Blocking MIR155HG/miR-155 axis inhibits mesenchymal transition in glioma

Human Tissue Samples and Cell Lines

In total, 225 glioma data with mRNA expression microarray were downloaded from the Chinese Glioma Genome Atlas (CGGA) data portal (http://www.cgga.org.cn, accessed February 21, 2017). Among all the samples, 158 gliomas contain miRNA expression data. Glioma gene expression datasets deposited at the Repository of Molecular Brain Neoplasia Data (REMBRANDT; http://caintegrator.nci.nih.gov/rembrandt/, accessed February 21, 2017) were also included in this study. Ten normal brain tissues (NBTs) and 10 primary glioblastoma (GBM) samples were obtained from the Department of Neurosurgery, The First Affiliated Hospital of Nanjing Medical University. Written informed consent was obtained from all patients. Our study was approved by the institutional review board and the ethics committee of Nanjing Medical University. The human U87 and U251 GBM cell lines were purchased from the Chinese Academy of Sciences Cell Bank. The primary GBM cells (pGBM-1) were established from a primary GBM surgical specimen.

Significance Analysis of Microarray and Gene Set Enrichment Analysis

The Significance Analysis of Microarray (SAM) was performed using a 2-class t-test in the R program (version 3.1.2) to identify differentially expressed genes in different MIR155HG expression groups. The Gene Set Enrichment Analysis (GSEA) of the differentially expressed genes was performed using software downloaded from the Broad Institute (http://www.broadinstitute.org/gsea/index.jsp, accessed February 21, 2017) with C2 (curated gene sets) and C5 (Gene Ontology gene sets) collections.

Oligonucleotides, Plasmids, and Transfection

The MIR155HG small interfering (si)RNA, and inhibitor, mimics and negative control (NC) RNAs of miR-155-5p and miR-155-3p were synthesized and purified by GenePharma. The RNA sequences mentioned above were as follows: si-MIR155HG: sense, 5′-CUGGGAUGUUCAACCUUAATT-3′, antisense, 5′-UUAAGGUUGAACAUCCCAGTT-3′; miR-155-5p inhibitor (anti–miR-155-5p): 5′-ACCCCUAUCA CGAUUAGCAUUAA-3′, miR-155-5p mimics: sense, 5′-UUAAUGCUAAUCGUGAUAGGGGU-3′, antisense, 5′-CCCUAUCACGAUUAGCAUUAAUU-3′; miR-155-3p inhibitor (anti–miR-155-3p): 5′-UGUUAAUGCUA AUAUGUAGGAG-3′, miR-155-3p mimics: sense, 5′-CUCC UACAUAUUAGCAUUAACA-3′，antisense, 5′-UUAAUGC UAAUAUGUAGGAGUU-3′. An siRNA that was unrelated to any human sequence was used as an NC: sense, 5′-UUCUCC GAACGUGUCACGUTT-3′, antisense: 5′-ACGUGA CACGUUC GGAGAATT-3′; miRNA inhibitor negative control (anti-NC): 5′-CAGUACUUUUGUGUAGUACAA-3′.

The protocadherin (PCDH)9 and PCDH7 overexpression plasmids (pcDNA3.1-PCHD9, pcDNA3.1-PCHD7), which included the full-length protein coding sequences, were constructed by GeneChem. Oligonucleotides and plasmids were transfected into cells using Lipofectamine 2000 (Invitrogen) following the manufacturer’s instructions.

Cell Experiments

RNA extraction, quantitative real-time (RT)-PCR, Cell Counting Kit-8 assay, wound healing assay, transwell assay, western blot assay, luciferase reporter assay, and assay by MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) are described in the Supplementary material.

Orthotopic Glioma Model and Treatment

Female nude mice at 4–6 weeks of age were used in this experiment, following the standard guidelines under a protocol approved by Nanjing Medical University. Lentivirus containing siMIR155HG segments (Lenti-siMIR155HG; siRNA sequence is 5′-CUGGGAUGUUCAACCUUAATT-3′) or NC sequence (Lenti-NC; negative control sequence is 5′-UUCUCCGAACGUGUCACGUTT-3′) was obtained from GeneChem. A total of 0.5 × 10⁵ U87 cells were co-infected with lentiviruses expressing luciferase and a Lenti-NC or Lenti-siMIR155HG and then implanted stereotactically to establish intracranial gliomas. Mice were imaged for Fluc activity using bioluminescence imaging on days 3, 7, 14, 21, and 28.

High-Throughput Molecular Dockings

The hairpin loop of the pre–miR-155 was chosen to build the 3D structure using the MC-Fold/MC-Sym pipeline.¹⁸ High-throughput molecular dockings for pre–miR-155 against the 1990 National Cancer Institute/diversity compounds were conducted using the AutoDock program. Protein–ligand interactions were analyzed and visualized using Jmol (http://jmol.sourceforge.net/, accessed February 21, 2017). Detailed high-throughput screening is described in the Supplementary materials.

Statistical Analysis

All data shown in the graphs are presented as mean ± standard error of 3 independent experiments. The 2-sided Student’s t-test was used to determine differences between 2 groups and one-way ANOVA was used to test differences among at least 3 groups. The log-rank test was employed to assess the statistical significance between stratified survival groups using the median value as the cutoff in a Kaplan–Meier analysis. Cox proportional hazards regression analyses were performed using SPSS software, and Pearson correlation was used to determine significant differences. P < .05 was considered significant for all statistical analysis.

Increased MIR155HG Expression Correlates with Glioma Grade

To examine MIR155HG expression in human glioma, we initially analyzed the whole genome gene profiling of 5 normal and 220 glioma tissues in the CGGA dataset. As shown in Fig. 1A, GBM tissues showed significantly increased MIR155HG transcript levels compared with normal tissues (P = .0051). Moreover, MIR155HG expression was significantly higher in high-grade glioma (HGG) than in low-grade glioma (LGG) (grade IV vs grade II, P < .001; grade III vs grade II, P = .0216). We next examined the association between MIR155HG expression level and glioma grade in another independent glioma gene expression dataset, REMBRANDT. We found that MIR155HG was significantly associated with tumor grade (one-way P < .0001), which was consistent with the CGGA data. Together these findings suggest that MIR155HG may play an important role in glioma progression.

MIR155HG Is an Independent Prognostic Factor for GBM Patients

To investigate the correlation between MIR155HG expression and overall survival, we conducted Kaplan–Meier survival curve analysis with a log-rank comparison in the CGGA and REMBRANDT datasets. A total of 83 primary GBMs with complete survival data in the CGGA set were divided into the high MIR155HG expression group (n = 42) and the low expression group (n = 41) based on the median ratio of relative MIR155HG expression in tumor tissues. We found that high expression of MIR155HG was inversely correlated with overall survival (P = .0196, Fig. 1B). Similar results were detected in the REMBRANDT data (P = .0065). These data indicated that overexpression of MIR155HG conferred a poor prognosis in GBM patients.

We next examined the relationship between MIR155HG expression and the features of clinical and molecular pathology of the 83 primary GBMs mentioned above. As shown in Supplementary Table S1, the expression of MIR155HG was significantly associated with male sex (P = .011) and wild-type isocitrate dehydrogenase 1 (IDH1) (P = .001), and was independent of the rest of the features. We then conducted a univariate Cox regression analysis using clinical and genetic variables for 83 primary GBMs and found that high expression of MIR155HG, high KPS score, and total resection were remarkably associated with overall survival, while gender or IDH1 mutation showed no association (Table 1).

Then we performed a multivariate Cox proportional hazards model to evaluate the factors that contributed to overall survival. MIR155HG expression, KPS score, and expression of proliferating cell nuclear antigen were correlated independently with overall survival (hazard ratio [HR] = 2.082, P = .023; HR = 0.450, P = .016; and HR = 2.455, P = .014, respectively) when considering age, resection, epidermal growth factor receptor, Ki-67, topoisomerase type II, and glutathione S-transferase pi (P < .3, univariate Cox regression analysis) (Table 1). Together these results implied that MIR155HG may serve as an independent prognostic factor for GBM patients.

MIR155HG Is a Mesenchymal Transition–Associated Long Noncoding RNA

Using the expression data from the CGGA dataset, we identified 1858 upregulated genes and 2120 downregulated genes in the high MIR155HG expression group (SAM, false discovery rate < 5%, fold change > 2; Fig. 1C). We named these genes “MIR155HG differentially expressed genes.” We performed GSEA to verify whether we could detect differences in the mesenchymal transition–associated genes listed by Cheng¹⁹ in the CGGA data, and we found that these genes were indeed significantly enriched in HGG compared with LGG (P < .0001, Supplementary Fig. S1A). Further, GSEA of the MIR155HG differentially expressed genes demonstrated remarkable enrichments in mesenchymal transition–associated genes in the high MIR155HG expression samples of GBM (P < .0001; Fig. 1D and Supplementary Fig. S1B).

GSEA was then performed to identify Gene Ontology terms and Kyoto Encyclopedia of Genes and Genomes and Reactome pathways that were differentially regulated by MIR155HG differentially expressed genes between the high MIR155HG expression group and the low expression group. The GSEA results revealed many mesenchymal transition–related categories (extracellular matrix, extracellular region, cell migration, response to wounding, extracellular matrix receptor interaction, and focal adhesion) (Supplementary Fig. S1C and D). Additionally, GSEA indicated that 6 distinct published mesenchymal transition–related gene signatures^20–25 were significantly enriched in high MIR155HG expression samples, strongly suggesting that MIR155HG induced a pervasive and sustained mesenchymal transition signaling program (Supplementary Fig. S1E). The work flow of these analyses is shown in Supplementary Fig. S1F.

Ten NBTs and 10 primary GBM samples from patients were used to detect the expression status of MIR155HG by RT-PCR. As shown in Fig. 1E, the level of MIR155HG was significantly upregulated in these GBM specimens compared with the level in NBTs (P < .0001). We chose 3 mesenchymal transition–associated markers (POSTN, MMP11, and FN1) from Cheng’s list and found higher expression levels in GBMs with high MIR155HG expression (n = 5) than those with low MIR155HG expression (n = 5) (Fig. 1F). Next, to directly test the functional role of MIR155HG in the mesenchymal transition, MIR155HG was downregulated by specific siRNA in glioma cells (Supplementary Fig. S2A). MIR155HG reduction triggered remarkable decrease of β-catenin, N-cadherin, vimentin, and matrix metalloproteinase 9 (Fig. 1G). These results strongly demonstrated that MIR155HG is a mesenchymal transition–associated lncRNA.

MIR155HG Promotes Glioma Progression In vitro and In vivo

We examined the role of MIR155HG in cell proliferation by Cell Counting Kit-8 assays. Cell proliferation was dramatically suppressed in U87, U251, and primary GBM cells after transfection with siMIR155HG compared with NC cells (Supplementary Fig. S2B). Glioma cells transfected with siMIR155HG presented significantly reduced scratch wound healing and lower invasion abilities compared with NC (Fig. 2A and B and Supplementary Fig. S2C and D).

To further examine whether tumor growth was inhibited by siMIR155HG, we performed in vivo analyses using a U87 orthotopic glioma model. A statistically significant difference in tumor volume as indicated by bioluminescence imaging was detected between the Lenti-NC and Lenti-siMIR155HG treated groups (Fig. 2C). Representative images of mice implanted with intracranial tumors are shown in Fig. 2D. In addition, treating with Lenti-siMIR155HG was associated with significantly longer survival times in the mice (Fig. 2E).

MiR-155-5p and miR-155-3p Are Key Derivatives of MIR155HG

MIR155HG is the primary miRNA of miR-155, and thus we next examined miR-155 expression. MIR155HG knockdown remarkably reduced miR-155-5p and miR-155-3p expression in U87 and primary GBM cells (Fig. 3A). As shown in Supplementary Fig. S3A and B, Pearson correlation assay revealed a significant and positive correlation between MIR155HG and miR-155-5p or miR-155-3p in 158 gliomas and 64 GBM tissues. One-way ANOVA analysis showed that both miR-155-5p and miR-155-3p were significantly associated with tumor grade (P = .0302 and P < .0001, respectively) and their expression levels were higher in GBM than in LGG (Supplementary Fig. S3C). Moreover, high expression of miR-155-3p correlated with a worse survival outcome in GBM patients (P = .0448), while miR-155-5p expression did not show significant correlation with overall survival (P = .8340; Supplementary Fig. S3D). After the expression of miR-155-5p or miR-155-3p was downregulated by anti–miR-155-5p or anti–miR-155-3p transfection, respectively (Supplementary Fig. S4A), cell proliferation, migration, and invasion were remarkably suppressed in glioma cells (Fig. 3B and C and Supplementary Fig. S4B and C).

To determine whether the phenotypes associated with MIR155HG are mediated by miR-155-5p and miR-155-3p, we cotransfected siMIR155HG with miR-155-5p or miR-155-3p mimics. As shown in Fig. 3D and E and Supplementary Fig. S4D and E, overexpression of miR-155-5p or miR-155-3p markedly reversed the inhibition of cell proliferation, migration, and invasion induced by siMIR155HG. Collectively, these data suggested that miR-155-5p and miR-155-3p are critical participants in the biological function of MIR155HG in glioma.

MiR-155-5p or miR-155-3p Targets Protocadherin 9 or 7, Respectively

To explore downstream targets of miR-155-5p and miR-155-3p, we performed bioinformatics analysis using 2 online algorithms, TargetScan and microRNA, to predict mRNA targets. We found that PCDH9 contains putative binding sites in its 3′ untranslated region (UTR) for miR-155-5p, and PCDH7 contains putative binding sites for miR-155-3p (Fig. 4A). CGGA data revealed remarkable and negative correlation between miR-155-5p and PCDH9 and between miR-155-3p and PCDH7 (Supplementary Fig. S5A and B). Furthermore, reduced miR-155-5p or miR-155-3p expression in glioma cells led to an increase in PCDH9 or PCDH7 protein (Fig. 4B). Luciferase reporter assays showed that overexpression of miR-155-5p led to a marked decrease of luciferase activity of the PCDH9–3′ UTR wild-type construct in U87 cells with no change in luciferase activity of PCDH9–3′ UTR mutant, which contains mutations in the putative binding sites (Fig. 4C). Similar data were observed with miR-155-3p. Our results strongly indicate that PCDH9 and PCDH7 are direct targets of miR-155-5p and miR-155-3p, respectively.

Protocadherin 9 and 7 Function as Tumor Suppressor Genes by Inhibiting the Wnt/ β-catenin Signaling Pathway

PCDH9 and PCDH7 were significantly and reversely correlated with glioma grade and were remarkably downregulated in HGG in the CGGA and REMBRANDT datasets (Supplementary Fig. S6A and B). Low expression of PCDH7 showed a poor prognosis for GBM patients in accordance with miR-155-3p, while PCDH9 didn’t show any association with prognosis for GBM patients, just like miR-155-5p (Supplementary Fig. S6C and D).

To test whether PCDH9 and PCDH7 mediate the biological function of miR-155 in glioma cells, PCDH9 and PCDH7 expression vectors were used in rescue experiments. The overexpression of PCDH9 and PCDH7 were validated by western blot (Supplementary Fig. S7). As shown in Supplementary Fig. S8A–F, overexpression of PCDH9 and PCDH7 suppressed glioma cell proliferation, migration, and invasion. These results were consistent with the effects of miR-155 reduction. Furthermore, when cells were cotransfected with miR-155-5p and PCDH9, or miR-155-3p and PCDH7, the effect of miR-155 in promoting cell proliferation, migration, and invasion was significantly attenuated. These results further suggested that PCDH9 and PCDH7 are the functional downstream targets of miR-155.

To study the underlying mechanism, we performed western blot analyses of potential mediating factors. The results revealed that overexpression of PCDH9 and PCDH7 remarkably downregulated the expression of β-catenin and cyclin D1 and increased the levels of phosphorylated β-catenin (pβ-catenin) (Fig. 4D). Moreover, concomitant overexpression of miR-155-5p/-3p and PCDH9/7 abrogated the downregulation of pβ-catenin and upregulation of β-catenin and cyclin D1 reduced by miR-155-5p/-3p. Together, these data support the idea that PCDH9 and PCDH7, which are direct targets of miR-155, function as tumor suppressor genes by inhibiting the Wnt/β-catenin signaling pathway.

Identification of NSC141562 as a Potent Inhibitor of miR-155

To screen small-molecule inhibitors of the MIR155HG/miR-155 axis, we conducted a high-throughput screening based on the 3D structure of the Dicer binding site on pre–miR-155 for small molecules targeting miR-155. The sequence of the pre–miR-155 hairpin loop (Dicer binding site on pre–miR-155) was generated from the miRBase database and then input into the MC-Fold/MC-Sym pipeline to construct a 3D model (Supplementary Figure S9A). We conducted high-throughput molecular dockings for pre–miR-155 against the 1990 National Cancer Institute/diversity compounds using the AutoDock program (Supplementary Table S2). The specific docking process is shown in Supplementary Video S1. Here, our dockings revealed 8 compounds with high-binding affinity (Supplementary Table S3). In addition, NSC141562 showed the highest affinity with pre–miR-155. Supplementary Figure S9B and C display the chemical structure of NSC141562 and its highest affinity with the best docking pose.

We next conducted MTT assays to evaluate the impact of these 8 compounds on cell viability in U87 cells. All the screened compounds inhibited glioma cell viability in a dose-dependent manner after 48 h of treatment (Supplementary Table S4). The top 3 with the lowest half-maximal inhibitory concentration values (NSC141562, NSC35450, and NSC10408) were further assessed for miR-155 inhibition efficacy in U87 and primary GBM cells. NSC141562 presented the strongest inhibition of mature miR-155 generation in a time-dependent manner, with an approximate 50% inhibition after 6 h of treatment (Fig. 5A). Additionally, NSC141562-treated glioma cells showed reduced proliferation, migration, and invasion (Fig. 5B and C and Supplementary Fig. S9D and E). Overexpression of miR-155-5p largely reversed NSC141562-induced suppressive effects on glioma cells (Supplementary Fig. S10A–D). Furthermore, we found that NSC141562 treatment increased the expression of PCDH9 and PCDH7, whereas it decreased the expression of mesenchymal transition‒associated markers, N-cadherin, vimentin, matrix metalloproteinase 9, and β-catenin (Fig. 5D). These data suggested NSC141562 as a potent small-molecule inhibitor of the MIR155HG/miR-155 axis.