MicroRNA 21 Promotes Glioma Invasion by Targeting Matrix Metalloproteinase Regulators^{▿ †}

Human tissue samples.

Fresh frozen human nonneoplastic brain tissue and human tumor samples were obtained from the Department of Pathology at Brigham and Women's Hospital. All human materials were used in accordance with the policies of the institutional review board at Brigham and Women's Hospital.

Cell cultures and oligonucleotide transfections.

The human A172, U87, and LN229 glioma cell lines, which express high levels of miR-21 (5), were used for in vitro and in vivo experiments. A172, U87, LN229, cervix carcinoma HeLa, breast carcinoma MCF7, and human osteosarcoma U2OS cells were obtained from the ATCC. The cells were maintained in either Dulbecco's modified Eagle's medium (DMEM) (U87, A172, LN229, HeLa, and U2OS cells) or RPMI medium supplemented with 10% fetal calf serum (MCF7 cells). For miR-21 inhibition, the cells were transfected with 50 nM of 2′-O-methoxyethyl (2′-O-MOE) oligonucleotides by Lipofectamine 2000 (Invitrogen), according to the manufacturer's protocol. 2′-O-MOE oligonucleotides with phosphate backbone were synthesized as previously described (10). The 2′-O-MOE sequences were as follows: for the miR-21 antisense oligonucleotide (ASO), 5′-TCAACATCAGTCTGATAAGCTA-3′; for the mismatch oligonucleotide, 5′-TCTTCATGAGTCAGATTACCTA-3′; and for the control oligonucleotide, 5′-ACATACTCCTTTCTCAGAGTCCA-3′. The sequences of the control and mismatch oligonucleotides have been analyzed using BLAST searches to exclude potential hits in the human transcriptome. For overexpression, the cells were transfected with a synthetic RNA duplex (obtained from IDT) corresponding to mature miR-21. For RECK RNA interference, the following sense sequence was used for targeting: 5′-GACCAGCCCUUGCCUCAAUU-3′.

mRNA arrays and computational data analysis.

A172 glioma cells treated with anti-miR-21, mismatch control antagomiR, or the miR-21 mimetic duplex were compared to mock-treated A172 glioma cells as a reference. Cells were harvested at 6 and 24 h posttreatment, comprising six pairs of samples in all. Agilent microarrays were hybridized as fluor-reversed pairs of arrays.

Microarray analysis was performed as described previously (22, 24). Regulated transcripts were identified in microarray gene expression signatures, using a P value cutoff (P < 0.05). Regulated transcripts were tested for enrichment relative to a background set, using hypergeometric distribution. miR-21 target regulation was measured by enrichment of transcripts containing miR-21 hexamer seed strings (stretches of 6 contiguous bases complementary to miRNA seed region nucleotides 1 to 6, 2 to 7, or 3 to 8) in transcripts having annotated 3′ UTRs. Biological function was detected by enrichment of transcripts from Gene Ontology database biological process functional categories (http://www.geneontology.org/). The set of genes represented by probes on the microarray was used as a background set.

qRT-PCR analysis of mRNA and miRNA expression.

Total RNA from tissues and cells was isolated using TRIzol reagent (Invitrogen) for both mRNA and miRNA analyses. Relative levels of mRNA were examined using Sybr green real-time quantitative reverse transcription-PCR (qRT-PCR) (Applied Biosystems) and normalized to levels of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA. For analysis of miRNA expression, real-time RT-PCR analyses were carried out using TaqMan miRNA assays (Applied Biosystems) according to the manual. Relative expression was calculated using the ΔΔC_T method (38) and normalized to the expression of snoRNA48 (Applied Biosystems) or relatively uniformly expressed miR-19b. All qRT-PCRs were performed in duplicate, and the data are presented as means ± standard errors of the means (SEM).

Western blot analysis.

Cells were transfected with either anti-miR-21 or control oligonucleotides in tissue culture plates and grown for 72 h. The cells were lysed by a standard procedure in radioimmunoprecipitation assay buffer containing protease inhibitor cocktail (Roche Diagnostics). Protein concentrations of total cell lysates were measured using a Micro BCA protein assay kit (Pierce Biotechnology), and 50 μg per lane of total cell lysates was resolved on sodium dodecyl sulfate-polyacrylamide gel electrophoresis gels (Invitrogen), followed by immunoblot detection and visualization with ECL Western blotting detection reagents (Pierce Biotechnology). Immunoblotting was performed with the following primary antibodies: mouse anti-RECK (1:250; BD Biosciences Pharmingen), rabbit anti-TIMP3 (1:1,000; Chemicon International, Inc.), and mouse antiactin (1:5,000; Abcam).

Luciferase miRNA target reporter assay.

Total cDNA from A172 cells was used to amplify the 3′ UTRs of RECK and TIMP3 by PCR using the forward primers ATTAACTAGTACCTCTATTCGCCACACAG and TATGACTAGTAGCCCAGTGATGCTTGTGTTG and the reverse primers CTACATCAGCACTGACATATTCTG and TATGAAGCTTATTCAGGAAAATGGCGGCATGTG, correspondingly. After digestion of the PCR product by SpeI and HindIII, the RECK and TIMP3 3′ UTRs were cloned into the SpeI and HindIII sites of pMir-Report (Ambion), resulting in pMir-Report-3′RECK and pMir-Report-3′TIMP3. Mutations were introduced in the potential miR-21 binding sites by using a QuikChange site-directed mutagenesis kit (Stratagene). HeLa and A172 cells were transfected with the pMir-Report vectors containing the 3′ UTR variants, and at 5 h after transfection, the cells were transfected again with 50 nM of anti-miR-21 or control oligonucleotides. After 24 h, the cells were lysed and luciferase activity was measured. pRenilla was cotransfected and used for normalization.

Scratch migration assay.

Confluent A172 cells were transfected with anti-miR-21 inhibitors or control oligonucleotides. At 5 h after transfection, a cell scratch spatula was used to make a scratch in the cell monolayer. Pictures of the scratches were taken (×4 magnification) by using a digital camera system coupled with a microscope. The cells were incubated for 24 to 72 h, and pictures were taken at different times. The software program MetaVue was used to determine migration distance (in micrometers).

Zymography.

To measure the activity of matrix metalloproteinase (MMP) in glioma cells (A172 and U87), 15 μl of conditioned medium was mixed 1:1 with Zymogram sample buffer (Bio-Rad Laboratories) and resolved on 10% Novex Zymogram gels (Invitrogen). Following electrophoresis, gels were incubated at room temperature in Zymogram renaturing buffer (Bio-Rad Laboratories) for 30 min. Next, the gels were equilibrated in Zymogram development buffer (Bio-Rad Laboratories) at room temperature for 30 min and then incubated at 37°C overnight. The gels were washed and stained in Imperial protein stain (Pierce) according to the manufacturer's protocol. Quantification analysis was performed using the software program ImageJ (NIH).

In vitro invasion assay.

A172 and U87 cells were transfected with either anti-miR-21 or control oligonucleotides, cultured for 48 h, and transferred on the top of Matrigel-coated invasion chambers (24-well insert, 8-μm pore size; BD Biosciences) in a serum-free DMEM. As a chemoattractant, DMEM containing 10% fetal calf serum was added to the lower chamber. After incubation for 20 h, noninvaded cells were removed from the inner part of the insert by using a cotton swab. Fixation and staining of invaded cells were performed using a Diff-Quick differential staining set (Dade Behring, Inc.).

In vivo imaging of MMP activity.

U87 cells were transfected with 50 nM anti-miR-21 inhibitors or control oligonucleotides. At 24 h after transfection, 150 μl phosphate-buffered saline containing 1 × 10⁶ cells was mixed with 50 μl Matrigel (Becton Dickinson) and directly injected subcutaneously in the flanks of nude mice. Each mouse was injected with U87 cells transfected with antisense anti-miR-21 (left flank) and control oligonucleotide U87 cells (right flank). At 3 days after injection of the cells, 150 μl of MMPsense 680 (Visen) was injected intravenously. After 24 h, the mice were subjected to fluorescence imaging using a Kodak fluorescence imaging system (32). The relative fluorescence intensities were determined using ImageJ software (NIH). In parallel, caliper measurements were performed to track tumor growth.

Microarray data accession number.

The microarray data have been deposited in the GEO database (http://www.ncbi.nlm.nih.gov/projects/geo/) and assigned accession number GSE11778.

miR-21 expression correlates with glioma grade.

We have previously demonstrated that miR-21 expression is upregulated in GBMs (5). To further assess the relevance of miR-21 in glioma tumorigenicity, we determined miR-21 levels in tumors of different grades (grade II [low-grade gliomas], grade III [anaplastic astrocytomas and oligodendrogliomas], and grade IV [GBMs]) compared to normal brain tissue by qRT-PCR. miR-21 levels remain low in grade II gliomas and in most grade III tumors and are significantly higher in GBMs (Fig. 1). This expression pattern is different from those of the other cancer-associated miRNAs, for example, miR-10b, which is often elevated in grade III and even grade II gliomas (Fig. 1) and dysregulated in some other cancers (41). The most important characteristics used for glioma grading and GBM diagnostics are microvascular proliferation, which is associated with reorganization of the extracellular matrix (ECM), and high proliferative capacity in the tumor cells (40). The high level of miR-21 expression in GBMs compared to that in anaplastic astrocytomas suggests the involvement of miR-21 in one of these key processes.

mRNA profiling of miR-21-modulated gliomas: apoptosis and ECM remodeling.

To identify molecular pathways and gain insights into direct mRNA targets affected by miR-21, we suppressed miR-21 in human glioma A172 cells by transfecting them with modified ASOs. We then used Merck Custom human (Agilent) 44K expression arrays to determine mRNA profiles and identify genes whose expression levels were persistently and reproducibly changed by suppression of this miRNA (see Table S1 in the supplemental material). The analysis was performed on RNA samples harvested at 6 and 24 hours posttransfection, consistent with the time frames commonly used for this type of analysis (18, 36, 37). Since modified oligonucleotides can induce a wide range of nonspecific, off-target-mediated effects, several ASO molecules of different chemistries have been used for miR-21 inhibition. To assess the degree of inhibition, we transfected into glioma cells the luciferase reporter of miR-21 activity, a vector that contained a perfect miR-21 binding site downstream of the luciferase gene (pMIR-Report; Ambion) and was therefore utilized for testing miR-21 activity and the efficacy of its inhibitors. Among the molecules tested, the 2′-O-MOE ASO was the most potent in repressing miR-21 activity (Fig. 2A), in agreement with previous data (10), whereas the uniform 2′-O-methyl ASO was the least potent, and mixed locked nucleic acid (LNA)-DNA oligonucleotides were variable (data not shown). In addition to being the most potent ASO, the 2′-O-MOE ASO also induced expression of an mRNA population significantly enriched in miR-21 seeds. In fact, three hexamers, corresponding to nucleotides 1 to 6, 2 to 7, and 3 to 8 of miR-21, were strongly overrepresented in the population of upregulated transcripts (with expectations of 1.45e−09, 1.37e−14, and 1.82e−12, respectively) but not in the population of downregulated transcripts (Fig. 2B). Such overrepresentation of the hexamers suggests an abundance of direct miR-21 targets among the genes upregulated by 2′-O-MOE anti-miR-21 and indicates that the effects of this molecule on glioma cells are indeed mediated through the miR-21 pathway. Other tested inhibitors did not show such a strong bias for regulating miR-21 seed-containing mRNAs (data not shown). We therefore used the 2′-O-MOE anti-miR-21 molecule and the corresponding mismatched control oligonucleotide in subsequent experiments.

Transcriptional profiling of knocked-down miR-21 cells revealed changes in the expression levels of ∼570 genes (P < 0.05) associated with various biological functions (Fig. 2C). DNA damage response genes, regulators of cell cycle arrest, and positive regulators of apoptosis have been enriched among genes upregulated at 24 h. Among downregulated genes, those involved in stress response, apoptosis, and regulation of signal transduction (particularly Jun N-terminal protein kinase cascade, mitogen-activated protein kinase kinase kinase cascade, and stress-activated protein kinase pathway) and, most significantly, genes associated with blood vessel morphogenesis and development have been strongly enriched (expectation value < 10⁻⁴) (Fig. 2C). This molecular profiling suggested that miR-21 regulates multiple genes involved in several cellular programs in glioma cells.

Concurrent with miR-21 inhibition, we have also tested the effects of miR-21 overexpression by transfecting synthetic miR-21 duplex molecules into the A172 glioma cells (Fig. 3). Since, as demonstrated earlier, the endogenous miR-21 is high in gliomas and many other cancer cell types (5, 54), we hypothesized that its overexpression might have additional regulatory effects on miR-21 targets. Notably, a large group of miR-21 octamer- and/or heptamer-containing mRNAs (at least 65) that were upregulated by anti-miR-21 (P < 0.05) has been also downregulated by miR-21 overexpression (P < 0.05) (Fig. 3). Such reverse correlation between modulated miR-21 level and expression levels of this set of mRNAs suggests that this mRNA population is enriched in direct miR-21 targets. An abundance of targets predicted by several commonly used computational algorithms supports this conclusion. Particularly, PDCD4, RECK, SOX2, and PELI1 have been predicted by TargetScan, PicTar, and Miranda (16, 17, 29, 34), and Yod1, PPARA, GPR64, RASGRP1, FAM63B, TIMP3, CDC25A, GLCCI1, TRIM59, CCDC14, PLEKHA1, CPEB3, MSH2, TNFRSF11B, ANKRD46, Sesn1, FAM3c, STAT3, and APAF1 have been predicted either by TargetScan, PicTar, and Miranda or by RNA Hybrid (25, 29, 34, 49). Several of these genes are directly associated with apoptosis. Prominent among them are apoptotic peptidase activating factor 1 (APAF1), a cytoplasmic protein that initiates apoptosis by forming apoptosome (26); PDCD4, a tumor suppressor recently validated as a direct miR-21 target (1, 14), whose expression is often lost in gliomas (15); and tumor necrosis factor receptor superfamily member 11B (TNFRSF11B), which was shown to have a role in different cancers (20).

In addition to apoptosis- and cell cycle-related genes, analysis of miR-21 array data identified a number of important ECM remodeling factors that are computationally predicted miR-21 targets (Fig. 3; also see Table S1 in the supplemental material). Remarkably, the mRNA levels of two major inhibitors of MMPs, RECK and TIMP3, have been affected by both miR-21 knockdown and its overexpression. MMPs are a group of peptidases involved in degradation of the ECM. Accumulated data suggested that MMP levels and activities are significantly elevated in human gliomas, which contributes to glioma cell invasion of the surrounding normal tissues, metastasis, and angiogenesis through cell surface ECM degradation (21, 45). mRNAs carrying RECK and TIMP3 are predicted conserved miR-21 targets with one and two putative binding sites, respectively (Fig. 4A). RECK is a membrane-anchored MMP inhibitor whose reduced expression or inactivation seems to be critical for the invasiveness and metastasis of various cancers, including glioma cancer (7, 46). Its expression level is also an important prognostic factor for multiple cancer types (6). However, molecular mechanisms that downregulate RECK level are not well studied. TIMP3 is a tissue inhibitor of MMPs that inhibits angiogenesis and tumor cell infiltration and also induces apoptosis (2, 48). It has been demonstrated to inhibit MMP activity and tumor growth in malignant gliomas in vivo (31, 32). Our mRNA array analysis suggested that RECK and TIMP3 mRNAs may be regulated by miR-21.

miR-21 targets MMP inhibitors.

In order to test our hypothesis, we examined RECK and TIMP3 mRNA levels in the progression of human gliomas (normal brain tissue versus gliomas of different grades) by qRT-PCR. RECK mRNA expression was significantly lower in GBMs than in normal brain tissue and, especially, in grade II and III gliomas. TIMP3 mRNA expression also tended to be lower in GBM tumors than in grade II and III gliomas (Fig. 4B). These data suggest that RECK and TIMP3 may represent natural miR-21 targets in vivo. To validate miR-21 regulation of RECK and TIMP3, we first confirmed RECK and TIMP3 mRNA elevation in miR-21 knocked-down A172 cells by qRT-PCR (Fig. 4C). RECK mRNA level increased consistently ∼2-fold and TIMP3 mRNA about 1.6-fold in the cells transfected with anti-miR-21. We then examined the protein levels of RECK and TIMP3 and found that both proteins were upregulated by miR-21 inhibition in glioma A172 and LN229 cells (Fig. 4D). In glioma U87 cells, TIMP3 was also upregulated by anti-miR-21; however, we could not detect distinct RECK bands by Western blot analyses. Nevertheless, using qRT-PCR, we observed a twofold upregulation of RECK mRNA expression by anti-miR-21 (data not shown), similar to the effect demonstrated for the A172 glioma cell line (Fig. 4C). To validate that miR-21 can bind directly to and regulate RECK and TIMP3 mRNAs through the predicted binding sites, we constructed luciferase reporter plasmids that contain RECK and TIMP3 3′ UTR sequences downstream of firefly luciferase and tested the luciferase activity in HeLa and A172 glioma cells, both expressing high levels of endogenous miR-21. We also constructed mutant versions of these reporters, altering bases in the putative miR-21 binding sites in a way that was predicted to abolish miR-21 binding. All constructs were transfected into cultured HeLa and A172 cells, along with a Renilla luciferase transfection control, and both firefly and Renilla luciferase levels were assayed by luminometry. These experiments have been performed with the miR-21 inhibitor and with a mismatched control oligonucleotide to ensure that any observed effects are driven by miR-21. These results indicate that RECK is a direct miR-21 target (Fig. 4E). The TIMP3 3′ UTR construct did not consistently respond to changes in miR-21 levels, indicating that TIMP3 is not likely to be a direct miR-21 target (Fig. 4E). However, since TIMP3 mRNA and protein levels were modulated by miR-21 expression (Fig. 4C and D), TIMP3 is likely to be a downstream effector of miR-21 signaling.

Inhibition of miR-21 leads to downregulation of MMP activities.

Since the expression levels of at least two MMP inhibitors, RECK and TIMP3, are regulated in glioma cells by miR-21, we next examined the effects of miR-21 inhibition on MMP proteolytic activities. Conditioned media from glioma A172 and U87 cells transfected with either miR-21 inhibitor or the control oligonucleotide were analyzed by gelatin zymography. In this assay, major bands that correspond to the active form of MMP-2 and/or pre-MMP-2 were detected in glioma cell lines. When miR-21 was suppressed, the amount of gelatin digested that reflects the levels of MMP-2 proteolytic activity was significantly reduced relative to the amount for the cells transfected with the control oligonucleotide (Fig. 5A). Notably, total protein levels of secreted MMP-2 were not affected by miR-21 suppression (data not shown). Concurring with previously published reports (12, 46), our data indicate that these MMP inhibitors regulate MMP activity posttranslationally.

We next studied the effects of miR-21 inhibition on MMP proteolytic function during glioma growth in vivo. In these experiments, we used tumorigenic U87 cells, a common model of human gliomas (23). U87 cells transfected with either anti-miR-21 or the control oligonucleotide were implanted into nude mice subcutaneously, and MMP activities were assessed by using an MMPSense 680 imaging probe (VisEn Medical). This probe is a protease-activated fluorescent agent, optically silent in its inactive state and highly fluorescent following MMP-mediated activation. It allows in vivo imaging of key MMPs, including MMP-2 as well as MMP-3, -9, -7, -12, and -13 (4, 32). Each mouse was implanted subcutaneously with anti-miR-21-transfected U87 cells on one side of the body and control U87 cells on the other side. At 3 days after cell implantation, MMPsense was injected intravenously, and the mice were visualized using a fluorescence imaging system 24 h later. All tumors produced from miR-21-inhibited cells demonstrated significantly lower MMP activity than control tumors (Fig. 5B and C). The relative fluorescence intensities were determined using ImageJ software (NIH), and MMP signals were normalized to tumor size, measured by a caliper. The anti-miR-21 oligonucleotide used in this study did not affect subcutaneous U87 tumor growth within the first week after cell implantation, when the imaging was performed (data not shown).

Inhibition of miR-21 reduces motility and invasiveness of glioma cells.

MMPs can degrade ECM components and therefore play important roles in mediating GBM tumor cell motility and invasion (45). Moreover, MMP-2 and MMP-9 activities correlate well with glioma cell migration and invasiveness (58). We have therefore tested miR-21 effects on glioma cell motility and invasion. First, we performed an in vitro scratch motility assay on A172 cells (35). Cells were plated, allowed to grow to confluence, and transfected with anti-miR-21 or the mismatched control oligonucleotide. The monolayer was then scratched to create a cleared area within the monolayer. Pictures were taken, and the two-dimensional movement of the transfected cells was quantified by measuring the migration distance at different time points and comparing it with the front area at time zero. These experiments demonstrated that inhibition of miR-21 reduced the motility of glioma cells (Fig. 6B and C). Notably, this reduced-motility phenotype was partly rescued by cotransfection of anti-miR-21 with small interfering RNA (siRNA) to RECK (Fig. 6B and D), indicating that miR-21 regulation of glioma cell motility is mediated, at least partly, by RECK.

To measure miR-21 effects on glioma cell invasiveness, a better indication of glioma migratory and invasive properties in vivo, we employed a Transwell invasion system. The system consists of two fluid-filled, stacked compartments separated by a porous membrane filter coated with Matrigel. Cells were grown in the upper chamber and assessed for invasion through the Matrigel toward a chemoattractant (10% serum) in the lower chamber. The numbers of invasive A172 and U87 cells in cultures with inhibited miR-21 were significantly reduced relative to those in cultures transfected with the control oligonucleotide (Fig. 7A and B). This reduced invasiveness was not observed, however, when anti-miR-21 was cotransfected with RECK siRNA (Fig. 7C), suggesting that RECK is the major factor that mediates miR-21 function in glioma invasion, at least in vitro.

mRNA expression profiling for revealing miR-21 functions.

In this study, we attempted to identify major mRNA targets and signaling pathways that mediate miR-21 regulation in gliomas. We used both gain- and loss-of-function approaches and combined them with 44K Merck Custom human expression array analysis to study miR-21 signaling in glioma cells (Fig. 2B and C and 3; also see Table S1 in the supplemental material). Such a bidirectional approach to manipulating miR-21 levels enabled narrowing of the lists of regulated genes and identification of those whose expression correlated negatively with miR-21 levels, e.g., upregulated under miR-21 inhibition conditions and downregulated by its overexpression. Though mRNA expression analysis does not allow identification of targets whose translation is regulated without modulation of mRNA levels, the advantages of such an approach are multifold. First, it enables bioinformatic analysis for determination of biological functions and signaling pathways affected and therefore can predict functions that a particular miRNA may play. Second, it allows sequence analysis and identification of genes showing predicted miRNA-binding motifs and enrichment in seeds (putative direct targets). Importantly, since both inhibition and overexpression of an miRNA with synthetic oligonucleotides can potentially cause significant off-target effects, enrichment of the predicted targets among negatively regulated genes validates the specificity of the response and enables an intelligent design of most specific miRNA inhibitors and mimics. Here, we show that a 2′-O-MOE ASO not only is potent but also upregulates miR-21 seed-containing mRNAs (Fig. 2 and 3).

Inhibition of miR-21 leads to a decrease of MMP activities.

Our mRNA expression analysis suggests that miR-21 regulates the expression of multiple mRNA targets and thereby induces several cellular programs in glioma cells, including those associated with the tumor invasiveness and microvascular proliferation typical for the GBM as well as those involved in cell cycle regulation and apoptosis. Many genes identified in this study are potent signal transducers and regulators of gene expression implicated in glioma biology and more generally in carcinogenesis. In this study, we have demonstrated in vitro and in vivo, using U87 glioma xenografts in nude mice, that miR-21 regulates MMPs and glioma cell invasiveness by directly controlling the MMP inhibitor RECK. RECK, a membrane-anchored regulator, and TIMP3, the ECM-bound protease regulator, are key inhibitors of several MMPs, and their expression is prognostic in a number of common cancers (28, 53). miR-21 negatively regulates the mRNA and protein levels of RECK and TIMP3 in cultured glioma cells (Fig. 4C and D). Inhibition of miR-21 decreases MMP activities both in vitro and in vivo (Fig. 5). Moreover, it leads to reduction of glioma cell motility and invasion, which is mediated by elevated RECK expression and was partly rescued by RECK siRNA (Fig. 6 and 7). In glioma progression from less invasive grade II gliomas to very invasive GBMs, RECK and TIMP3 levels drop (Fig. 4B), whereas miR-21 expression increases (Fig. 1). This implies that miR-21 regulates RECK and TIMP3 expression and thus MMP activities and glioma tumor invasiveness.

In addition to MMP inhibition, TIMP3 may contribute to blocking of cancer progression by participating in a number of cellular processes, including apoptosis. Its ectopic expression in glioma cells induces upregulation of genes associated with angiogenesis, growth factors, cytokines, death receptors, and substrates of the various MMPs (31). Importantly, it inhibits tumor necrosis factor alpha-converting enzyme and induces apoptosis through the stabilization of tumor necrosis factor alpha receptors on the cell surface (42). TIMP3 has been demonstrated to activate caspases and therefore induce apoptosis of human glioma cells in vitro and the growth inhibition of human glioma tumors in a mouse model (31, 32).

Biological functions of other miR-21 targets.

In addition to RECK and TIMP3, many genes respond to both increased and decreased levels of miR-21 (Fig. 3; also see Table S1 in the supplemental material), including several that play important functions in glioma biology and in carcinogenesis. For example, APAF1 (apoptotic protease activating factor 1) is the molecular core of apoptosome. It is typically required for activation of the caspases that initiate apoptosis (3, 26). The APAF1 3′ UTR contains a strong miR-21 binding site (9-mer binding at miR-21 5′ end), and therefore, it is likely one of the direct miR-21 targets. In gliomas, APAF1 is often inactivated or downregulated (56). Our data suggest that this can be at least partly due to miR-21 regulation, in addition to the reported chromosome 12q22-23 loss of heterozygosity and hypermethylation (56). Its overexpression by viral transduction induces apoptosis in glioma cells and may be beneficial in glioma treatment (50). STAT3, the other gene that may play a tumor suppressor function in GBM (11), is also negatively regulated by miR-21 according to the microarray data (see Table S1 in the supplemental material) and is a predicted miR-21 target (29, 34). Interestingly, STAT3 can potentially mediate IL-6-miR-21 autocrine feedback. As mentioned, both STAT3 and IL-6 are usually elevated in malignant gliomas (57) and can be among the factors inducing miR-21 expression (39). miR-21, in turn, may downregulate STAT3 (see Table S1 in the supplemental material). Such a regulatory loop between miR-21 and IL-6/STAT3 may provide a feedback mechanism for stabilizing miR-21 expression and balancing STAT3 signaling.

Several mRNAs regulated by miR-21 and anti-miR-21 in glioma cells (Fig. 3; also see Table S1 in the supplemental material) have recently been identified in an mRNA expression analysis of miR-21 knocked-down MCF7 breast carcinoma cells (14). Among these mRNAs are PDCD4, APAF1, FAM3C, SESN1, and GLCCI1. Our analysis also confirmed tropomyosin 1 (TPM1) regulation by miR-21, previously reported to occur in MCF7 cells (59, 60). As identified in our study, TIMP3 or both TIMP3 and RECK are also induced by anti-miR-21 in MCF7 (breast carcinoma) and U2OS (osteosarcoma) cells (Fig. 4D). However, we have not detected any correlation between miR-21-modulated expression in several glioma cell lines and protein level of phosphatase and tensin homolog (data not shown), the reported miR-21 target in hepatocellular carcinomas (44). These combined results indicate that miR-21 regulates overlapping though not identical sets of genes in different cells and therefore shares common signaling pathways in different cancers.

Issues of miRNA inhibitors and glioma models.

We note that, though 2′-O-MOE miR-21 ASO induced many proapoptotic and cell cycle arrest genes (Fig. 2C; also see Table S1 in the supplemental material), it did not significantly and consistently change the viability of glioma cell lines, which was reported for other, less potent miR-21 inhibitors (5, 8). This strong and specific 2′-O-MOE anti-miR-21 ASO also did not cause a reduction of subcutaneous tumor growth in a U87 glioma model (data not shown). In contrast, an LNA-DNA anti-miR-21 oligonucleotide was previously shown to reduce tumors in an intracranial U87 glioma model (8). Although the locations of the tumors in the mice were different, not fully allowing a direct comparison, the likely explanation for these conflicting results might be that the suboptimally designed LNA-DNA oligonucleotide used in the earlier study killed glioma cells by an off-target mechanism. We should also note that subcutaneous U87 tumors are noninfiltrative, noninvasive, and poorly vascularized. Therefore, though commonly used and very robust, they do not represent an accurate model for studying invasiveness and its effects on tumor growth of human GBMs. Better models will be developed and employed in the future to study the effects of miR-21 inhibition on glioma growth and invasion.

Considered together, this study and previous reports show that differences in nucleic acid chemistry can, in some cases, profoundly alter cellular responses to miRNA inhibition. More-careful analyses that combine mRNA expression profiling with studies of inhibitor potency should help to ensure that cellular responses to miRNA inhibition are specific and genuine. While the goal of this study was to better understand miR-21 function in gliomas, future research is required to address the therapeutic potential of modulating miR-21. Our data suggest that by modulating expression and/or activity of one miRNA with specific tools, e.g., inhibitors or mimics, it is possible to control the expression of multiple oncogenic or tumor-suppressing genes and simultaneously trigger several cellular programs that these genes mediate.