ADAM9 Up-Regulates N-Cadherin via miR-218 Suppression in Lung Adenocarcinoma Cells

Lung cancer is the leading cause of cancer death worldwide, and brain metastasis is a major cause of morbidity and mortality in lung cancer. CDH2 (N-cadherin, a mesenchymal marker of the epithelial-mesenchymal transition) and ADAM9 (a type I transmembrane protein) are related to lung cancer brain metastasis; however, it is unclear how they interact to mediate this metastasis. Because microRNAs regulate many biological functions and disease processes (e.g., cancer) by down-regulating their target genes, microRNA microarrays were used to identify ADAM9-regulated miRNAs that target CDH2 in aggressive lung cancer cells. Luciferase assays and western blot analysis showed that CDH2 is a target gene of miR-218. MiR-218 was generated from pri-mir-218-1, which is located in SLIT2, in non-invasive lung adenocarcinoma cells, whereas its expression was inhibited in aggressive lung adenocarcinoma. The down-regulation of ADAM9 up-regulated SLIT2 and miR-218, thus down-regulating CDH2 expression. This study revealed that ADAM9 activates CDH2 through the release of miR-218 inhibition on CDH2 in lung adenocarcinoma.


Introduction
Lung cancer represents the leading cause of cancer-related death in the Western world. This disease has a 5-year overall survival rate of only 15%, and this has not improved during recent decades [1]. In Taiwan, lung cancer is also the leading cause of cancer death [2], and adenocarcinoma is the major histological type (52.5%). Metastasis is a major cause of morbidity and mortality in lung cancer. Surgical resection of primary lung cancer is frequently followed by tumor recurrence at distant sites, such as the lymph nodes [3], bone [4], and brain [5]. Approximately 30% of patients with lung cancer develop brain metastasis [5]. However, the mechanisms mediating lung cancer metastasis to the brain remain unclear.
Cancer invasion into distant sites requires the degradation of extracellular matrix components, which may be mediated by matrix metalloproteinases, and the loosening of epithelial cell-cell junctions and adhesions to generate mesenchymal cell types, which is referred to as the epithelial-mesenchymal transition [6,7]. Currently, several genes related to lung cancer brain metastases have been identified, such as CDH2 and ADAM9 [8,9]. Neural cadherin (N-cadherin), encoded by the CDH2 gene, is a transmembrane protein and plays an important role in cell adhesion [10]. In most cancers, the expression of CDH2 increases during tumor progression [11] and induces cell migration and invasion as a mesenchymal marker in the epithelial-mesenchymal transition [6,12]. These observations indicate that CDH2 plays a critical role in metastasis [11,12]; therefore, its expression needs to be tightly regulated. CDH2 expression can be regulated by methylation, transcription factors, and microRNAs (miRNAs). For example, the expression of CDH2 in gastric cancer cells was up-regulated following demethylation [13]. Additionally, CDH2 expression is regulated by several transcription factors, such as Twist 1 [14], TP63 [15], and CTNNB1 [16]. Currently, little is known about how miRNAs regulate CDH2. Only miR-145 has been reported to target CDH2 in gastric cancer [17], and it remains unclear whether other microRNAs can regulate CDH2.
MiRNAs are a class of small non-coding RNAs that are approximately 22 nucleotides in length [18] and originate from longer primary miRNA transcripts located in either intergenic or intronic regions. Intergenic miRNAs are located in the regions between genes, and intronic miRNAs are found in the introns of genes [19]. Intronic miRNAs are co-expressed with the genes in which they are located and are regulated from the same promoters as their host genes [19]. Initially, the primary miRNA is transcribed in the nucleus, is modified by the RNAase III endonuclease Drosha, and subsequently forms a hairpin-like precursor miRNA (pre-miRNA) [20]. Pre-miRNAs are transported to the cytoplasm by exportin 5, where they are further modified into their mature form by dicer. The mature miRNA combines with the RNA-induced silencing complex (RISC) and suppresses its target mRNAs by binding the 3' untranslated region  of the target genes. This binding leads to the suppression of translation and/or the degradation of the mRNA [21].
A disintegrin and metalloprotease 9 (ADAM9) is a member of the ADAM family of type I transmembrane proteins and plays an important role in the regulation of the cell-cell and cell-matrix interactions that are critical determinants of malignancy. The disintegrin domain of ADAM9 adheres to cells by binding to integrins [22], and the metalloprotease domain functions by releasing a variety of cell surface proteins, such as growth factors, cytokines, cell adhesion molecules, and receptors [23]. Overexpression of ADAM9 has been observed in many cancers [24] and is correlated with brain metastasis [8]. However, the molecular mechanism underlying this association is not clearly understood.
In the current study, we aimed to better understand the relationship between CDH2 and ADAM9 in lung cancer brain metastasis. We hypothesized that miRNAs may play a role in ADAM9-CDH2 regulation, and we identified several differentially expressed miRNAs in aggressive lung adenocarcinoma using miRNA microarrays. We further demonstrated that ADAM9 could inhibit the expression of miR-218 and its precursor pri-miR-218-1 and could, in turn, up-regulate the expression of CDH2 to increase the mobility of lung adenocarcinoma cells.

Cell culture
Several human lung adenocarcinoma cell lines were used, including A549, H1299, CL1-0, F4, and BM7. A549 and H1299 cells were obtained from Bioresource Collection and Research Center (Hsinchu, Taiwan). BM7 cell line was a brain-metastatic clone derived from a high metastatic subline F4, which had higher invasion capability than its parental cell line CL1-0. CL1-0 cells were a gift from Dr. Pan-Chyr Yang (National Taiwan University, Taipei, Taiwan) [25]. F4 cells with stable high level luciferase expression were established as previously described [26].

Illumina human v2 microRNA expression beadchip and data analysis
Cells were flash frozen in liquid N 2 and stored at 280uC until RNA extraction. Total RNA was extracted using TRIZOL Reagent (Ambion, Carlsbad, CA, USA). The RNA concentration and quality were determined using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE) and an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA), which was used to calculate an RNA integrity number (RIN). Total RNA with an A260/A280 between 1.7 and 2.1 and a RIN .7.0 was adjusted to 40-200 ng/ml with DEPC-treated H 2 O. A total of 1 mg of RNA was used for the microRNA assay. Input RNA was polyadenylated and converted into cDNA using standard methods. A single miRNA-specific oligo (MSO) was used to assay each miRNA on the panel. All MSOs were hybridized to the sample in parallel, and a solid-phase primer extension step further increased the specificity and reduced the noise. After eluting the extended products and performing PCR with fluorescently labeled universal primers, the double-stranded PCR products were bound to a solid phase, and the labeled, single-stranded PCR products were prepared for Human v2 microRNA expression beadchip hybridization (Illumina, San Diego, CA). After 14-20 hours of hybridization, the beadchip was washed and coated with xylene solution. The intensities of the bead fluorescence were determined using the Illumina BeadArray Reader, and the results were analyzed using GenomeStudio v2010.1 software. The microarray data in this study are MIAME compliant [27] and have been submitted to the Gene Expression Omnibus (GEO) database (accession number GSE51666).
Quantile normalization was performed using Partek Genomics software (Partek, St. Louis, MO, USA). MiRNAs were selected when their expression change was greater than 2-fold in the three miRNA microarrays. The array results from the brain metastatic lung adenocarcinoma cells were compared to the results from the parental F4 cell line.
Luc-CDH2 vector. The CDH2 39-UTR was amplified by PCR from genomic DNA isolated from human blood. The pMIR-CDH2-39UTR construct was digested with SpeI and MluI, and the generated fragment was inserted into the SpeI-MluI sites of the pMIR-REPORT miRNA Expression Reporter Vector (Applied Biosystems, Carlsbad, CA, USA). Three miR-218 binding sites in the CDH2 39-UTR were predicted using miRSystem [29], and these sites were located at 2,671-2,691 bp, 2,740-2,760 bp, and 3,571-3,591 bp relative to the transcription start site. Mutations were made in the miR-218 binding sites in the CDH2 39-UTR using the QuikChange Site-Directed Mutagenesis Kit (Agilent Technologies, Santa Clara, CA, USA) according to the manufacturer's protocol.
Cell transfection. BM7 and H1299 cells were seeded in antibiotic-free medium at 70-80% confluence. The cells were transfected with using Lipofectamine LTX with Plus Reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's instructions.
shRNA-mediated gene silencing of ADAM9 HEK293T packaging cells (ATCC # CRL-11268) were cultured in high-glucose DMEM supplemented with 10% FBS. HEK293T were transfected using Turbofect (Thermo Scientific) according to the manufacturer's instructions. The specific lentiviral shRNA constructs targeted against ADAM9 were obtained from the National RNAi Core Facility in Taiwan. The target sequences for ADAM9 were shADAM9-C (59-GCCAGAATAACAAAGCC-TATT-39) and shADAM9-E (59-CCCAGAGAAGTTCCTATA-TAT-39). Lentivirus was packaged in HEK293T cells following the guidelines of the National RNAi Core Facility (http://rnai. genmed.sinica.edu.tw/protocols), and the culture supernatants containing the lentivirus were collected at 48 and 72 h posttransfection. BM7 cells were infected with the lentiviruses overnight in the presence of 8 mg/ml polybrene (Sigma) and were cultured in fresh medium for an additional 24 h. The infected cells were then selected in medium containing 0.4 mg/ml puromycin until the uninfected cells were completely dead.

Luciferase reporter assay
HEK293 cells were co-transfected with 300 ng of miRNA, 100 ng of the reporter vector containing the CDH2 39-UTR or the mutant CDH2 39-UTR, and 25 ng of the Renilla luciferase vector as an internal control. After 48 h, the cells were collected, and the luciferase activities were measured using the Dual-Luciferase Reporter Assay System (Promega, Madison, WI, USA).

Cell migration assay
Migration assays were performed using 24-well transwell migration chambers (Corning, Corning, New York, USA) with polyethylene membranes (8 mm pore size). The upper chambers were seeded with 5610 4 cells/well in 200 ml of serum-free DF12 or RPMI medium, and the lower chambers were filled with 600 ml of complete medium, which was used as a chemoattractant. The cells were allowed to migrate for 24 h at 37uC. Following incubation, the medium in the upper and lower chambers was removed by aspiration. A methanol-acetic acid (1:3) mixture was added into the lower chamber to fix the cells. After incubation at room temperature for 20 min, the inserts were washed twice with ddH 2 O. After the well was dried, 0.1% crystal violet (upper: 150 ml; lower: 650 ml) was added, and the inserts were incubated for 20 min at room temperature. After two washes with ddH 2 O, 200 ml of destaining solution was added into the lower chamber of each well to destain the membrane, and the wells were read at an emission wavelength of 570 nm.
For the time-lapse migration assay, BM7 cells with stable, tetracycline-inducible miR-218 expression were cultured on 6-cm dishes coated with collagen (10 mg/ml, 3 ml) and were treated with 20 mg/ml tetracycline for four days. After tetracycline induction, cell movements were monitored using inverted microscopes (Axio Observer Z1, Zeiss, Jena, Germany) with CCD video cameras (AxioCam MRm, Zeiss) at 20 min intervals for a total of 16 h in a 37uC chamber. The accumulated distance was determined by tracking the positions of cell nuclei using the Track Point function of ImageJ.

ADAM9 activated CDH2 in aggressive lung adenocarcinoma cells
To understand whether the expression of ADAM9 and CDH2 were correlated with the malignancy of lung adenocarcinoma, we detected the endogenous expression levels of ADAM9 and CDH2 using real-time PCR and western blot analyses. Brain-metastatic cell line BM7 and H1299 cells [30] are more aggressive cell lines with high migration ability, whereas the CL1-0 and A549 cell lines were used as controls. As shown in Fig. 1A and 1B, the RNA and protein levels of CDH2 in the BM7 cells were up-regulated compared with their levels in CL1-0 cells. Similarly, the RNA and protein levels of CDH2 were more abundant in another lung adenocarcinoma cell line, H1299, compared with A549 cells (Fig.  S1 A & B). The amounts of both the long and short forms of ADAM9 were also increased in the aggressive cell lines, including BM7 (Fig. 1B) and H1299 (Fig. S1B).
Next, we assessed whether the expression of CDH2 changed when the levels of ADAM9 were altered. First, we used shRNA to knock down ADAM9 in BM7 cells, and two primer sets were used to measure the expression of CDH2. As shown in Fig. 1C, the expression of CDH2 was significantly down-regulated by both shADAM9 constructs (shADAM9-C & shADAM9-E). The protein levels of ADAM9 and CDH2 also decreased when ADAM9 was knocked down, according to western blot (Fig. 1D) and immunohistochemistry (Fig. 1E) analyses. The amount of CDH1 (E-cadherin) and VIM (vimentin) protein did not change (Fig. 1D). Furthermore, to confirm the relationship between CDH2 and ADAM9, the expression of CDH2 was measured in BM7 that overexpressed ADAM9. The amount of CDH2 increased in the ADAM9-expressing cells (Fig. 1F). These results indicated that ADAM9 is able to activate CDH2 in aggressive lung adenocarcinoma cells.

Identification of the differentially expressed miRNAs in aggressive lung adenocarcinoma cells
To investigate which miRNAs could regulate CDH2 expression in brain metastatic lung cancer cells, we examined the miRNA expression profiles in these cells and their parental cells using an Illumina miRNA microarray. The endogenous expression levels of all miRNAs were examined, and 146 miRNAs were determined to have a .2-fold change in expression in the brain metastatic lung cancer cells. Furthermore, we used several algorithms in the miRSystem program [29] to predict which miRNA targeted CDH2. The program uncovered 44 miRNAs that were predicted to target CDH2. In total, we identified nine miRNAs that both targeted CDH2 and showed significant expression changes between the highly metastatic cells and their parental cells ( Fig. 2A). Of these miRNAs, four were down-regulated and five were up-regulated in the brain metastatic lung cancer cells (Fig. 2B).
Because CDH2 was up-regulated in the BM7 cells and miRNAs down-regulate their target genes, we focused on the miRNAs that were down-regulated in the BM7 cells. Of these down-regulated miRNAs, six computational algorithms [29], including DIANA, miRanda, miRBridge, PicTar, rna22, and TargetScan, predicted that miR-218 was the most likely to target CDH2. Therefore, we focused on miR-218 for further experiments. We first compared the endogenous expression levels of miR-218 in several lung cancer cell lines (Fig. S2). The results of quantitative RT-PCR validated the down-regulation of miR-218 in the aggressive lung cancer cells, including the BM7 (Fig. 2C) and H1299 (2D) cell lines, compared with their control lines, CL1-0 and A549.

MiR-218 directly regulated CDH2 in aggressive lung adenocarcinoma cells
To identify whether miR-218 can bind and regulate CDH2, we first used computational algorithms to predict the potential binding sites in the CDH2 39-UTR and examined their interaction using luciferase assays. The locations of the potential binding sites were 2,671-2,691 bp, 2,740-2,760 bp, and 3,571-3,591 bp relative to the transcription start site of CDH2 (Fig. 4A). Because the seed region of the miRNA, which includes 2 to 8 nucleotides at the 59-end of the miRNA [21], must be complementary to the 3'-UTR of the target genes, we mutated these binding sites to evaluate which binding sites played important roles (Fig. 4A). By cotransfecting the miR-218 plasmids and the reporter construct, which contained the CDH2 39-UTR behind the luciferase gene (Fig. 4A), we showed that miR-218 was better able to inhibit the luciferase activity compared with the miR-empty vector control (Fig. 4B). When we mutated all the binding sites, the luciferase activity was recovered. Mutation of site A or site C alone, but not site B alone, could relieve the suppression of luciferase activity (Fig. 4B). This result suggested that site B was not a binding site for miR-218. Therefore, we showed that miR-218 can bind to the 39-UTR of CDH2 at two binding sites.
To further confirm that CDH2 could be inhibited by miR-218, we over-expressed miR-218 in metastatic BM7 cells. Real-time PCR showed that miR-218 was significantly up-regulated at 48 h after transfection (Fig. 4C), and the relative mRNA levels of CDH2 were decreased 0.6-fold in the BM7 cells (Fig. 4D). Western blot analysis also showed that the protein levels of CDH2 were decreased (0.54-fold) following over-expression of miR-218 in BM7 cells (Fig. 4E). Similarly, administration of miR-218 mimic oligonucleotides in BM7 cells resulted in decreased CDH2 expression (Fig. 4F). Furthermore, we over-expressed miR-218 in another lung adenocarcinoma cell line, H1299 (Fig. S3A), and we found that CDH2 was also down-regulated both at the RNA and protein levels ( Fig. S3B & C). To confirm this regulation, we further used miR-218 inhibitors to block the levels of miR-218 in lung cancer cells F4 and A549. Real-time PCR showed that miR-218 was significantly decreased at 48 h after transfection (Fig. 4G), and the relative mRNA levels of CDH2 were increased in these cells (Fig. 4H). These results indicate that miR-218 can downregulate CDH2 in aggressive lung adenocarcinoma cells.

MiR-218 inhibited the migration ability of aggressive lung adenocarcinoma cells
Previous reports showed that CDH2 was up-regulated in metastatic cells and induced cell migration [12]. Therefore, we evaluated whether miR-218 could suppress cell migration by targeting CDH2. After transfection of miR-218 in both BM7 and H1299 cells, we measured cell migration using transwell migration assays. As shown in Fig. 5A and 5C, the number of migrated cells in the group over-expressing miR-218 was decreased. We quantitated the cell migration ability by detecting the dye used to stain the migrated cells. As shown in Fig. 5B and 5D, the relative cell migration was decreased 0.2-fold in the BM7 cells and 0.3-fold in the H1299 cells. Furthermore, using a tet-on construct to over-express miR-218 in the presence of tetracycline (Fig. S4A  & B), we also observed that cell mobility was significantly (P,0.01) decreased in the BM7 cells over-expressing miR-218 (Fig. 5E & F).
In contrast, blocking miR-218 expression with miR-218 inhibitors in F4 and A549 cells, the migration ability was significantly enhanced in miR-218 inhibitor group compared to negative control (NC) group in F4 (Fig. 5G & H) and A549 cells (Fig. 5I &  J). These results indicate that miR-218 can inhibit cell migration by repressing the expression of CDH2.

Discussion
In this study, we demonstrated that endogenous ADAM9 expression was significantly up-regulated in aggressive lung adenocarcinoma cells, and ADAM9 could activate the expression of CDH2. Down-regulation of miR-218, which resulted from low transcription of pri-mir-218-1, led to CDH2 over-expression in aggressive lung cancer cells. Thus, over-expression of miR-218 could inhibit CDH2 expression and tumor cell mobility. Here, we illustrate the mechanism by which ADAM9 activates CDH2, which may be due to the release of miR-218 inhibition of CDH2.
Previously, miR-218 was mostly regarded as a tumor suppressor in many cancers. For example, miR-218 could inhibit migration, invasion, and proliferation of glioma cells [32], head and neck squamous cell carcinoma cells [33], cervical squamous cell carcinoma cells, nasopharyngeal cancer cells [34], and gastric cancer cells [35]. MicroRNA-218 could also inhibit cell cycle progression, promote apoptosis in colon cancer [36], and increase the chemosensitivity of cervical cancer cells to cisplatin [37]. In primary non-small cell lung cancer, miR-218 was deleted or downregulated, and its expression could be used to predict survival and relapse [38]. When miR-218 expression was low in lung cancer patients, their clinical outcomes were poor [38]. Our findings were consistent with these previous results, thus confirming the tumor suppressor role of miR-218. In contrast, only one study reported that miR-218 was a potent activator of Wnt signaling, contributed to osteoblastogenesis, and facilitated the metastasis of breast cancer cells into the bone [39].
Several targets of miR-218 have been reported, including BMI1 [36], PXN [38], BIRC5 [34], GJA1 [34], laminin-332 [33], and ROBO1 [34,35]. In particular, the miRNA-218 and ROBO1 signaling axis has been studied extensively and correlates with metastasis and vascular patterning in pancreatic and nasopharyngeal cancers [40,41]. In this study, we demonstrated that miR-218 can directly bind to the 39-UTR of CDH2 at two binding sites (2,671-2,691 bp and 3,571-3,591 bp) using luciferase reporter assays. Interestingly, the binding site at 3,571-3,591 bp has also been reported in bovine cells [42], which supports our finding that miR-218 targets CDH2. Furthermore, over-expressing miR-218 by In this study, we observed low expression levels of miR-218 in the aggressive lung cancer cell lines BM7 and H1299 ( Fig. 2A &  2B). We further explored this down-regulation by examining the expression of the precursor miRNAs of miR-218. The miR-218 transcripts are located within the introns of SLIT2 (pri-mir-218-1) and SLIT3 (pri-mir-218-2), which were reported to function as tumor suppressors [43]. The expression levels of SLIT2, SLIT3, pri-mir-218-1, and pri-mir-218-2 were detected using real-time PCR. We found that the down-regulation of miR-218 in lung adenocarcinoma cells was related to the expression of SLIT2. Hyper-methylation of the CpG-islands in SLIT2 [44] and copy number losses of SLIT2 have been reported [45]. Additionally, SLIT2 could suppress cell migration through the regulation of beta-catenin [46], the AKT-GSK3b signaling pathway [47], and the ROBO1 signaling pathway [34]. However, in gastric cancer and thyroid cancer, it was shown that down-regulation of miR-218 was attributed to low expression levels of SLIT3 [31,34], and restoring the expression of miR-218-2 and SLIT3 could repress cell  invasion and migration [31]. The difference between lung cancer and gastric cancer may be due to the tissue specificity of the miRNA precursors that result in mature miR-218.
ADAM9 has two isoforms, including a shorter ADAM9secreted (ADAM9-S) transcript and a transmembrane protein, ADAM9-long (ADAM9-L). ADAM9 is typically regarded as oncogene in many cancers, such as oral squamous cell carcinomas [48], breast tumors [49], prostate cancer [50], and renal cell cancer [51]. Inhibition of ADAM9 expression can sensitize prostate cancer cells to radiation and chemotherapy [50]. However, the ADAM9 splice variants have opposing effects on breast cancer cell migration [52]. ADAM9-S promoted breast cancer cell migration, whereas ADAM9-L suppressed cell migration. Therefore, a key determinant in the manifestation of aggressive migratory phenotypes is the relative levels of the membrane-tethered and secreted variants of ADAM9. In our results, the relative ratio of the short form to the long form was higher in the BM7 cell line compared with the CL1-0 cell line, which corresponded to the aggressiveness of BM7. Moreover, we found that down-regulation of ADAM9 could up-regulate SLIT2. However, there is no direct evidence indicating that ADAM9 can regulate SLIT2; thus, more experiments are needed to explore this relationship.
In conclusion, brain metastasis of lung cancer is one of the main reasons for the high mortality of this disease. MicroRNAs have been reported to modulate tumor metastasis. We demonstrated that down-regulation of miR-218 was attributed to low expression of its host gene, SLIT2, and its precursor, pri-mir-218-1. Although there was no direct evidence that ADAM9 regulates SLIT2, the down-regulation of ADAM9 resulted in the up-regulation of SLIT2 and miR-218, which in turn down-regulated CDH2 (Fig. 6). Overall, this study increases our understanding of how lung cancer cells metastasize to the brain and may result in the development of new therapeutic strategies for lung cancer.