cells were incubated in the presence of doxycycline (1 μg/mL) for 2 days, and transduced cells were puromycin-selected and passaged one time. MiRNA mimic or siRNA transfected cells were collected by trypsinization 72 hours after transfection. We seeded 40,000 cells into the invasion chambers in serum-free medium. The invasion chambers were placed in a 24-well plate containing 500 μL medium with 10% FCS as a chemo-attractant. As a control, 40,000 cells were seeded in a 24-well plate. After 48 hours of incubation, cells in the inner compartment of the invasion chamber were removed by aspiration and cleaning with cotton swabs. Invasive cells were quantified by incubating the bottom of the invasion chamber in CellTiter-Glo (CTG) reagent (Promega), and then analyzing luminescence on a Victor3 multilabel counter (PerkinElmer). The percentage of invasive cells was calculated as the CTG activity of the invasive cells, normalized for the CTG activity of the input control (i.e., cells grown in a 24-well plate).

Animal experiments were approved by the ethical committee on animal research of the Radboud university medical center (Nijmegen, NL). For this experiment, we used 24 female, 6- to 8-week-old NOD-SCID immunodeficient mice (Charles River laboratories) that were divided into four groups of 6 mice. Groups 1 and 2 were given doxycycline-containing (0.2 mg/mL) drinking water during the entire experiment, whereas groups 3 and 4 received doxycycline-free water. Groups 1 and 3 were injected with 0.8 × 10⁶ (200 μL volume) T24-imiR-520f cells and groups 2 and 4 with 0.8 × 10⁶ (200 μL volume) T24-Tet-On cells, all via the lateral tail vein (17). Mice were monitored daily, and when the first mice started to suffer from serious tumor burden, all animals were sacrificed. The number and size of metastases were determined macroscopically (i.e., by visual examination through a dissection microscope) and microscopically on hematoxylin and eosin (H&E)–stained sections (4-μm) of Tissue-Tek–embedded, frozen mouse lungs. Lung tumor burden in the H&E-stained sections was determined by scanning the sections using a Pannoramic 250 Flash II digital slide scanner and analyzing the percentage of tumor area using Pannoramic viewer (3DHISTECH). Twenty 20-μm sections from Tissue-Tek–embedded frozen mouse lungs were used for TRIzol RNA isolation.

Gene expression was examined by Affymetrix GeneChip Human Exon 1.0 ST arrays (Affymetrix). RNA was processed using the GeneChip WT PLUS Reagent Kit (Affymetrix). The arrays were further processed using the GeneChip Hybridization, Wash, and Stain Kit (Affymetrix). All kits were used according to the manufacturer's protocols. GeneChips were scanned with a GeneChip Scanner (Affymetrix), generating CEL files for each array. Gene-expression values were derived from the CEL file using the model-based Robust Multiarray Average (RMA) algorithm as implemented in Partek software (Partek Genomics Suite 6.6). RMA performs normalization, background correction and data summarization. Differentially expressed genes between conditions were calculated using ANOVA. The data discussed in this publication have been deposited in NCBI's Gene Expression Omnibus (18) and are accessible through GEO Series accession number GSE92988 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE92988).

The ADAM9 and TGFBR2 3′-UTR (untranslated region) regions and mutant variants were cloned into the psiCHECK-2 reporter vector (Promega). Mutant variants of the ADAM9 3′-UTR were created using SOEing PCR (Supplementary Fig. S2; Supplementary Table S6). A more detailed description can be found in the Supplementary Data. For the 3′-UTR reporter assays, 5,000 T24 cells per well were seeded in a 96-well plate. One day after seeding, cells were transfected with 100 ng psiCHECK-2-3′-UTR reporter DNA (using X-tremeGENE 9 DNA transfection reagent; Roche), and 20 nmol/L miRNA mimic. The luciferase activities were measured 48 hours posttransfection using the dual-luciferase reporter assay system (Promega). firefly and Renilla luciferase signals were measured on a Victor3 multilabel reader (PerkinElmer).

The data are presented as means ± SEM from at least three independent experiments. Two-tailed t tests were performed using GraphPad Prism (GraphPad Software, Inc.). Clinical data, ADAM9 and TGFBR2 expression of muscle-invasive bladder urothelial carcinoma and pancreatic adenocarcinoma patients included in The Cancer Genome Atlas (TCGA) database were downloaded from cBioPortal (19). Kaplan–Meier analyses were performed using GraphPad Prism (Graphpad Software, Inc.). A P value of <0.05 was considered statistically significant.

To identify miRNAs that can reverse EMT, we developed a screening model to test their ability to induce transcriptional activation of CDH1. We cloned the core element of the CDH1 promoter that is responsible for cell type–specific expression (20) in an expression vector to drive firefly luciferase expression with an HSV-Tk promoter–driven Renilla luciferase cassette as an internal control (Fig. 1A; Supplementary Fig. S3A). This reporter construct was stably transfected into the bladder cancer cell line T24 (T24-pEcad-luc/Rluc). T24 has a mesenchymal phenotype, as determined by the CDH1/VIM ratio (Supplementary Fig. S3B), which has been shown to correlate well with EMT status (21). Because of its low endogenous CDH1 expression, T24 is a suitable model cell line for studying re-activation of CDH1 and reversal of EMT.

We performed a functional screen, using a lentiviral-based miRNA expression library containing 1120 human miRNA precursors in our model cell line T24-pEcad-luc/Rluc (Fig. 1A). The miR-200 family members miR-141 and miR-200c were considered positive controls because they are known to re-activate CDH1 expression by targeting ZEB1 and ZEB2. Re-activation of CDH1 by each miRNA was quantified by calculating the firefly luciferase/Renilla luciferase (FLuc/RLuc) ratio, where miRNAs with a z-score > 2.0 were considered positive. We identified 29 positive miRNAs in our screen, including miR-141 and miR-200c (Fig. 1B). These miRNAs were further validated by studying their effects on the endogenous expression of CDH1 in wild-type T24 cells. Of the 29 positive miRNAs from the screen, 10 miRNAs also induced endogenous CDH1 expression (Fig. 1C). MiR-520f showed the highest induction of endogenous CDH1 expression (Fig. 1C). The identity of miR-520f was confirmed by Sanger DNA sequence analysis of the proviral DNA of transduced cells. MiR-520f is a primate-specific miRNA that lies in a miRNA cluster on chromosome 19 (C19MC). It is not expressed in most normal or malignant human tissues, with the exception of testicular germ cell tumors and some thymomas (data from TCGA and own data, not shown).

To confirm that miR-520f is able to reverse EMT, we analyzed the expression of CDH1 and several EMT-related genes as well as E-cadherin protein levels in T24 cells transfected with miRNA mimics. Our results obtained with the miR-520f mimics confirmed the upregulation of CDH1 that we found using lentiviral based miR-520f expression (Fig. 2A). Western blot analysis showed that protein levels of E-cadherin are also elevated by miR-520f (Fig. 2B). Moreover, we were able to confirm these results in the pancreatic cancer cell line PANC-1 (Fig. 2A and B), which is characterized as an intermediate mesenchymal cell line based on its CDH1/VIM ratio (Supplementary Fig. S3B). In both cell lines, miR-520f mimic transfection markedly increased miR-520f levels (Supplementary Fig. S1E). In T24, upregulation of CDH1 was accompanied by downregulation of SNAI2, while in PANC-1, it was accompanied by downregulation of ZEB1 and ZEB2 (Fig. 2A). In T24, SNAI1 was upregulated (Fig. 2A), yet expression of SNAI1 remains very low in T24 cells (data not shown). Furthermore, miR-520f significantly downregulated the mesenchymal markers CDH2 and VIM in PANC-1, but not in T24 (Fig. 2A).

To further show EMT reversal, we looked at the morphology of our cell line models using phase-contrast microscopy. T24 cells transfected with miR-520f gained a more epithelial-like morphology (Supplementary Fig. S1F). PANC-1 cells, which have a typical cobblestone and epithelial-like morphology, maintained their morphology after transfection with miR-520f (Supplementary Fig. S1F). In addition, we observed an increased membranous E-cadherin staining in miR-520f transfected PANC-1 cells (Supplementary Fig. S4).

Because miR-520f is able to re-activate CDH1 and down regulates several EMT-inducing factors, we hypothesized that miR-520f may reduce tumor cell invasion. To test this, we infected T24 cells with the miR-520f-precursor-containing lentivirus and studied their invasion capacity using Matrigel-coated cell invasion chambers. Tumor cell invasion was inhibited up to 65% by miR-520f (Fig. 2C). To validate these findings, we also tested the invasion capacity of T24 cells with doxycycline-inducible miR-520f expression (T24-imiR-520f) and PANC-1 cells transfected with miR-520f mimics. MiR-520f inhibited invasion over 40% in the induced T24-imiR-520f cells and about 55% in mimic transfected PANC-1 cells (Fig. 2D and E).

To further study the role of miR-520f in EMT and tumor cell invasion, we identified potential target genes of miR-520f by microarray analysis of miR-520f–transfected PANC-1 cells. Genes that were downregulated and contained miR-520f binding sites, as assessed using in silico prediction tools, were selected as candidate targets and validated by qPCR analysis. Using this approach, ADAM9 (encoding a disintegrin and metalloprotease domain-containing protein 9) and TGFBR2 (encoding Transforming growth factor beta receptor II) were identified as potential targets of miR-520f (Fig. 3A and B). Western blot analysis showed that ADAM9 was also downregulated at the protein level (Fig. 3C).

To determine whether ADAM9 and TGFBR2 are direct targets of miR-520f, we performed 3′-UTR luciferase reporter assays. We used psiCHECK-2 reporter constructs containing either the wild-type or mutated ADAM9 3′-UTR or TGFBR2 3′-UTR fragments (Supplementary Fig. S5A). The mutants contained several nucleotide substitutions in the predicted miR-520f–binding sites, which should disrupt miRNA binding (Fig. 3D). In T24 cells transfected with the ADAM9 3′-UTR wild-type construct, which contains two miR-520f–binding sites, luciferase reporter activity was decreased up to 60% by miR-520f, compared with control mimic transfected cells (Fig. 3E). Regulation by miR-520f of the mutant ADAM9 3′-UTR constructs, where only one of the two predicted miR-520f–binding sites was disrupted, was reduced compared to the wild-type construct (Fig. 3E). When both binding sites in the ADAM9 3′-UTR were disrupted, the effect of miR-520f was completely abolished (Fig. 3E). In the TGFBR2 3′-UTR wild-type construct, containing one miR-520f-binding site, miR-520f decreased luciferase reporter activity by 35%, compared with control mimic transfected cells (Fig. 3F). In the mutant construct, the effect of miR-520f was completely abrogated (Fig. 3F).

To assess the role of ADAM9 and TGFBR2 in tumor cell invasion inhibition mediated by miR-520f, we transfected PANC-1 cells with siRNAs targeting ADAM9 and TGFBR2. Western blot and qPCR analyses showed that ADAM9 was successfully silenced by both siRNAs (Fig. 4A). Interestingly, tumor cell invasion of cells in which ADAM9 was silenced, was reduced (Fig. 4B). Similarly, knockdown of TGFBR2 using siRNAs also resulted in a reduction of tumor cell invasion (Fig. 4A and B). To assess the clinical relevance of ADAM9 and TGFBR2 targeting by miR-520f, we used TCGA data to examine the overall survival of muscle-invasive bladder urothelial carcinoma and pancreatic adenocarcinoma patients with high or low ADAM9 expression. High or low ADAM9 expression was defined as the top quartile or lower quartile, respectively, of the ADAM9 expression levels. High ADAM9 expression was associated with a significantly worse overall survival in both cancer types (Fig. 4C). Furthermore, high TGFBR2 expression (defined as top 10% vs. lower 10% in TCGA gene-expression datasets) was associated with a significantly worse overall survival of pancreatic cancer, but not of bladder cancer patients (Fig. 4D).

We next performed a pilot study to test the antimetastatic activity of miR-520f in vivo, by injecting tumor cells with or without miR-520f expression in the tail vein of NOD-SCID mice and monitoring the formation of lung metastases (Fig. 5A). For this purpose, we used T24 cells with doxycycline-inducible miR-520f expression (see materials and methods). To control not only for the effect of miR-520f, but also for Tet-on activity, we used T24 cells transfected with only the pTet-On Advanced Vector (T24-Tet-On) next to T24 cells transfected with both the pTet-On Advanced Vector and the pmRi-ZsGreen1-miR-520f vector (T24-imiR-520f). For both cell lines, we included a group that received doxycycline treatment and one that did not.

After RNA analysis, we found that tumors in mice that were injected with cells containing the pmRi-ZsGreen1-miR-520f vector, but that did not receive doxycycline, showed expression of miR-520f, most likely due to leaky promoter activity (Supplementary Fig. S6A). Furthermore, it was found that doxycycline reduced metastasis formation (Supplementary Fig. S6B). Therefore, to determine the effect of miR-520f on tumor metastasis, we compared animals injected with T24-Tet-On cells that received no doxycycline (as a negative control), with animals injected with T24-imiR-520f cells that also received no doxycycline. The body weight of the animals in all groups was comparable during the study (Supplementary Fig. S6C). The total tumor area, as assessed on H&E sections, was reduced in the animals injected with T24-imiR-520f cells (Fig. 5B and C). Gene expression analysis in tumor foci confirmed the high expression levels of miR-520f in the T24-imiR-520f cells (Fig. 5D). Furthermore, the miR-520f–positive tumors displayed a trend toward upregulation of CDH1 expression (Fig. 5E), a significant downregulation of SNAI2 (Fig. 5F), and a trend toward downregulation of ADAM9 expression (Fig. 5G).

Despite continuous efforts in developing new therapeutic strategies, metastatic disease remains the leading cause of death from cancer. As EMT plays a fundamental role in metastasis, reversing this process may be a promising approach to target metastatic tumor cells. In this study, we identified miR-520f as a novel EMT reversing miRNA. We demonstrated that miR-520f is able to transcriptionally activate CDH1 expression in our screening model. Furthermore, miR-520f elevated endogenous levels of E-cadherin in cell lines with a strong and intermediate mesenchymal phenotype.

Because several transcription factors are known to repress CDH1, we hypothesized that miR-520f regulates these so called EMT inducers, and studied their expression in our model cell lines. Interestingly, miR-520f downregulated SNAI2 in T24 cells, and ZEB1 and ZEB2 in PANC-1 cells. However, we assume that these are not direct targets of miR-520f, as there are no predicted binding sites of miR-520f in the 3′-UTRs of SNAI2 and ZEB1 (TargetScan.org 7.0). Moreover, downregulation of these repressors was mutually exclusive in our cell line models. This is in line with the fact that the EMT inducing transcription factors are differentially expressed in different tumor types (22). Nevertheless, these results indicate that miR-520f can regulate different EMT inducers in multiple cell lines, and that other targets of miR-520f are likely to regulate these repressors. Furthermore, we have shown that miR-520f inhibits tumor cell invasion in multiple cell line models. Our data demonstrate that miR-520f reverses the EMT phenotype, which suggests that miR-520f has antimetastatic activity, making it an attractive therapeutic drug for different cancers.

We identified ADAM9 as a direct target of miR-520f. Our results show that miR-520f can bind to at least two seed-complementary sites in the 3′-UTR of ADAM9. Point mutations of several nucleotides in the miR-520f-binding sites were sufficient to abrogate the effect of miR-520f, showing that ADAM9 is a direct target. Elevated levels of ADAM9 are observed in multiple cancers and have been correlated with cancer progression and metastases (23–27). Furthermore, knockdown of ADAM9 has been shown to reduce cellular migration and invasion (28–31). This indicates that ADAM9 plays an important role in the mechanism by which miR-520f inhibits invasion. Consistent with these data, we also show that siRNA mediated knockdown of ADAM9 inhibits tumor cell invasion. Over expression of ADAM9 has been shown to enhance growth factor-mediated disruption of cell–cell contacts and internalization of E-cadherin (32). Therefore, targeting ADAM9 may inhibit tumor metastasis by regulating E-cadherin-mediated cell–cell adhesion and cellular motility and invasion. This is supported by our finding that siRNA mediated knockdown of ADAM9 increased membranous E-cadherin staining (Supplementary Fig. S4). Furthermore, our TCGA clinical data analysis shows that high ADAM9 expression is associated with a significantly worse overall survival in both bladder cancer and pancreatic cancer patients.

In addition to ADAM9, we also identified TGFBR2 as a direct target of miR-520f. Our results show that miR-520f can bind to at least one seed-complementary site in the 3′-UTR of TGFBR2. Point mutations of several nucleotides in the miR-520f–binding site were sufficient to abrogate the effect of miR-520f, showing that TGFBR2 is also a direct target of miR-520f. TGFβ signaling is known to stimulate cancer progression through the induction of EMT (33). TGFβ receptor II plays an important role in the TGFβ signaling pathway and has been shown to drive cancer progression (34, 35). As TGFβ signaling is known to induce transcriptions factors ZEB1, ZEB2, and SNAI2, miR-520f might downregulate these transcription factors through targeting TGFBR2 (36, 37). Furthermore, knockdown of TGFBR2 has been shown to inhibit tumor cell invasion (38, 39). This is in line with our results showing that siRNA mediated knockdown of TGFBR2 inhibits tumor cell invasion. Furthermore, high TGFBR2 expression is associated with a significantly worse overall survival in pancreatic cancer patients.

The antimetastatic potential of miR-520f was further supported by our in vivo results. In our mouse metastasis model, miR-520f reduced lung metastasis formation. The effect on metastasis, however, was not significant due to the large variation in the formation of metastases in the control group and the small group sizes. Expression analysis of RNA derived from the mouse lung tumors showed that miR-520f expression was raised significantly in tumor cells carrying the miR-520f expression vector. Furthermore, we observed a trend toward CDH1 upregulation and ADAM9 downregulation, and significant SNAI2 downregulation in miR-520f–expressing tumor cells, consistent with our in vitro data. This suggests that targeting EMT was successful and a feasible approach to inhibit metastasis. In this model, we were unable to determine the effect of miR-520f on intravasation, because the tumor cells were injected directly into the circulation. However, miR-520f might also be effective in restraining cells from intravasation by targeting EMT. It would be interesting to assess the effect of miR-520f in an orthotropic cancer model.

Although miR-520f has not been linked to EMT or invasion before, it has been described to play a role in drug resistance. Lai and colleagues (40) reported that in PANC-1 cells, miR-520f enhanced sensitivity to gemcitabine. Furthermore, Harvey and colleagues (41) found that miR-520f increased the sensitivity of a cisplatin resistant neuroblastoma cell line to etoposide and cisplatin. This drug-sensitizing effect of miR-520f might be through inhibition of EMT. EMT has been shown to give rise to cancer cells with a stem cell-like phenotype, including resistance to therapy (3, 4). Interestingly, Josson and colleagues (42) described that knockdown of ADAM9 in a prostate cancer cell line induced sensitivity to doxorubicin, cisplatin, taxotere, gemcitabine, and VP-16. This indicates that miR-520f might also enhance drug sensitivity by targeting ADAM9. Taken together, these studies emphasize the therapeutic potential of miR-520f in advanced cancers, as it not only inhibits tumor cell invasion, but it also re-sensitizes resistant tumor cells to chemotherapy.

In conclusion, our study showed for the first time that miR-520f is able to reverse EMT and inhibit metastasis, underlining the therapeutic potential of this miRNA. The effect of miR-520f can be, at least partially, explained by targeting ADAM9 and TGFBR2, but other factors are likely to play a role as well (Fig. 6). Our results warrant further research to explore the therapeutic potential of miR-520f.

R.Q.J. Schaapveld has ownership interest (including patents) in InteRNA Technologies BV. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J.G.M. van Kampen, R.Q.J. Schaapveld, J.A. Schalken, G.W. Verhaegh

Development of methodology: J.G.M. van Kampen, P.I. van Noort, R.Q.J. Schaapveld, G.W. Verhaegh

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.G.M. van Kampen, O. van Hooij, F.P. Smit, P.I. van Noort, G.W. Verhaegh

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.G.M. van Kampen, O. van Hooij, C.F. Jansen, F.P. Smit, P.I. van Noort, I. Schultz, R.Q.J. Schaapveld, G.W. Verhaegh

Writing, review, and/or revision of the manuscript: J.G.M. van Kampen, O. van Hooij, P.I. van Noort, I. Schultz, R.Q.J. Schaapveld, J.A. Schalken, G.W. Verhaegh

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.G.M. van Kampen, O. van Hooij, C.F. Jansen

Study supervision: P.I. van Noort, I. Schultz, R.Q.J. Schaapveld, J.A. Schalken, G.W. Verhaegh

Part of the results shown here are based upon data generated by the TCGA Research Network: http://cancergenome.nih.gov/. We are grateful to Jeroen Mooren and Debby Smits of the Central Animal Laboratory Nijmegen for their help and expertise on the animal work. Furthermore, we would like to thank Diede van Bladel, Joost Hendriks, and Kirsten van Niekerk for their help with experiments.

This work was supported by InteRNA Technologies BV and by the Dutch Technology Foundation STW (project number: 12439).

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.