MiR-212-3p functions as a tumor suppressor gene in group 3 medulloblastoma via targeting nuclear factor I/B (NFIB)

Haploinsufficiency of chromosome 17p and c-Myc amplification distinguish group 3 medulloblastomas which are associated with early metastasis, rapid recurrence, and swift mortality. Tumor suppressor genes on this locus have not been adequately characterized. We elucidated the role of miR-212-3p in the pathophysiology of group 3 tumors. First, we learned that miR-212-3p undergoes epigenetic silencing by histone modifications in group 3 tumors. Restoring its expression reduced cancer cell proliferation, migration, colony formation, and wound healing in vitro and attenuated tumor burden and improved survival in vivo. MiR-212-3p also triggered c-Myc destabilization and degradation, leading to elevated apoptosis. We then isolated an oncogenic target of miR-212-3p, i.e. NFIB, a nuclear transcription factor implicated in metastasis and recurrence in various cancers. Increased expression of NFIB was confirmed in group 3 tumors and associated with poor survival. NFIB silencing reduced cancer cell proliferation, migration, and invasion. Concurrently, reduced medullosphere formation and stem cell markers (Nanog, Oct4, Sox2, CD133) were noted. These results substantiate the tumor-suppressive role of miR-212-3p in group 3 MB and identify a novel oncogenic target implicated in metastasis and tumor recurrence.


Introduction
Medulloblastoma (MB), the most common malignant brain tumor of childhood, accounts for 20% of pediatric central nervous system (CNS) neoplasms with annual age-adjusted incidence ranging from 0.38 to 0.42 per 100,000 persons [32,39]. Wingless-type (WNT), sonic hedgehog (SHH), group 3, and group 4 are the classic subgroups of MB with distinctive clinicopathologic and molecular features [37]. The most aggressive tumors fall into group 3 (non-SHH/WNT MB), which account for approximately 25-30% of all MB cases and belong to a high-risk subgroup punctuated by haploinsufficiency of 17p (20-50% incidence), c-Myc amplification (15-20% incidence), and metastases close to or at diagnosis (30-40%), all resulting in very poor prognosis with < 50% 5-year survivorship [3,25,38,45]. Current treatment regimens involve surgical resection followed by a combination of craniospinal radiation and multi-agent chemotherapy, including vincristine, cisplatin, and either cyclophosphamide or lomustine [40,41]. Recent studies have implicated MB tumor-initiating (stem) cells in evading chemotherapeutic regimens, facilitating recurrence [15,31]. Recurrence can further reduce 5-year survival to < 10% [25,43]. Thus, there is an urgent need to understand group 3 MB pathobiology and key signaling pathways involved in disease progression and tumor recurrence to provide a more accurate risk-adapted targeted treatment approach that can mitigate the dismal survivorship of these patients.
To date, no studies have examined the role of miR-212-3p in MB pathogenesis. Given its location on a highly afflicted chromosomal locus in high-risk MB, we hypothesized that miR-212-3p possesses tumor-suppressive properties. Here, we have focused on thoroughly elucidating the anti-neoplastic properties of miR-212-3p in group 3 MB and revealed a new oncogenic target of this miR, Nuclear Factor I/B (NFIB).

Human tissue samples and molecular subgrouping of the MB tissues
Frozen and formalin-fixed paraffin-embedded (FFPE) samples of normal cerebellum (pediatric = 12, adult = 5) and pediatric MB specimens (WNT = 1, SHH = 7, group 3 = 10, group 4 = 14, unknown = 7) were collected from the Children's Hospital and Medical Center and the University of Nebraska Medical Center, Omaha. Tumor samples were sub-classified into four subgroups using genome-wide DNA methylome analysis (Illumina Methylation EPIC 850 K bead arrays) as previously described [24]. Normal cerebellum specimens were obtained at autopsy. For expression profile of HDACs, EZH2, and NFIB, we cross-analyzed our local MB dataset (Kanchan et al., GSE148390) with publicly available MB datasets (Drusco et [9,24,52]. For Kaplan-Meier Survival Analysis, we used the R2 database (Cavalli et al.,GSE85217) [3].

DNA methylation profiling
Genome-wide DNA methylome analysis was performed using the Illumina Methylation EPIC 850K bead arrays as previously described [36]. Genomic DNA was extracted from normal cerebellum and MB FFPE samples using RecoverAll ™ Total Nucleic Acid Isolation Kit (Invitrogen). Results are presented as percent methylation at each CpG measured.

Colony formation assay
After transfection/Dox treatment, MB cells (1 × 10 3 cells/well) were reseeded in 6-well plates and grown for 7-10 days in a humidified atmosphere (95% humidity) at 37 °C and 5% CO 2 . Cells were washed, fixed with 2.5% methanol and stained with 0.5% crystal violet. Cell staining was dissolved using 10% acetic acid and quantified by measuring the absorbance at 590 nm.

Cell migration and invasion assay
For transwell migration/invasion assay, transfected/Dox treated stable cells (5 × 10 5 cells) in serum-free media were seeded in the upper chamber of an insert (8 mm pore size; BD Bioscience) coated with Fibronectin (BD Bioscience) or Matrigel (Invasion Chamber Matrigel Matrix) followed by addition of a chemoattractant (10% FBS in complete media) to the lower chamber. After overnight incubation (16 h), cells that migrated/invaded into the lower chamber were stained using Diff-Quik Stain Set (Siemens Healthcare Diagnostics, Inc.); images were captured using an EVOS FL Auto Imaging System (Life Technologies). Results were quantitated by taking an average cell count, measuring cell numbers from fourfield/images/well (10 × magnification).

Wound healing assay
After transient transfection, cells (5 × 10 5 cells/well) were plated in a 6-well plate, and upon reaching 80% confluence, a vertical scratch was made using a 10 µL pipette tip. For stable cells, 3 × 10 4 cells/well were seeded in a culture-insert (ibidi culture-insert 2 well), and after overnight incubation the culture-insert was removed and washed with PBS to remove non-adherent cells. Fresh growth medium (with and without Dox (500 ng/ml)) supplemented with 5% serum was added to the plates. The wound closure area was photographed at denoted time intervals using an EVOS FL Auto Imaging System (Life Technologies). Quantitative measurements (% wound closure) were determined by measuring three fields per well.

Apoptosis and cell cycle analysis
Annexin-V/Cy ™ 5 (BD Biosciences) and propidium iodide (PI) (Roche Diagnostics) were used to measure apoptosis and cell cycle profile as previously described [24]. Briefly, miR-212-3p mimic or scramble control transfected cells were incubated for 96 h. Following incubation, cells were washed and resuspended in calcium-binding buffer consisting of Annexin-V/Cy ™ 5 and PI (apoptosis assay) or fixed with 70% ethanol and stained with PI (cell cycle analysis). Stained cells were analyzed using a FACS Canto ™ Flow cytometer (BD Bioscience).

Immunohistochemistry
Unstained tissue slides (3 µm) were deparaffinized in xylene, rehydrated in a series of alcohol dilutions, and heated in citrate buffer (pH 6.2) for 20 min. Following antigen retrieval, slides were incubated with blocking buffer (Horse serum, Vector Labs) for 60 min at room temperature. Subsequently

Orthotopic medulloblastoma mouse model
All experimental animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC). Briefly, four to six-week-old NSG (NOD-scid Il2rg null ) mice were anesthetized using ketamine and xylazine (i.p.) and transferred to a stereotactic frame (Stoelting Co, IL, USA). Hamilton syringe (26-gauge needle) with stable Tet-on miR-212 inducible HDMB03 cells (1.0 × 10 5 cells/5 µl of PBS) was implanted into the cerebellum (2-mm lateral and 2-mm posterior to lambda, and 2.7-mm deep from the surface of the skull), with a speed of 1 µl/min. Needle was removed after 5 min. After surgery, tumor growth, based on total photon flux, was measured using an IVIS Spectrum imaging system  9:195 (Caliper life sciences; PerkinElmer, MA, USA). Based on tumor size, mice were randomized into two groups (i) control and (ii) DOX treatment group; doxycycline (2 mg daily through oral gavage) induced expression of miR-212-3p. Tumor growth was monitored using an IVIS Spectrum imaging system. On appearance of first neurological symptoms (as per IACUC guidelines), animals were euthanized. Brains were fixed for analysis.

Statistical analysis
All experiments were conducted in triplicate. Values are presented as mean ± SD. Statistical analyses were performed using Prism 7.0b (GraphPad Software). For data normalization, control group was set at "1" and experimental groups were represented as fold-change compared to control with error bars reflecting deviation from mean triplicate measurements; statistical analyses were conducted prior to normalization. Differences between groups were compared using a two-tailed, unpaired Student's t-test, Mann-Whitney U test or one-way analysis of variance followed by a least significant difference post hoc test. Statistical significance was established at *p < 0.05, **p < 0.01, and ***p < 0.001.
To explore the regulation of miR-212-3p in MB, we first conducted an in silico expression analysis of miR-212-3p in CSF samples of patients with MB (NC = 14 vs. MB = 15; GSE62381), revealing reduced expression compared to normal (Fig. 1A (i)). Subsequent expression examination in our local pediatric MB cohort (GSE148390) revealed specific downregulation in group 3 (n = 9) and group 4 (n = 13) MB tumors (Fig. 1A (ii)). We cross validated these results using RT-PCR, noting near-abrogation of expression of miR-212-3p in group 3 MB (n = 10) and group 4 MB samples (n = 12) compared to pediatric cerebellum (n = 10) (Fig. 1B). In vitro, we recapitulated these findings in a panel of MB cell lines with characteristic high-risk features, including i17q and c-Myc amplification (group 3 MB-D341, D425, HDMB03; group 3/4 MB-D283). These were in stark contrast to normal human astrocytes (NHA) and an SHH MB-type cell line, i.e. Daoy, with high expression of miR-212-3p (Fig. 1C). These results substantiated a decreasedto-absent expression of miR-212-3p specific to group 3 and 4 MB tumors and cell lines. We thus chose to focus the rest of our study on group 3 tumors.
To elucidate a mechanism for expression silencing, we studied epigenetic mechanisms, i.e. hypermethylation vs. histone modification. We initially performed DNA methylation profiling in our group 3 MB patient samples (n = 6) but found no perturbation to the methylation of the miR-212-3p promoter region when compared to normal pediatric cerebellum (n = 4). In vitro, these findings were supported by a lack of expression restoration in HDMB03 cells treated with the global demethylating agent, 5-AzaC (5 µM, 96 h) (Additional file 2: Fig. S1B).
Several studies have linked aberrant methylation and acetylation of key histones involved in chromatin structure, including H3K4 and H3K27, with enhanced disease aggressiveness mainly in group 3 and group 4 tumors [10,22]. Moreover, studies targeting histone deacetylases (HDACs) have revealed growth inhibition of MYC-driven medulloblastomas [11,12,42]. Thus, we shifted our focus to histone modification-mediated epigenetic regulation. Our initial in silico analysis of HDAC expression revealed high expression of HDAC 1, 3, 5, 9, D425) were compared with normal human astrocytes (NHA) and an SHH MB cell line (Daoy). CHIP-qRT-PCR analysis revealed important differences in the methylation status of H3K27 and H3K9 and in the acetylation status of H3K9 in group 3 MB cell lines compared to NHA and Daoy cells (Fig. 1E). Specifically, HDMB03 and D425 cells, with baseline reduced miR-212-3p expression, showed enriched methylated H3K27 and H3K9, with a concomitant decrease in acetylated H3K9, compared to NHA. In contrast, Daoy cells, with high intrinsic miR-212-3p expression, were hypomethylated at H3K27 and showed no change to the acetylation pattern of H3K9 compared to NHA. Consequently, treatment of HDMB03 and D425 cell lines with pan-HDAC inhibitors (TSA, 100 nM; Belinostat, 1 µM; Vorinostat, 1 µM) substantially increased miR-212-3p expression compared to vehicle control ( Fig. 1F and Additional file 2: Fig. S1D). Silencing EZH2 expression (siRNA-EZH2, 20 nM) accomplished the same (Fig. 1F). Together, these findings strongly implicated a unique pattern of histone modification as a silencing mechanism for miR-212-3p in group 3 MB tumors.

MiR-212-3p expression restoration inhibits group 3 MB cell growth and proliferation
To highlight the tumor-suppressive properties of miR-212-3p in group 3 MB, we employed two methods to restore miR-212-3p expression, i.e. by transient transfection with miR-212-3p mimic or by Dox-inducible stable expression in group 3 MB cell lines. With successful miR-212-3p expression restoration ( Fig. 2A), cancer cell growth and proliferation were significantly impacted in a time-and dose-dependent manner (Fig. 2B). Similarly, colony formation, transwell migration, and wound closure assays all recapitulated this anti-neoplastic phenotype with reduced colonies, cellular migration, and wound closure, respectively ( Fig. 2C-E). Although overexpressing miR-212-3p in Daoy cells decreased cell proliferation and colony formation, a dose-and timedependent effect was absent. Instead, growth inhibitory effects seemed to plateau within 24 h and with a 25 nM concentration of miR-212-3p, suggesting that SHH cell lines do not experience the same growth inhibition as group 3 cell lines do (Additional file 3: Fig. S2A-C). These data provided compelling in vitro evidence for the tumor-suppressive properties of miR-212-3p, expressly in group 3 MB.

MiR-212-3p induces cell cycle arrest and destabilizes c-Myc to favor apoptosis in group 3 MB
Myc amplification is a cardinal high-risk feature of group 3 MB, and its phosphorylation status influences downstream tumor phenotype [43,44]. More specifically, c-Myc phosphorylated at serine 62 increases stability leading to tumor aggressiveness, while phosphorylation at threonine 58 destabilizes the protein, leading to ubiquitin-mediated degradation and subsequent cellular apoptosis [14,23]. Transient transfection of HDMB03 cells with miR-212-3p resulted not only in a reduction in total c-Myc protein but also a concomitant increase in the ratio of phosphorylated T58 to phosphorylated S62 (Fig. 3A). Concurrently, the active upstream kinases responsible for c-Myc stabilization via phosphorylation at S62, i.e. p-Erk and p-Akt [46], were both significantly decreased in these cells (Fig. 3A). In miR-212-3p stablyexpressing HDMB03 cells, c-Myc expression was almost completely abrogated, with consequent reductions in both phosphorylated species of c-Myc. As prior, p-AKT and p-ERK were also significantly reduced. Together, these data purported a c-Myc deregulatory function for miR-212-3p. Of note, in gastric cancer, c-Myc has been revealed as a direct target of miR-212 [55]. We then focused on progression through the cell cycle, given the destabilization of c-Myc and a reduced proliferative phenotype noted in the presence of miR-212-3p. Not surprisingly, we revealed arrest in transientlytransfected miR-212-3p mimic-treated cells (25 nM and 50 nM) at the G 0 /G 1 phase of the cell cycle in HDMB03 cells (Fig. 3B). We confirmed the phase of arrest using Western blotting, which showed decreased expression of the complementary checkpoint markers, CDK4, CDK6, and cyclin D1 ( Fig. 3A and C). In support, hierarchical clustering and pathways analysis revealed that miR-212-3p expression restoration led to the enrichment of gene clusters (4,5,8) involved in the regulation of cell cycle phase (Additional file 4: Fig. S3A and B, and Additional file 5: Table S1).
We concluded our study of the anti-neoplastic properties of miR-212-3p by analyzing its effect on cancer cell apoptosis. We first examined the expression of c-Myc binding partners that signal for apoptosis, i.e. Bin-1 and p19 ARF , along with various pro-apoptotic proteins. MiR-212-3p restoration in HDMB03 cells significantly increased expression of Bin-1 and p19 ARF concurrent with rises in cleaved PARP and cleaved caspase 3 (Fig. 3D). These results were validated using Annexin V-cy5/PI staining, demonstrating a two-and threefold increase in late and early apoptosis, respectively, compared to control (Fig. 3E).
Taken together, our results identified an apoptotic mechanism for miR-212-3p, either by altering c-Myc phosphorylation states to destabilize c-Myc, by reducing the total c-Myc expression, and/or by increasing the expression of its pro-apoptotic binding partners. In parallel, miR-212-3p played a role in cell cycle arrest at the G 0 /G 1 phase.

NFIB is a downstream target of miR-212-3p
To identify oncogenic targets of miR-212-3p, we started with the targets that are common to miRNA target prediction databases (TargetScan and mirDIP) and that are significantly downregulated (Log 2 fold change < −1.0, p < 0.05) by miR-212-3p expression restoration in HDMB03 cells (Fig. 4A). This comparison revealed 37 putative targets whose expression and associated pathways were studied in two group 3 MB patient cohorts (Fig. 4B (i) and (ii), and Additional file 6: Fig. S4A and B). Out of the 37 common targets, fourteen were upregulated in group 3 MB (summarized in Additional file 7: Table S2). NFIB was chosen for further study based on the following characteristics that bestow a high likelihood of oncogenic potential in group 3 MB: (i) there are two conserved binding sites for miR-212-3p on the 3′UTR of NFIB mRNA (TargetScan); (ii) expression of NFIB was significantly increased across multiple MB datasets ( (Fig. 4F); and (vi) immunohistochemical staining of NFIB in group 3 MB tissues (n = 9) revealed intense nuclear staining when compared to normal cerebellum (n = 8) (Fig. 4G). We further validated this target by the dual-luciferase assay (Fig. 4H) and showed significant attrition in miR-212-3p mimic treated HDMB03 cells at the transcription (Fig. 4I) and translation levels (Fig. 4J). In this manner, we identified NFIB as an oncogenic target of miR-212-3p.

NFIB possesses oncogenic potential in group 3 MB cancer cells
Nuclear Factor I/B (NFIB) belongs to the nuclear factor I (NFI) family of transcription and replication proteins that recognize palindromic sequences on various promoters capable of activating transcription and replication throughout organ development [16,49]. An oncogenic role for NFIB has been shown in triple-negative breast cancer (TNBC), small cell lung cancer (SCLC), colorectal and gastric cancers, and melanoma by enhancing tumor growth, epithelial-mesenchymal transition (EMT), migration, and invasion [13,29,30,47,53]. With this prior evidence, we elucidated the oncogenic role of NFIB in group 3 MB cancer cells by studying the effect of silencing expression on proliferation, transwellmigration, and invasion. Silencing NFIB in HDMB03 cells transiently resulted in a significant reduction in cell proliferation up to 48 h (Fig. 5A). Transwell migration and invasion assays demonstrated similar reductions in cell migration and invasion (Fig. 5B).
Mechanistically, NFIB overexpression has been shown to increase chromatin accessibility, promote the expression of pro-metastatic genes, and drive metastasis in SCLC tumors [8,47,54]. Studies in SCLC have also revealed NFIB as a downstream target of c-Myc, which directly regulates its expression and contributes to tumor aggressiveness and rapid metastases [35]. However, we have shown that miR-212-3p has a deregulatory effect on c-Myc. To delineate if NFIB inhibition plays a role in Myc regulation, we silenced NFIB in group 3 MB cells using transient siRNA-NFIB treatment. As with miR-212-3p stable expression, NFIB silencing decreased total c-Myc levels with a concomitant decrease in phosphorylation of c-Myc at S62 and T58; moreover, upstream c-Myc regulators i.e. p-Akt and p-Erk, were also significantly reduced (Fig. 5C).
Given the direct inhibitory effect of miR-212-3p on c-Myc and NFIB's promoting role in metastasis, we sought to study the effect of the miR-212-3p/NFIB axis on cancer stem cell (CSC) maintenance and self-renewal. We first analyzed the expression of stem cell markers in miR-212-3p stably-expressing and NFIB-silenced HDMB03 cells. RT-PCR and Western blotting analyses revealed concurrent de-regulation in the expression of several stem cell markers, including CD133, β-catenin, Sox-2, Oct-4, and Nanog ( Fig. 5D and E). CSCs selfrenewal ability was further assessed through tumor sphere-forming assay. Again, in both NFIB-silenced and miR-212-3p stably-expressing cancer cells, substantially decreased numbers of spheres were noted ( Fig. 5F and G). These results further elucidated the anti-neoplastic properties of miR-212-3p with its inhibitory role on CSC maintenance and implicated NFIB as a strong oncogenic target of miR-212-3p in group 3 tumors.

Restoration of miR-212-3p expression inhibits the growth of orthotopic group 3 MB tumors
To determine the role of miR-212-3p in group 3 MB tumor growth and progression in vivo, we implanted stable miR-212-inducible HDMB03 cells into the cerebellum of mice as detailed in Fig. 6A. MiR-212-3p expression was restored in mice via daily doxycycline supplementation (oral gavage). At day 16, Dox-fed mice showed significant reduction in tumor growth compared to control ( Fig. 6B and C). Moreover, Kaplan-Meier survival analysis revealed significantly higher survival in the Dox-fed group compared to control (Fig. 6D). Tumor histology revealed smaller tumor margins in Dox-fed mice compared to control (Fig. 6E). In addition, Dox-fed mice developed tumors with substantially reduced staining for Ki-67, c-Myc, and NFIB with a concomitant increase in staining for cleaved caspase-3 (Fig. 6F). These in vivo results substantiated the anti-proliferative, pro-apoptotic, and tumor-suppressive properties of miR-212-3p (Fig. 6G) in group 3 MB tumors.

Discussion
Haploinsufficiency of chromosome 17p bestows a highrisk phenotype upon group 3 tumors [3,25,45]. Several loci populated by tumor suppressor genes can be found within the afflicted short arm. Studies have linked few genes residing on the terminal locus (17p13.3) with MB, such as ROX/MNT [7] or HIC1 [51]. None have studied microRNAs on this locus. We previously elucidated the anti-neoplastic properties and oncogenic targets of miR-1253, a microRNA found on the terminal part of this locus, in MB [24]. In the present study, we have described the tumor-suppressive properties of miR-212-3p in group 3 MB; we have additionally identified a target with strong oncogenic potential. An initial in silico analysis of a publicly available MB dataset revealed significantly reduced expression for miR-212-3p compared to normal. These findings were specifically recapitulated in non-SHH/WNT tumors both in pediatric samples (ex vivo) and in classic MB cell lines (in vitro). Focusing on the higher risk group 3 tumors, we sought a mechanism for miR-212-3p silencing. In many cancers, microRNAs undergo epigenetic silencing via hypermethylation along CpG islands or chromatin rearrangement through histone modifications [24,26]. Methylation profiling in high-risk tumors showed no differences in methylation pattern of the miR-212-3p locus between tumors and normal tissue. Contrarily, histone modifications at critical lysine residues within miR-212-3p provided compelling evidence for an epigenetic silencing mechanism. More specifically, increased methylation at H3K27 and H3K9 and a concomitant decline in acetylation of H3K9 was shown by ChIP-RT-PCR. Either treatment with HDAC inhibitors or EZH2 silencing reliably restored miR-212-3p expression. Together, these data revealed miR-212-3p silencing in group 3 tumors and assigned an epigenetic mechanism via histone modifications. Our findings aligned with prior reported patterns of miR-212-3p epigenetic silencing in lung cancer [19].
Restoring miR-212-3p expression in group 3 MB cells by either transient transfections or stable induction resulted in a cadre of anti-neoplastic effects. First, cancer cell proliferation, wound healing, migration, and colony formation were significantly reduced. Subsequently, cell proliferative markers, p-Akt and p-Erk, were downregulated in miR-212-3p restored cells, eventually leading to cell cycle arrest (G 0 /G 1 cell cycle phase) with dysregulated expression of checkpoint regulatory kinases, CDK4, CDK6, and cyclin D1. Next, we revealed regulatory effects of miR-212-3p on c-Myc phosphorylation states, with transient expression resulting in a shift in phosphorylation patterns from S62 (active, favoring proliferation) to T58 (inactive, favoring apoptosis), and stable expression leading to a near-abrogation in total c-Myc expression. Of note, p-Akt and p-Erk, which were downregulated by both transient and stable miR-212-3p expression, are upstream kinases that phosphorylate c-Myc at S62, rendering it resistant to degradation [14,23]. Moreover, in miR-212-3p-expressing cells, key apoptotic binding partners of c-Myc, p19 ARF and Bin-1, were elevated in tandem with cleaved PARP and cleaved caspase-3, resulting in a high apoptotic signal in cancer cells. As prior, our results aligned with studies on miR-212-3p in colorectal cancer [34], nasopharyngeal carcinoma [20], and lung cancer [21] where similar effects were seen on cancer cell proliferation, migration and invasion, cell cycle arrest, and apoptosis. Given the high association between c-Myc amplification and poor prognosis in group 3 MB [3,45], our data suggested a plausible mechanism contributing to high-risk MB aggressiveness, wherein silencing miR-212-3p may lift its regulation of c-Myc, allowing for unchecked Myc-driven signaling.
We next identified an oncogenic target of miR-212-3p using in silico, ex vivo, and in vitro approaches. In so doing, we revealed Nuclear Factor I/B (NFIB), a transcription factor that can bind to specific overlapping DNA repeat sequences (5′-TTGGCNNNNNGCCAA-3′) throughout the genome to activate transcription and replication, most notably in normal lung and brain development [16,49]. In fact, the 14 target genes of miR-212 isolated via high-throughput analyses of independent MB datasets seemed most intricately involved in these processes (Additional file 6: Fig. S4B). Given the association of high NFIB expression in group 3 tumors with poor survival, we showed that silencing NFIB expression dramatically decreased cancer cell proliferation, migration, and invasion in vitro. Similar effects have been shown for NFIB in gastric cancers [53], TNBC [29], SCLC [47], and colorectal cancer [30]. In colorectal cancer, melanoma, and gastric cancers, overexpression of NFIB was associated with epithelial-mesenchymal transition (EMT), migration and invasion [13,30,53]. In melanoma, specifically, NFIB-targeted upregulation of EZH2 led to the epigenetic silencing of MITF, promoting a highly invasive phenotype [13]. In SCLC, NFIB overexpression cooperated with Rb/p53 deletion to increase chromatin (See figure on next page.) Fig. 5 Effect of NFIB silencing on cancer cell phenotype in group 3 MB. A Transient silencing of NFIB (siRNA-NFIB, 20 nM) showing significant reduction in HDMB03 cell viability compared to siRNA-control cells. B Transwell migration/invasion assay recapitulating the same with decreased migrating and invading cells in siRNA-NFIB-treated cells compared to control. C Western blotting analysis of c-Myc showing reduction in total c-Myc, p-S62, p-T58 and reduction in upstream activators of c-Myc, i.e., p-Akt and p-Erk. RT-PCR and Western blotting analyses demonstrating reduced stem cell markers, i.e. Nanog, Oct4, Sox2, CD133 and β-catenin, in D siRNA-NFIB-treated HDMB03 cells and E Dox-treated stably-expressing miR-212-3p HDMB03 cells. Medullosphere assays demonstrating dramatically reduced sphere formation in F siRNA-NFIB-treated and G Dox-treated stably-expressing miR-212-3p HDMB03 cells. β-actin served as an internal loading control. Data presented as mean ± SD from experiments done in triplicate and analyzed by Student's t-test; *p < 0.05; **p < 0.01; ***p < 0.001. Scale bar: 100 μm   9:195 accessibility to pro-metastatic genes. Moreover, c-Myc was shown to regulate NFIB in SCLC and further contribute to rapid metastases [35]. Metastases at diagnosis and tumor recurrence are cardinal prognostic features contributing to dramatically increased mortality in group 3 MB [3,25,38,[43][44][45]. MB recurrent tumors, either in the primary or metastatic site, derive from a subpopulation of cancer stem cells (MBSCs), which can evade current chemotherapeutics and radiation therapy [1]. MBSCs express stemness markers, such as CD133, CD15, and Sox2, and can possess an inordinate capacity to form aggressive tumors with increased self-renewal ability, facilitating MB relapse and rapid demise [48,50]. In NFIB-silenced and miR-212-3p restored group 3 MB cells, we noted decreased expression of stemness markers, i.e. CD133, Sox2, Oct4, Nanog, and β-catenin, concurrent with reduced tumor cell self-renewal capacity as evidenced by sphere-forming assays. These results introduce miR-212-3p's role in hampering MBSC maintenance and self-renewal, possibly through NFIB regulation.
These in vitro findings were translated into an orthotopic mouse model wherein miR-212-3p induction led to significantly smaller tumors and substantially elevated survival. Moreover, miR-212-3p-expressing tumors exhibited decreased staining for Ki-67, c-Myc, and NFIB and elevated staining for cleaved caspace-3, strengthening our hypothesis of the tumor-suppressive role of miR-212 in group 3 MB tumors.
Intriguingly, we have shown EZH2 as an upstream regulator of miR-212-3p and revealed miR-212-3pmediated destabilization of c-Myc. At the core of these critical contributors to tumor aggressiveness lies NFIB, a downstream target of c-Myc (SCLC [35]) and an inducer of EZH2 (melanoma [13]). Whether NFIB plays a role in inducing EZH2 in group 3 tumors to epigenetically silence miR-212-3p, in turn stabilizing c-Myc, is an important mechanism that is being presently evaluated in detail.

Conclusions
This study has uncovered a novel tumor suppressor gene in group 3 MB, i.e. miR-212-3p. We have shown expression silencing by histone modifications, as opposed to hypermethylation. The anti-neoplastic properties of miR-212-3p were exhibited by destabilization of c-Myc, cancer cell arrest at G 0 /G 1 cycle, and robust apoptosis, resulting in attrition in cancer cell invasiveness and tumor growth with improved survival in vivo. By targeting NFIB, a wellstudied metastatic driver [8], a regulator of EZH2 [13], and downstream effector of c-Myc [35], miR-212-3p decreased cancer cell aggressiveness, stem cell maintenance, and renewal. Our studies provide support for the use of miRNA-based therapies as a targeted approach to not only addressing group 3 MB tumor aggressiveness but also unburdening young patients from the harmful side-effects of current cytotoxic therapies.