miR340 Suppresses the Stem-like Cell Function of Glioma-Initiating Cells by Targeting Tissue Plasminogen Activator

Glioblastoma multiforme (GBM) is the most common and aggressive malignant brain tumor. Despite the most intensive current therapeutic efforts, the median survival time (∼14 months) has not changed significantly in decades (1). The extremely poor prognosis of patients with GBM is likely because of the presence of glioma-initiating cells (GIC; also known as glioma stem-like cells), which are equivalent of the cancer stem cells (CSC) observed in other cancer types (2–4). CSCs are not only highly resistant to chemo- and radiotherapy (5–7) but are also highly tumorigenic; thus, these cells are thought to be the primary cause of tumor recurrence and progression. To improve the poor prognosis of patients with GBM, it is important to understand the mechanism that GICs are activated to have high tumorigenesis and invasion.

miRNAs (miR) are small, noncoding RNAs of approximately 22 nucleotides in length that mediate the posttranscriptional silencing of specific target mRNAs and that are currently recognized as important regulators of tumorigenesis and development (8). Many miRNAs have been reported to be aberrantly expressed in malignant gliomas and play a role in determining the degree of malignancy. These include miR21, miR221/222, miR124, and miR128 (9–12). These miRNAs can act as either oncogenes or tumor suppressors in gliomas, depending on their effects on cell proliferation and apoptosis. However, there are few studies on integrated functions of miRNAs that significantly regulate tumorigenesis and tumor development initiated by GICs.

In the present study, we comprehensively analyzed the miRNA expression profiles of human GICs (hGIC) and identified miR340 as a novel miRNA that is significantly downregulated in hGICs compared with human neural stem cells (NSC). Functional analyses revealed that miR340 suppressed hGIC proliferation, invasion, and migration in vitro, as well as hGIC tumorigenesis in vivo, in the mouse brain. Furthermore, we defined tissue plasminogen activator (PLAT) gene as a direct target of miR340. These results indicate that GICs with the decreased level of miR340 promote glioblastoma formation in the mouse brain and that miR340 downregulation can induce a variety of malignant processes, such as cell proliferation and diffuse invasion. We also demonstrated that these effects are primarily mediated by increased levels of the target molecule PLAT. These findings reveal miR340 and its target gene PLAT as potentially useful therapeutic candidates for the treatment of glioblastoma.

The mice were obtained from CLEA Japan, Inc. The mouse experiments were performed according to protocols approved by the Animal Care and Use Committees of Ehime University (Ehime, Japan) and of Hokkaido University (Hokkaido, Japan). Chemicals and growth factors were purchased from Invitrogen and PeproTech, respectively, except where otherwise indicated.

Seven primary human glioma samples [five GBMs (E1-4, 6), one anaplastic oligodendroglioma (AO), and one diffuse astrocytoma (DA)] were used to prepare the hGICs. These seven samples were obtained from Ehime University Hospital with the patients' consent according to the Research Ethics Committee guidelines and were used in compliance with the research guidelines of the Ehime University Graduate School of Medicine and the Institute for Genetic Medicine of Hokkaido University. Tumor samples were dissociated using a papain dissociation system (Worthington) according to the manufacturer's instructions. The dissociated cells were cultured to form tumor spheres in serum-free DMEM/Ham's F-12 (Wako) supplemented with human basic fibroblast growth factor (bFGF; 10 nmol/L), human EGF (10 nmol/L), heparin (5 μmol/L), N2 supplement (Wako), 10 μg/mL insulin (Wako), GlutaMAX supplement, 100 U/mL penicillin G, and 100 μg/mL streptomycin. For IHC, GICs were cultured on poly-D-lysine (15 μg/mL, Sigma)-coated and fibronectin (1 μg/mL)-coated 8-well chamber slides (Nunc). The human glioma cell lines U251 and U87 (obtained from ATCC) were cultured in DMEM with 10% FBS (Thermo Scientific). The mouse GIC lines NSCL61 and OPCL61, that have been established by overexpression of oncogenic HRas^L61 in p53-deficient mouse NSC and oligodendrocyte precursor cell, were cultured as described previously (13, 14). Human NSCs (H9 human embryonic stem cell derived; Invitrogen) were cultured according to the supplier's instructions. The cells were maintained at 37°C under a humidified 5% CO₂/95% air atmosphere for all of the experiments. The characterization of human GICs is presented in the Supplementary Fig. S1.

Total RNA was extracted from GICs, glioma cells, and glioma tissues using the TRIzol Plus RNA Purification System (Invitrogen). The miRNA microarrays were manufactured by Agilent Technologies, and 100 ng total RNA was hybridized using the miRNA Microarray Kit protocol for use with Human miRNA Microarray Release 16.0 or Mouse miRNA Microarray Release 16.0. Hybridization signals were detected using a DNA microarray scanner (Agilent Technologies), and the scanned images were analyzed using Agilent Feature Extraction software.

For the gene expression analyses, total RNA was amplified and labeled with cy3 using a one-color Agilent Low Input Quick Amp Labeling Kit (Agilent Technologies) according to the manufacturer's instructions. Labeled cRNAs were fragmented and hybridized to the Agilent Human GE 8×60K Microarray. After washing, the microarrays were scanned using a DNA microarray scanner. Intensity values for each scanned feature were quantified using Agilent Feature Extraction software, which also performs background subtraction corrections.

Data normalization was conducted using Agilent GeneSpring GX version 11.0.2 software. After normalization, hierarchical sample clustering of the expressed genes was performed using the Euclidean distance and average linkage methods (Agilent GeneSpring GX). We used the freely available GenMAPP 2.1 database (http://www.genmapp.org/) and ConPath Navigator software, in combination with custom software programs, for the pathway analysis.

The microarray data have been submitted to NCBI GEO and available under GSE61078, GSE61079, and GSE61080.

Total RNA was extracted from GICs, glioma cells, and glioblastoma and normal brain tissues using Isogen reagent (Nippon Gene). Then, total RNA was reverse-transcribed using MultiScribe Reverse Transcriptase (Applied Biosystems) and MMLV RT (Invitrogen) according to the suppliers' instructions. miR340 expression was analyzed using TaqMan small RNA assays (Applied Biosystems) and a MiniOpticon RT-PCR System (Bio-Rad) according to the suppliers' instructions. PLAT expression was analyzed using FastStart Universal SYBR Green (Roche) with a MiniOpticon RT-PCR System. The PCR conditions were as follows: 10 minutes at 95°C; followed by 40 cycles of 95°C for 15 seconds and 60°C for 1 minute. All of the expression values were normalized against β-actin mRNA expression levels. The RT-PCR primers for PLAT and β-actin are listed in Supplementary Table S1.

Either the precursor form of miR340 or a control miRNA was overexpressed in GICs and glioma cells using miRNA lentiviral particles (pEZX-MR03; GeneCopoeia) according to the supplier's instructions. In certain experiments, cells were transfected with a miR340 expression vector (pBApoCMV-Neo-miR340, Takara) using a Nucleofector device (Lonza) and cultured in the presence of neomycin (300 μg/mL; Sigma) for 10 days.

GICs and glioma cells (1×10⁴ cells/well) were transfected with either a control vector or a miR340 expression vector and cultured for 72 hours. Cells were collected on days 1, 2, and 3 and counted using a hemocytometer. Bromodeoxyuridine (BrdUrd) staining was performed as previously described (15). Cell-cycle analysis was performed using propidium iodide staining and FACS analysis (BD FACSCalibur, Cycletest Plus DNA Reagent Kit, BD Biosciences) according to the supplier's instructions.

Cell invasion and migration were assessed using a BioCoat Matrigel Invasion Chamber (Becton-Dickinson) and the BioCoat Cell Culture Inserts (Becton-Dickinson), respectively. Briefly, transfected GICs and glioma cells (5–10×10⁴) were resuspended in DMEM/Ham's F-12 medium, which was supplemented with 1 mg/mL BSA (Sigma), and transferred to the upper chamber of each well. Then, medium containing 10% FBS was added to the lower chamber. After incubation for 12 hours, the cells on the upper membrane surface were mechanically removed. Cells that had invaded or migrated to the lower side of the membrane were fixed, stained with 0.1% crystal violet, and counted the amount of cells in five random fields under a microscope (×400).

Control and miR340-overexpressing cells (1×10⁵) were suspended in 5 μL of culture medium and injected into the brains of 6 to 8-week-old female NOD/SCID mice that had been anesthetized with 10% pentobarbital. The stereotactic coordinates of the injection site were 2 mm forward from lambda, 2 mm lateral to the sagittal suture, and 5 mm deep. MRI was taken to confirm the tumor formation before decapitation of the mice.

Dissected mouse brains were fixed in 4% paraformaldehyde at 4°C overnight. After fixation, the brains were embedded in paraffin, cut into 6-μm thick coronal sections, and stained with hematoxylin and eosin (H&E).

IHC was performed as previously described (15). The following antibodies were used to detect cellular antigens: mouse monoclonal anti-Nestin (1:200; Pharmingen), mouse monoclonal anti-GFAP (1:200; Sigma), mouse monoclonal anti-Sox2 (1:200; Stem Cell Technology), rabbit polyclonal anti-PLAT (1:50; Sigma), and rabbit monoclonal anti-caspase-3 (cleaved; 1:100; Cell Signaling Technology). The antibodies were detected using either Alexa 488-coupled goat anti-rabbit IgG (1:500; Life Technologies) or sheep anti-rabbit IgG (1:1; Nichirei). The nuclei were counterstained with DAPI (1 μg/mL) or hematoxylin.

Cells transfected with either a control or a miR340 overexpression vector were cultured in the presence of 400 μg/mL G418 for 10 days. Then, selected cells were fixed and stained using a SA-β-gal staining kit (Calbiochem) according to the supplier's instructions.

Cell or tissue lysates were separated by 10% SDS-PAGE and the proteins were transferred to a polyvinylidene fluoride membrane (Amersham). The membrane was blocked in 5% nonfat milk, and then incubated in an anti-Nestin (1:1000), anti-Sox2 (1:500), anti-GFAP (1:500), anti-PLAT (1:250), or anti-GAPDH antibody (1:2000; Chemicon). After washing, the membrane was incubated in IgG-HRP (1:2000; Santa Cruz Biotechnology). An enhanced chemiluminescence system (Amersham) was used to detect the protein bands.

Full-length human plat was amplified from human glioma cDNA using RT-PCR and KOD-Plus-Ver.2 polymerase (Toyobo) according to the manufacturer's instructions and cloned into the pcDNA3-2×FLAG-c vector, resulting in pcDNA3-hPLAT-2×FLAG-c. The primers used for amplification of the full-length human plat gene are listed in Supplementary Table S2.

To knockdown human plat expression, hairpin sequences were inserted into the psiRNA-h7SKhygro G1 expression vector (InvivoGen) to produce psiRNA-h7SKhygro-platsh. The shRNA target sequence was 5′-GAATTCGATGATGACACTT-3′.

To construct a firefly luciferase-human plat 3′ untranslated region (3′UTR) expression vector, human plat 3′UTR genomic DNA was amplified using the KOD plus DNA polymerase and cloned into the pT7Blue-2 vector (Novagen), according to the manufacturer's instructions. We also amplified a mutant form of the plat 3′UTR genomic DNA, with a deleted miR340-binding site, and cloned it into the pT7Blue-2 vector. The wild and mutant types of plat 3′UTR DNA were inserted with the firefly luciferase cDNA into the pcDNA3.1-hyg vector, resulting in pcDNA3.1-hyg-Luc-PLAT 3′UTR (Luc-PLAT) and pcDNA3.1-hyg-Luc-ΔPLAT 3′UTR (Luc-ΔPLAT), respectively. The primers used to generate plat 3′UTR reporter construct are listed in Supplementary Table S3.

hGICs (E3) were transfected with pEF-Rluc (a kind gift from Koji Shimozaki and Shigekazu Nagata), a luciferase-PLAT vector (Luc-PLAT or Luc-ΔPLAT) and a miR expression vector (control vector or pBApoCMV-Neo-miR340), and their luciferase activities were measured as previously described (16).

The Mann–Whitney test was used to compare two-paired groups. Kaplan–Meier curves were used to estimate the unadjusted time-to-event variables. Log-rank tests were applied to compare each time-to-event variable between groups. P values less than 0.05 (two sided) were considered significant. All statistical analyses were performed using the StatMate software program.

To identify novel miRNAs that are aberrantly expressed in GICs, we analyzed differences in miRNA expression between mouse and human GICs and control NSCs using miRNA microarrays. We identified seven miRNAs that were aberrantly expressed in both human and mouse GICs, three of which were upregulated and four of which were downregulated, compared with NSCs. Among these miRNAs, miR340 was identified as a novel miRNA that was significantly downregulated in all human and mouse GICs and in the human glioma cell lines (Fig. 1A). Using qPCR analysis, we confirmed that miR340 expression was significantly decreased in all examined human and mouse GICs and human glioma cell lines compared with NSCs (Fig. 1B). Furthermore, we determined that miR340 expression was significantly decreased in human GBM tissues compared with normal human brain tissues (Fig. 1C).

To examine the function of miR340 in hGICs, we infected hGICs and glioma cell lines with a recombinant lentivirus encoding either GFP alone or GFP and miR340 together. Then, we purified the GFP-positive cells using flow cytometry and evaluated miR340 expression in the infected cells (Supplementary Fig. S2). During the cell culture, we observed that miR340-overexpressing cells became flatter and larger than control cells (Fig. 2A). IHC analysis and Western blot assays revealed that miR340-overexpressing hGICs partially decreased Nestin expression and lost the expression of Sox2, but remained positive for GFAP, an astrocyte marker (Fig. 2B–D). miR340-overexpressing hGICs ceased proliferating during the first 3 days of culture (Fig. 3A). The BrdUrd-incorporation and cell-cycle analyses revealed that miR340 overexpression significantly arrested the cell cycle at the G₁–S transition, as indicated by a marked accumulation of cells in the G₁ peak and by a reduction of cells in the S phase (Fig. 3B and C). In addition, we examined invasiveness and motility in the miR340-overexpressing hGICs and confirmed that miR340 significantly inhibited both invasion and migration in these cells (Fig. 3D and E). We further demonstrated that miR340 overexpression activated the expression of SA-β-gal, a marker of cellular senescence, in hGICs (Fig. 3F). With the exception of cellular senescence, similar results were observed in the miR340-overexpressing human glioma cell lines. Taken together, these data indicate that miR340 negatively regulates the proliferation and invasiveness of both hGICs and glioma cell lines but induces cellular senescence only in hGICs.

Next, we investigated whether miR340 overexpression inhibits GIC-mediated tumorigenesis in vivo. We transplanted either miR340-overexpressing hGICs or their control hGICs into the brains of immunodeficient mice. The control hGICs without miR340 overexpression continuously formed malignant gliomas with high invasiveness and many mitotic cells with a histopathology closely resembling that of the original tumor, causing death in mice within 48 days (Fig. 4A, top, Fig. 4B–D). In contrast, none of the mice injected with miR340-overexpressing hGICs formed apparent tumors (Fig. 4A, bottom), and these mice survived for more than 80 days (Fig. 4D). Histologic examination 3 days after transplantation of miR340-overexpressing hGICs demonstrated that the injected tumor cells were positive for active form of caspase-3 immunostaining, indicating that miR340 induced GIC apoptosis (Fig. 4E). In contrast, tumor formation of miR340-overexpressing human glioma cell lines in mouse brains was not completely suppressed, resulting in the death of the mice although the mice survived 2 or 4 weeks longer than the control (Supplementary Fig. S3A and S3B). These findings indicate that miR340 is a strong suppressor of tumorigenesis in GICs, although this antitumorigenic effect is less obvious in glioma cell lines.

To identify functional targets of miR340, we compared the gene expression profiles of miR340-overexpressing hGICs with the profiles of their control GICs (Fig. 5A). We selected candidate target genes that were significantly upregulated or downregulated in miR340-overexpressing hGICs and analyzed these candidates using ConPath Navigator software, which is based on the GenMAPP software program (www.genmapp.org). These data revealed that miR340 regulates the genes related to two signaling pathways. One is the cell-cycle pathway, which involves p21/Cip1 and cyclin A (17, 18). The other is the cell adhesion/ECM remodeling pathway, which includes PLAT and matrix metalloproteases (Supplementary Fig. S4A and S4B; refs. 19, 20). Of the molecules that were strongly regulated by miR340, we focused on PLAT because its expression was significantly decreased in miR340-expressing hGICs (Supplementary Fig. S5A). In addition, the 3′UTRs of both the mouse and human plat mRNAs contained potential miR340 target sequences (Fig. 5B). IHC analysis and Western blot assays confirmed that miR340 overexpression decreased PLAT expression in hGICs (Fig. 5C and D). Using a reporter vector encoding the firefly luciferase gene with the wild-type plat 3′UTR, we demonstrated that miR340 overexpression inhibited luciferase activity in hGICs, whereas a deletion in the predicted binding site of miR340 in the 3′UTR of the plat gene abrogated the aforementioned inhibitory effect of miR340 (Fig. 5E). Taken together, these data strongly indicate that PLAT is a novel direct target of miR340 in hGICs. We confirmed that PLAT was more highly expressed in human glioblastomas than in normal brain tissues at the level of both mRNA (Fig. 5F) and protein (Fig. 5G) and that the tumor cells positive for PLAT immunostaining coexisted with those cells expressing CD15 (SSEA-1, hGIC marker; ref. 21; Supplementary Fig. S5B).

We subsequently examined the functions of PLAT in hGICs. PLAT knockdown using the plat-sh expression vector (Supplementary Fig. S6A and S6B) inhibited hGIC proliferation and invasion in a manner similar to that induced by miR340 overexpression (Fig. 6A–D). The inhibitory effects were fully recovered by introducing mutant PLAT, in which the shRNA target sequences were deleted (Supplementary Fig. S7). Furthermore, miR340 transfection in hGICs overexpressing the PLAT gene that lacked the 3′UTR recovered the inhibitory effects of miR340 on cell functions (Supplementary Fig. S8). These results indicate that PLAT is a key molecule that promotes malignancy of GIC by directly mediating the functions of miR340 downregulation. In addition, transplanting PLAT-knockdown hGICs into mouse brains resulted in the formation of tiny tumors localized at the injection site, but the growth rate was extremely slow, indicating that PLAT promotes GIC tumor growth in vivo (Fig. 6E).

In the current study, we identified miR340 as a novel miRNA whose expression was significantly lower in both hGICs and glioma cell lines than in NSCs. We also observed that miR340 suppressed not only GIC proliferation and invasion in vitro, but also GIC-initiated tumor formation in mouse brains. These findings indicate that miR340 can act as a suppressor of malignant functions in GICs, particularly gliomagenesis and extensive tumor invasion. Furthermore, we determined that PLAT is the most significant target of miR340 in hGICs from the results of luciferase assays and PLAT-knockdown experiments with plat-sh, which phenocopied the suppressive effects of miR340 overexpression in vitro and in vivo. In addition to PLAT, miR340 overexpression in hGICs decreased the expression of Sox2, c-Met, CD44, and DNMT1, which are miR340 target genes (Supplementary Fig. S9). c-Met and CD44 regulate cell invasion and migration (22, 23), and Sox2 and CD44 play crucial roles in the maintenance of CSC stemness (24). Although miR340 overexpression in hGICs did not markedly decrease Nestin expression, Sox2 and CD44 expression significantly decreased, indicating that miR340 might reduce GIC stemness.

Transplanting miR340-overexpressing hGICs into mouse brains did not result in tumor formation. However, the miR340-overexpressing glioma cell lines generated tumors, although the growth rate was considerably suppressed, finally resulting in tumor-caused death of the mice. The mechanisms underlying the differences in miR340-induced gliomagenesis inhibition in GICs and in glioma cell lines remain unclear. Histopathologically, the transplanted hGICs with miR340 overexpression showed apoptosis in the early stage of transplantation, suggesting that antitumorigenic effects of miR340 in GIC might be partly due to tumor cell apoptosis. Recently, the inhibition of Sox2, which is a stem cell marker, was demonstrated to reduce the tumorigenic potential of human gastric cancer cells (25). In the glioma cell lines, the expression levels of Sox2 were initially quite low; therefore, Sox2 function was likely not highly affected by miR340 in these cell lines. These findings suggest that the tumorigenic mechanisms differ between GICs and more differentiated glioma cell lines.

Interestingly, the present study demonstrated that miR340 overexpression in hGICs promoted senescence in these cells, whereas miR340 introduction into glioma cell lines did not induce senescence (data not shown). Cellular senescence can inhibit tumor formation by modulating the redox state in CSCs and by regulating p27 in glioma cells (26, 27). In addition, GICs derived from wild-type mice expressing the senescence-promoting factor esophageal cancer-related gene 4 (Ecrg4) caused significantly reduced tumor formation in mouse brains compared with cells from Ecrg4-knockdown mice (28). We observed that miR340 overexpression in hGICs decreased DNA methyltransferase 1 (DNMT1) expression (Supplementary Fig. S9D); decreased DNMT1 expression can promote cellular senescence (29). These findings suggest that miR340-induced cellular senescence might explain the reduced tumorigenic activity of miR340-overexpressing GICs.

Currently, many miRNAs that are down-regulated in GSCs have been reported, including miR128, miR124, miR137, miR34a, and miR451 (11), (30–32). The overexpression of these miRNAs inhibits the proliferation and self-renewal, reduces the viability, or induces the differentiation of GSCs but does not cause the irreversible death of GSCs (31–33). Chan and colleagues demonstrated that miR138 acts as a prosurvival oncomiR in GSCs and that functional inhibition of miR138 prevents not only tumorsphere formation in vitro, but also tumorigenesis in vivo by inducing apoptosis and suppressing the proliferation of GSCs (34). It has been recently reported that miR218 inhibits the self-renewal of GSCs by targeting Bmi1 and regulates glioma cell development by Wnt pathways (35). In addition, Wang and colleagues reported that miR33a is highly expressed in GICs and that antagonizing miR33a function in GICs reduces self-renewal and tumor progression in mice by inhibiting the PKA and NOTCH pathways, which promote GIC maintenance (36). Many tumor cell functions, including a high proliferation rate, increased cell motility, and suppressed apoptosis and senescence, may be tightly linked during tumor development. Our findings indicate that miR340 regulates the expression of several key target genes that participate in cell proliferation, invasion, migration, and senescence, including PLAT, c-Met, CD44, Sox2, and DNMT1. Furthermore, p21 and ANXA2 genes, which have important roles in cell proliferation and in cell invasion, respectively, are downstream of PLAT. Both of these genes were also significantly upregulated or downregulated by miR340 overexpression in hGICs (Supplementary Fig. S9), indicating that PLAT plays central roles in cell proliferation, invasion, and migration. The loss of miR340 as a regulator controlling the expression of these genes in GICs promotes the development of these cells into malignant tumors and advances tumor progression (Fig. 7).

The National Cancer Institute's Repository for Molecular Brain Neoplasia Data (REMBRANDT) database (https://caintegrator.nci.nih.gov/rembrandt/; ref. 37) indicates that the prognosis of glioma patients with increased PLAT levels is poorer than that of patients with decreased PLAT levels (Supplementary Fig. S10). These findings suggest that the introduction of miR340, which is capable of directly suppressing PLAT expression, may be a useful therapeutic tool for improving the prognosis of patients with patients. Furthermore, future studies of the functional significance of miR340 target molecules, including PLAT, might provide a greater understanding of the complex mechanisms of GBM development and progression.

No potential conflicts of interest were disclosed.

Conception and design: T. Kondo, J. Tanaka, T. Ohnishi

Development of methodology: T. Kondo

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Yamashita, T. Kondo, S. Ohue, M. Ishikawa, R. Matoba, S. Suehiro, S. Kohno, J. Tanaka

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Yamashita, T. Kondo, S. Ohue, M. Ishikawa, R. Matoba, T. Ohnishi

Writing, review, and/or revision of the manuscript: D. Yamashita, T. Kondo, S. Ohue, T. Ohnishi

Study supervision: T. Kondo, H. Takahashi, S. Kohno, H. Harada, T. Ohnishi

Other (financial support): T. Kondo

The authors thank Tetsuo Moriguchi for his technical suggestions and Naoki Ohtsu for his technical assistances.

This work was supported in part by funds from the Adaptable and Seamless Technology Transfer Program through target-driven R&D (A-STEP; T. Kondo and R. Matoba) and by Grants-in-Aid for scientific research from the Ministry of Education, Culture, Sports, Science, and Technology of Japan [Program for Scientific Research (C) No. 25462266 to T. Ohnishi and S. Ohue].