MiR-134/487b/655 Cluster Regulates TGF-β–Induced Epithelial–Mesenchymal Transition and Drug Resistance to Gefitinib by Targeting MAGI2 in Lung Adenocarcinoma Cells

Lung cancer was reported to be the leading cause of cancer-related death worldwide in males, and the second leading cause in females (1). Molecular-targeted therapies have been developed recently for non–small cell lung cancer (NSCLC). Thus, patients with NSCLC with epidermal growth factor receptor (EGFR) gene mutations have shown a dramatic response to EGFR tyrosine kinase inhibitors (EGFR-TKI) such as gefitinib and erlotinib (2, 3). Our group and others have recently reported that first-line gefitinib treatment in patients with advanced NSCLC with EGFR mutations improved progression-free survival in randomized phase III studies (4–6). Therefore, gefitinib has been recognized as a first-line treatment of patients with NSCLC with EGFR activating mutations. These results represent a milestone toward personalized medicine in NSCLC oncology. Despite the advances in chemotherapy and molecular-targeted therapy, the prognosis of patient with advanced NSCLC is still poor because of metastasis and drug resistance.

Epithelial–mesenchymal transition (EMT), especially that induced by transforming growth factor β1 (TGF-β1), is a progressive biological phenomenon that includes loss of epithelial cell adhesion and induction of a mesenchymal phenotype (7). Previous studies demonstrated that EMT was involved in cell invasion, migration, and metastasis in several types of cancer, including NSCLC (8, 9). Several studies have demonstrated that EMT was associated with reduction of drug sensitivity and acquisition of resistance to EGFR-TKIs in NSCLC, whereas retention of an epithelial phenotype ensured a good response to EGFR-TKIs even in patients with tumors harboring wild-type EGFR genes (10–12). These reports suggested that EMT might be a mechanism of resistance to EGFR-TKI regardless of EGFR status.

MicroRNAs (miRNA) are single-strand 18–24nt noncoding molecules that post-transcriptionally modulate gene expression through binding to 3′ untranslated regions (UTR) of target mRNAs (13). miRNAs, which usually induce gene silencing, can function as either tumor suppressors or oncogenes (14). Recent studies revealed that miRNAs were diagnostic, prognostic, and therapeutic biomarkers in lung cancer (15, 16). Our previous study demonstrated that inhibition of miR-21 expression could be used as a therapeutic strategy in connection with EGFR-TKI treatment (17). Several miRNAs have been also identified as regulators of EMT in cancer. EMT was suppressed by miR-200 family members, targeting the transcriptional activators of EMT, ZEB1, and ZEB2 in lung cancer (18, 19). Despite previous reports of miRNAs associated with EMT and cancer, miRNAs and their target genes involved in TGF-β1–induced EMT, resulting in resistance to EGFR-TKIs, are still not fully understood.

In this study, we analyzed miRNA changes to clarify which miRNA was associated with TGF-β–induced EMT followed by acquired resistance to EGFR-TKI. We used four lung adenocarcinoma cell lines with or without the TGF-β–induced EMT phenomenon and demonstrated that the miR-134/487b/655 cluster was associated with TGF-β–induced EMT and played a critical role in EMT through the targeting of membrane-associated guanylate kinase, WW, and PDZ domain–containing protein 2 (MAGI2).

Eight lung adenocarcinoma cell lines, A549, LC2/ad, PC3, PC9, RERF-LCKJ, RERF-LCMS, PC14, and ABC-1, were used in this study. A549, LC2/ad, and RERF-LCKJ were obtained from the RIKEN Cell Bank; PC3, PC9, and PC14 were obtained from Immuno-Biological laboratories; and RERF-LCMS and ABC-1 were obtained from the Health Science Research Resources Bank. These cell lines have been obtained from 2003 to 2011, were amplified and frozen, and one aliquot of each was thawed for this project, although no authentication was done by the authors. All cells were routinely screened for the absence of mycoplasma. These cell lines were maintained in RPMI 1640 medium (GIBCO) supplemented with 10% fetal bovine serum (FBS).

Total RNA was extracted from the four lines with TRIzol reagent (Invitrogen) as described previously (20, 21). Concisely, 5 μg of total RNA was used for hybridization on miRNA microarray chips containing 667 probes with the TaqMan Array Human MicroRNA A + B Cards Set v2.0 (Life Technologies) on a 7900 Real Time PCR System (Applied Biosystems). Processed slides were scanned with a PerkinElmer ScanArray XL5K Scanner. Experimental data were analyzed by DataAssist Software (Life Technologies) using RNU44 and RNU48 as endogenous controls. Ct values were provided from all miRNAs represented on the cards and fold changes in expression were calculated using the ΔΔC_t method. Expression levels of MammU6 on the array card were defined as positive controls for the purpose of ddCt calculation. The microarray data have been deposited in NCBIs Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo/) and are accessible through GEO Series accession number GSE51828 (release date October 30, 2013).

The MAGI2 mRNA expression and miR-134/478b/655 levels were measured by real-time quantitative reverse transcription PCR (qRT-PCR) using TaqMan Gene Expression Assay and TaqMan MicroRNA Assay (Applied Biosystems), respectively. GAPDH (glyceraldehyde-3-phosphate dehydrogenase) or RNU66 was determined as an internal control (Applied Biosystems). miRNA and MAGI2 mRNA expressions were quantified and reported by the relative standard curve method.

Cells were dissolved in buffer containing 50 mmol/L Tris-HCl, pH 7.6, 150 mmol/L NaCl, 0.1% sodium dodecyl sulfate, 1% Nonidet P-40, and 0.5% sodium-deoxycholate. The lysates were cooled in ice for 30 minutes, and then centrifuged at 13,000× g for 30 minutes. Each protein (10 μg) of the collected supernatant was separated by SDS-PAGE on 12% gels and then transferred to a nitrocellulose membrane. The membrane, after a block with 5% skimmed milk, was incubated with antibodies to E-cadherin, N-cadherin, vimentin, β-actin, PTEN, p-PTEN (Ser 380; Cell Signaling Technology), and MAGI2 (AIP1; Abcam Biochemicals). Each protein was detected by immunoblotting with ECL-Plus reagents (GE Healthcare Bio-Science Corp).

TGF-β1 was purchased from R&D system. Cells were exposed to 5 ng/mL TGF-β for the indicated time. siRNA targeting MAGI2 and negative control were purchased from Ambion and Invitrogen, respectively. miR-134 and miR-487b precursors (pre-miR-134 and pre-miR-487b), their negative control (pre-NC), miR-134 and miR-487b inhibitors (anti-miR-134/487b) and their cognate negative control (anti-NC) were synthesized by Ambion. All precursors and inhibitors were treated with lipofectamin 2000 transfection reagent 24 hours after seeding, according to the manufacturer's instructions (Life Technologies). The precursor and inhibitor complexes were transfected into cells at a final concentration of 40 nmol/L. The final concentration of miRNA was determined according to the recommend dose of lipofectamin 2000 transfection protocol (Invitrogen). The transfection medium was replaced 6 hours later. After the transfection at 24 hours, 5 ng/mL TGF-β1 was added to the medium, which was then incubated at 37°C for 48 hours.

Gefitinib was purchased from Seleck Chemicals. A549 cells (5,000 cells/well) were seeded into 96-well plates for 24 hours. The cells were incubated in the various concentrations of gefitinib for 72 hours at 37°C after exposure to pre-miR-134, pre-miR-487b, or pre-miR-ctl at a final concentration of 40 nmol/L for 24 hours. Likewise, the cells with both anti-miR-134 and anti-miR-487b, or anti-miR-ctl were exposed to 5 ng/mL TGF-β1 for 48 hours, and then received the gefitinib treatment. After the incubation in various concentrations of gefitinib, each well was transfected with 20 μL of MTT solution (5 mg/mL in PBS), then continuously maintained for a further 4 hours at 37°C as previously described (22, 23). Finally, the absorbance was measured using a microplate reader with a test wavelength of 560 nm.

Luciferase reporter constructs containing portions of the MAGI2 3′UTRs were generated by GeneCopoeia, Inc.. A549 cells were cultured in 24-well plates for 24 hours and cotransfected with 800 ng/uL of MAGI2 3′UTR reporter constructs and 50 nmol/L of pre-miR-134 or pre-miR-487b, or pre-NC using lipofectamine 2000 for 24 hours. After transfection, cells were harvested, lysed, and assayed with the Dual-Luciferase Reporter Assay Kit (Promega) according to the manufacturer's instructions. Firefly luciferase activity was normalized to renilla luciferase activity for each transfected well. Each experiment was performed in duplicate and repeated three times.

Statistical analysis was performed using SigmaPlot software (version 12.0; Systat Software, Inc.). Data were expressed as a mean (SD) of three independent experiments and evaluated with the Student t test.

Four lung adenocarcinoma cell lines, A549, LC/ad2, PC3, and PC9, were stimulated by TGF-β1. Figure 1A shows the conversion from epithelial cells to mesenchymal cells, characterized by spindle-type cell morphology in A549 and LC2/ad cells. On the other hand, PC3 and PC9 cells did not show this morphological change (Fig. 1B). We evaluated the expression levels of EMT markers in A549 and LC2/ad cells treated with TGF-β1 to confirm the EMT phenomenon. A549 and LC2/ad cells after TGF-β1 stimulation displayed reduced epithelial marker E-cadherin and increased mesenchymal cell marker N-cadherin expressions, when compared with PC3 and PC9 cells without TGF-β1 exposure (Fig. 1C). The time courses of decreased expression of E-cadherin and increased expression of N-cadherin, through TGF-β1 transfection, were observed in A549 and LC2/ad cells (Fig. 1D). A549 and LC2/ad cells were therefore recognized as lung cancer cells with the TGF-β1–induced EMT phenomenon. Next, we evaluated miRNA expression profiles in EMT-positive lines (A549 and LC2/ad) and EMT-negative lines (PC3 and PC9) before and after exposure to TGF-β1 using microRNA arrays. After the TGF-β1 treatments, the expression levels of 11 miRNAs in EMT-positive cells were increased significantly more than the expression levels in EMT-negative cell (Fig. 2A). The miRBase database (http://www.mirbase.org) indicated that among the 11 miRNAs, miR-134, miR-487b, and miR-655 belonged to the same miRNA cluster on chromosome 14q32. The expression of these three miRNAs before and after exposure to TGF-β1 was validated by qRT-PCR (Fig. 2B–D). In contrast, the expression levels of these three miRNA in EMT-negative cells were unaffected by exposure to TGF-β1 (Fig. 2B–D). Moreover, we evaluated miR-134/478b/655 expression levels in eight lung adenocarcinoma cells without TGF-β1 stimulation, previously performed by the TaqMan MicroRNA Array (24). Previous reports demonstrated that RERF-LC-KJ and RERF-LC-MS were recognized as EMT-positive cells (25, 26). Four EMT-positive cells exhibited higher expression levels of three miRNAs than did those of four EMT-negative cells (Supplementary Fig. S1). Thus, we have identified miR-134, miR-487b, and miR-655 as candidate miRNAs associated with the EMT phenomenon in lung cancer.

We further examined whether miR-134 and miR-487b regulated TGF-β1–induced EMT. Unfortunately, the role of miR-655 on TGF-β1–induced EMT could not be evaluated because the specific miR-655 precursor or inhibitor was commercially unavailable. We confirmed that mature miR-134/487b of A549 cells were significantly increased by miR-134 or miR-487b precursors (pre-miR-134 or pre-miR-487b) and decreased by miR-134 or miR-487b inhibitors (anti-miR-134 and anti-miR-487b; Supplementary Fig. S2A and S2B). First, A549 cells were transfected with pre-miR-134 or pre-miR-487b and control precursor (pre-NC) and exposed to 5 ng/mL of TGF-β1 for 72 hours. After the TGF-β1 treatment, A549 cells transfected with pre-miR-134 or pre-miR-487b seemed to be more uniformly mesenchymal than A549 cells transfected with pre-NC (Fig. 3A). Western blot analysis showed that A549 cells treated with pre-miR-134 or pre-miR-487b had diminished E-cadherin expression, and that A549 cells treated with pre-miR-134 after the TGF-β1 stimulation had increased expression of vimentin, the mesenchymal cell marker (Fig. 3B). Next, we investigated the effect of inhibition of the two miRNAs on TGF-β1–induced EMT in A549 cells. A549 cells transfected with a combination of anti-miR-134/487b or control inhibitor (anti-NC) were exposed to TGF-β1 for 72 hours. Figure 3C demonstrates that transfection of anti-miR-134/487b inhibited the conversion from epithelial phenotype to mesenchymal phenotype after TGF-β1 stimulation. Western blot analysis also showed that E-cadherin expression was still present after the anti-miR-134/487b treatments, resulting in prevention of TGF-β1–induced EMT in A549 cells (Fig. 3D). These results demonstrated that miR-134 and miR-487b actively contributed to TGF-β1-induced EMT in A549 cells.

Our results indicated that miR-134 and miR-487b promoted TGF-β1–induced EMT. We next proceeded to identify potential targets using the microRNA.org database (http://www.microrna.org/microrna/home), a comprehensive resource of miRNA target predictions and expression profiles. We found that miR-134, miR-487b, and miR-655 belonged to the same cluster of miRNAs that commonly target MAGI2. Furthermore, MAGI2 could be predicted as a target of miR-134 and miR-655 by another database, Target Scan 5.0–5.2 (http://www.targetscan.org/). Figure 4A shows the regions of 3′UTR of the MAGI2 gene that could serve as binding sites for the three miRNAs based on the prediction of microRNA.org. To confirm whether TGF-β1 exposure diminished the expression levels of MAGI2 mRNA via upregulating these miRNAs, we evaluated MAGI2 mRNA expression by qRT-PCR in four types of lung cancer cells before and after treatment with TGF-β1. TGF-β1 stimulation decreased MAGI2 expression in the EMT-positive cells, but not in the EMT-negative cells (Fig. 4B). Next, we examined whether overexpression of miR-134 or miR-487b diminished MAGI2. The qRT-PCR and Western blot analyses showed that the pre-miR-134 or pre-miR-487b treatments induced downregulation of both mRNA and protein expression of MAGI2 in A549 cells (Fig. 4C). In contrast, expression levels of MAGI2 mRNA and protein were increased in A549 cells after treatment with anti-miR-134 and anti-miR-487b with or without TGF-β1 exposure (Fig. 4D). We also performed a luciferase reporter assay to verify that miR-134/487b directly targets MAGI2. We found that cotransfection of miR-134 or miR-487b precursor and MAGI2 3′UTR vector significantly decreased the luciferase activity in A549 cells as compared with the control (Fig. 4E). These data showed that MAGI2 is a direct target of miR-134/487b. MAGI2, a multidomain scaffolding protein, contains nine potential protein–protein interaction modules and inhibits cell migration and proliferation via PTEN in HCC cells (27). TGF-β1 stimulation or inhibition of MAGI2 was attenuated in A549 cells with PTEN phosphorylation (Fig. 4F and G). PTEN silencing played a critical role in the EMT phenomenon caused by miR-134 and miR-487b–induced MAGI-2 inhibition in A549 cells. These findings suggested that expression of miR-134 and miR-487b consequently played a key role in TGF-β1–induced EMT by targeting MAGI2, leading to PTEN inactivation in lung adenocarcinoma cells.

Finally, we evaluated whether miR-134 and miR-487b elicited resistance to EGFR-TKI. A549 cells treated with pre-miR-134 or pre-miR-487b as well as TGF-β1 were more resistant to gefitinib than were A549 parent cells (Fig. 5A). We also measured the response to gefitinib in A549 cells after treatment with miR-134 and miR-487b inhibitors in combination with TGF-β1. Anti-miR-134 plus anti-miR-487b transfection restored the resistance to gefitinib caused by TGF-β1 (Fig. 5B). Therefore, miR-134 and miR-487b activated by TGF-β1 could lead to resistance to gefitinib.

Identification of mechanisms of drug resistance to EGFR-TKIs and development of methods to overcome resistance are required for improvement of the prognosis of patients with advanced NSCLC. Approximately 50% of patients with EGFR-mutant lung cancer who develop acquired resistance to EGFR-TKI have a second-hit mutation, T790M, in exon 20 (28). Amplification of MET was observed in up to 20% of NSCLC specimens that had developed acquired resistance to EGFR-TKI (29). Other mechanisms of resistance that are operative in the remaining 30% to 40% of tumors with acquired resistance to EGFR-TKI are under active investigation. EMT has also been associated with acquired resistance to EGFR-TKI (10–12). Recent studies revealed that overexpression of AXL led to resistance to EGFR-TKI in NSCLC cells undergoing EMT, and that AXL was a potential therapeutic target in patients with acquired resistance to EGFR-TKIs (30, 31). Our present study also showed that the TGF-β1–induced EMT phenomenon encompassed drug insensitivity to gefitinib in A549 cells with wild-type EGFR genes.

Several miRNAs have been identified as regulators of EMT. miR-200 family members have been described as main suppressors of EMT by targeting of ZEB1 and ZEB2 (18, 19, 32). In breast cancer, antagonists of miR-21 reversed EMT through targeting PTEN, inactivating AKT and ERK1/2 pathways (33). Our previous study showed that miR-23a regulated TGF-β1–induced EMT by targeting E-cadherin, which encompassed resistance to gefitinib treatment in A549 cells (34). However, the contribution of these miRNAs to resistance to EGFR-TKI was still unclear. In this study, we demonstrated that activation of miR-134 and miR-487b modulated the EMT status and drug resistance to gefitinib through directly targeting MAGI2. Correlation between the 14q32 chromosome miRNA cluster and TGF-β–induced EMT has not been reported in human cancer. A recent study showed that overexpression of miR-134 in A549 cells led to promotion of cell proliferation, inhibition of cell apoptosis, and degradation of migration ability (35). Similar alterations of miR-487b and miR-655 have not been found to affect EMT. Previous studies demonstrated that genomic aberrancies located at 14q32 were observed in patients with NSCLC, however, no relationship between 14q32 miRNA cluster alterations and response to EGFR-TKI has been reported (36, 37).

Our present study highlighted a scaffold protein MAGI2, which has nine domains that are potential protein–protein interaction modules, including six PDZ domains, two WW domains, and a guanylate kinase-like domain (38). Expression of MAGI2, which belongs to a membrane-associated guanylate kinase (MAGUK) superfamily, is high in brain and low in other tissues (39). A recent study revealed that MAGI2 gene rearrangements were implicated in prostate tumorigenesis (40). Inhibition of MAGI2 led to cell migration and proliferation of human hepatocarcinoma cells (27). In our experiments, expression of MAGI2, diminished by TGF-β1, pre-miR-134, and pre-miR-487b treatments, was associated with the EMT phenomenon and acquired resistance to EGFR-TKI. Binding of PTEN to the PDZ domain of MAGI2 enhanced PTEN protein stability and the activity of the downregulating PI3K/Akt signaling cascade (41, 42). Previous studies demonstrated that degradation or loss of PTEN caused PI3K/Akt activation, elicited cell migration and proliferation, and induced the EMT phenomenon (27, 43–45). Absence of PTEN-MAGI2 binding induced PTEN phosphorylation and affected PTEN stability (41, 46). Loss or reduction of PTEN has been associated with acquired EGFR-TKI resistance in our in vitro study (47). We confirmed that TGF-β1 stimulation or silencing of MAGI2 led to PTEN phosphorylation and promotion of EMT in A549 cells. MAGI2 may also be associated with EMT by targeting other binding proteins. Previous reports showed that overexpressed MAGI2 interacted with Smad3, a TGF-β downstream molecule, and suppressed Smad3-induced transcriptional activity (48), which mediated TGF-β signaling and regulated many cellular processes (49). An inverse correlation between MAGI2 and Smad3 was reported in tumor samples (50). Taken together, our findings demonstrated that suppression of MAGI2 by the miR-134/487/655 cluster reduced PTEN activity and played an important role in the TGF-β–induced EMT phenomenon.

One limitation of our study is that we were unable to do functional experiments in vivo or on clinical samples. To detect EMT in vivo, we should view epithelial cells transitioning into fibroblasts and migrating to the interstitium in real time. Unfortunately, this is difficult with current technology. Further studies will be performed using clinical samples to clarify the association between miR-134/487/655 expression and the response to EGFR-TKI in lung adenocarcinoma.

In conclusion, the miR-134/miR-487b/miR-655 cluster promoted the TGF-β1–induced EMT phenomenon and induced resistance to EGFR-TKI in the EMT process by directly targeting MAGI2, in which suppression encompassed loss of PTEN stability. A link between MAGI2 and the miRNA cluster was identified as one of the EMT-inducing mechanisms and as a mechanism for development of acquired resistance to EGFR-TKI in lung adenocarcinoma cells without EGFR mutant genes. Our data consequently implied that this link could be a new therapeutic target in patients with advanced lung adenocarcinoma cells undergoing EMT. Further studies should be undertaken to clarify the mechanism of the connection between the miR-134, miR-487b, and miR-655 cluster and EMT in lung adenocarcinoma.

Kaoru Kubota has honoraria from speakers' bureau from AstraZeneca. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K. Kitamura, M. Seike, T. Okano, H. Mizutani, A. Gemma

Development of methodology: K. Kitamura, T. Okano, H. Mizutani

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Kitamura, T. Okano, K. Matsuda, A. Miyanaga, R. Noro

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Kitamura, T. Okano, K. Matsuda

Writing, review, and/or revision of the manuscript: K. Kitamura, M. Seike, A. Gemma

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Minegishi, A. Gemma

Study supervision: M. Seike, K. Kubota

The authors thank C. Soeno and M. Fujishiro of Nippon Medical School for technical assistance.

This study was supported in part by a grant-in-aid from the Ministry of Education, Culture, Sports, Science, and Technology of Japan (grant no. 24591179 to M. Seike; grant no. 25461172 to A. Gemma), MEXT-Supported Program for the Strategic Research Foundation at Private Universities (to M. Seike and A. Gemma), and Smoking Research Foundation (to A. Gemma).

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.