MicroRNA 128a Increases Intracellular ROS Level by Targeting Bmi-1 and Inhibits Medulloblastoma Cancer Cell Growth by Promoting Senescence

1Medulloblastoma is the most common malignant brain tumor of childhood. While outcomes have improved, there is significant therapy-related morbidity,. In addition, patients with high-risk features continue to have a poor prognosis. Recent advances indicate that medulloblastoma arises from cerebellar granule cell precursors or neural stem cells located in the cerebellum,,,. While the molecular mechanisms involved in medulloblastoma tumorigenesis are not well defined, it is clear that there is abnormal control of normal developmental mechanisms.

2Recently work from our lab and others have implicated microRNAs as important regulators of medulloblastoma cell growth,,. MicroRNAs consist of 18–22 nucleotide RNA molecules with post-transcriptional gene silencing activity. Most commonly they control gene expression through association with the 3′-untranslated region (3′ UTR) of genes and inhibit protein translation. MicroRNAs can also destabilize and mediate the degradation of RNA transcripts. In addition to their role in normal development, microRNAs are also associated with carcinogenesis,. Many microRNAs are under expressed in human tumors compared to normal tissues, while some are over expressed. Importantly, dysregulation of microRNA processing results in enhanced tumorigenesis. In addition, a growing number of microRNAs are associated with specific human cancers. For example, microRNA 21 (miR-21) is over expressed in glioblastoma, and inhibition of miR-21 inhibits glioblastoma growth in vitro and in vivo. Others have shown that miR-137 and miR-124 induces differentiation of glioma stem cells. We have previously shown that miR-124 also acts as a tumor suppressor in medulloblastoma cells while other recent studies have implicated the miR 17-92 polycistron as an oncogene in sonic hedgehog mediated medulloblastoma,,.

3These and other recent studies suggest that microRNAs are critical regulators of tumorigenesis and that their contribution to medulloblastoma tumorigenesis is important. Therefore, we decided to investigate microRNA function in medulloblastoma. We explored the possibility that expression of brain-enriched microRNAs is altered in medulloblastoma and that some of these microRNAs may play a critical regulatory role in this tumor. We found that several microRNAs associated with normal neuronal differentiation are significantly down regulated in medulloblastoma. Of these microRNA 128a (miR-128a) was strongly down regulated. While the role of miR-128a has been investigated in astrocytic tumors such as glioblastoma, its role in neuronal tumors such as medulloblastoma is not known. Here we describe the functional role of miR-128a in medulloblastoma for the first time. We found that miR-128a inhibits growth of medulloblastoma cells by targeting the Bmi-1 oncogene and thereby increasing the steady-state levels of superoxide and promoting cellular senescence.

4As a first step to addressing the role of microRNA in medulloblastoma we performed microarray-based analysis of microRNA in medulloblastoma cells and compared it to normal adult human cerebellum. We chose to examine primary medulloblastoma cell explants at passage 2 in culture to avoid noise from contaminating elements of normal brain in biopsy samples. Out of the 385 human microRNAs assayed, 167 were expressed at lower levels in medulloblastoma when compared to normal cerebellum. Hierarchical clustering of the microRNA expression correctly separated the normal cerebellum from the medulloblastoma samples (Figure S1-A). Further analysis revealed that decreased expression of 90 microRNAs in medulloblastoma were statistically significant (p<0.05). This list was then examined to separate the microRNAs that were expressed in all three cerebellar samples and statistically decreased in all three medulloblastoma samples. This narrowed the list of statistically significant human microRNAs with decreased expression in all three medulloblastoma samples to 30 microRNAs (Figure S1-B). Of note, many of the microRNAs had previously been identified as being enriched in normal brain. Some but not all of these microRNAs were also reported previously as being decreased in medulloblastoma as compared to normal cerebellum,. Comparison of the data sets of decreased microRNAs reported previously and our data revealed that only five microRNAs were detected by all three groups (Figure S2). These are miR-124, miR-129, miR-138, miR-150 and miR-323. We have previously shown that miR-124 is functionally important in medulloblastoma while the biology of the other four common miRs is not yet clear. Ferretti et al had an additional 12 microRNAs in common with our data set where as Northcott et al had only 3 additional microRNAs in common with us (Figure S2). Furthermore there were 10 additional microRNAs that were only identified by us as being significantly decreased in medulloblastoma (Figure S2).

5We next performed real time RT-PCR on a cohort of these miRNA in additional samples to validate our microarray data (Figure 1A). Comparing primary medulloblastoma cells to both adult and pediatric normal cerebellum revealed a significant down regulation of brain enriched microRNAs in medulloblastoma. There are four highly down regulated microRNAs namely, miR-125, miR-128a, miR -139 and let-7g. Of these four microRNAs, miR-139 was identified by us but not in two previous reports,. Furthermore we as well as Ferretti et al but not Northcott et al detected miR-128a as decreased in medulloblastoma. Interestingly adult cerebellum had higher expression of many of these microRNAs when compared to pediatric cerebellum. The expression of highly repressed microRNAs was further validated in a panel of medulloblastoma cell lines. Similar to the primary explants all four microRNAs were decreased in the cell lines when compared to normal cerebellum (Figure 1B). Consistent with previously published data we also found miR17-5p to be over-expressed in medulloblastoma.

6Next we examined the expression of let-7g, miR-125 and miR-128a in primary medulloblastoma tumors and normal cerebellar tissue. Using qRT- PCR we found that all three miRNAs are significantly decreased in expression in 10 archived medulloblastoma patient samples (ANOVA, p<0.001, Figure 1C). Given the small sample set, it is difficult to develop any correlation with tumor sub-type or patient outcomes. However the samples were not Gli1 high and thus not in the SHH subcategory,. These data indicate that these microRNAs could have an important biological role in medulloblastoma. Based on the extent of miR-128a repressed in medulloblastoma in contrast to its high expression in normal cerebellum, we chose to further investigate the functional role of miR-128a in medulloblastoma.

7To determine whether re-expression of microRNAs alters tumor cell growth, we transfected Daoy medulloblastoma cells with microRNA precursor oligonucleotides. These microRNA precursor molecules are designed to mimic endogenous microRNAs. Re-expression of miR-128a decreased medulloblastoma cell growth as measured by the MTT assay (Figure 2A). Similar findings were noted in the D283 medulloblastoma cell line (data not shown). To further evaluate the growth inhibitory effect of miR-128a, we transfected Daoy cells with control miR or miR-128a oligonucleotides and counted cells for 5 days using the trypan blue dye exclusion method. Consistent with the MTT data, miR-128a decreased Daoy cell growth in a dose-dependent manner (Figure 2B).

8To better evaluate the impact of microRNAs on medulloblastoma cells we used a lentiviral plasmid, which expresses the miR-128a cassette and performed colony formation assays. As shown in Figure 2C and D, transfection with miR-128a potently inhibited colony formation by medulloblastoma cells indicating that miR-128a is a putative tumor suppressor in medulloblastoma. In addition miR-128a potently decreased the ability of Daoy cells to grow in soft agar further suggesting a tumor suppressive role for miR-128a in medulloblastoma (Figure S3).

9We next performed an analysis of potential microRNA target sites using two commonly used prediction algorithms, TargetSCAN (http://www.targetscan.org) and PicTar (http://pictar.bio.nyu.edu),. Both algorithms predicted the polycomb gene Bmi-1 to be a target for miR-128a. The target site meets the seed match criteria (Figure 3A). To experimentally test whether our predicted targets were regulated by miR-128a, the Bmi-1 target site was cloned in the 3′UTR of the Renilla luciferase gene in the siCHECK vector. Daoy cells were transfected with control or miR-128a oligonucleotides. Co-transfection with miR-128a significantly decreased Renilla luciferase activity in the Bmi-1 target site vectors but not in the mutated site vectors (Figure 3B). This data suggests that Bmi-1 is a target of miR-128a. One other recent study of miR-128a in glioma also reported that Bmi-1 is a target for miR-128a. This is especially exciting since polycomb genes play a vital role in stem cell renewal during embryogenesis and are also known to be biologically important in neural cell proliferation and medulloblastoma pathogenesis–.

10To further validate Bmi-1 as a target of miR-128a, we performed immunoblot analysis of control or miR-128a transfected cells. Daoy cells were transfected with control or miR-128a oligonucleotides and harvested 48 hours later. Western blot analysis was performed using anti-Bmi-1 antibody. MiR-128a decreased Bmi-1 protein levels in Daoy cells consistent with the luciferase data. (Figure 3C, D).

11Bmi-1 is known to represses p16 expression. In the absence of Bmi-1, p16 is up regulated in neuronal stem cells reducing the cell proliferation. Also, it has been shown that Bmi-1 mediated repression of p16 can lead to increased aggressive behavior of melanoma stem cells. Daoy and ONS76 medulloblastoma cells transfected with miR-128a up-regulated p16 as detected by western blotting (Figure 4A). The same cell lines when transfected with Bmi-1 completely repressed p16 expression. To further confirm that miR-128a prevents the repression of p16 by inhibiting Bmi-1, both Daoy and ONS76 cell lines were co-transfected with miR-128a and Bmi-1. This resulted in a moderate increase in p16 protein level compared to that of cells transfected only with Bmi-1.

12Previous studies indicate that Bmi-1 inhibits p16 and thereby enhances E2F1 activity. The E2F transcription factors play a key role in the regulation of cellular proliferation and terminal differentiation. Therefore the effect of miR-128a on E2F1 transcriptional activity was examined. It was found that transfection of miR-128a into Daoy cells significantly decreased E2F1 activity as measured by luciferase reporter activity, p<0.05 (Figure 4B). This is consistent with our findings that miR-128a decreases Bmi-1 and increases p16.

13To further evaluate whether Bmi-1 repression is a functional component of the miR-128a growth arrest phenotype we investigated whether Bmi-1 could rescue the miR-128a growth arrest in medulloblastoma cells. Medulloblastoma cells were transfected with miR-128a, miR-128a and Bmi-1 or the corresponding controls and the cells subjected to the colony focus assay. Co-transfection of miR-128a with Bmi-1 resulted in attenuation of the miR-128a mediated growth arrest in both Daoy cells (Figure 5A and B). Daoy cells co-transfected with Bmi-1 lacking its 3′UTR and miR-128a also rescued the proliferation of cells as inhibited by miR-128a alone (Figure S4A). This clearly suggests that miR-128a targets Bmi-1 and decrease the cell survival. Another medulloblastoma cell line ONS76 cells also showed the same trend in plating efficiency when transfected with miR-128a and Bmi-1 (Figure S4B). This data functionally places Bmi-1 in the miR-128a cascade.

14Recently it has been shown that the absence of Bmi-1 leads to increased reactive oxygen species (ROS) levels in cells derived from Bmi-1 knockout mice. ROS are well known to modulate a variety of cellular functions including cancer cell biology. Given that miR-128a decreased Bmi-1 protein levels, we wanted to test whether a similar rise in ROS levels was evident in Daoy cells treated with miR-128a. We performed spin trapping of free radicals and electron paramagnetic resonance (EPR) spectroscopy to identify whether there was an increase of the ROS level in cells re-expressing miR-128a. The spin trap, DMPO was chosen because of its low cytotoxicity, accessibility to the cell, and reaction with hydroxyl radicals (·OH) to yield a distinctive DMPO-OH spin adduct. The reaction of DMPO with superoxide yields a product (DMPO-OOH) that is converted to DMPO-OH by GPx in cells. Therefore, DMPO is a useful probe for detecting oxygen-centered radicals arising from superoxide and H2O2. The EPR spectra show a typical 1∶2∶2∶1 DMPO-OH quartet (closed circles, Figure 6A-ii). EPR results clearly show an increase in the DMPO-OH quartet intensity in Daoy cells expressing miR-128a compared to that of the empty vector (Figure 6A-i). To identify precisely whether the DMPO-OH quartet signal could be due directly to superoxide or to H2O2, Daoy cells transfected with miR-128a were incubated with CuZnSOD for 30 min prior to the addition of DMPO. Addition of SOD significantly decreased the EPR signal height suggesting that the DMPO-OH signal is directly due to superoxide (Figure 6A-iii). The quantitative analysis of the quartet signal height (Figure 6B) demonstrated a significant increase (p<0.05) of superoxide in cells transiently transfected with miR-128a when compared to cells treated with the control vector. This increase in signal intensity was inhibited by the addition of SOD. It is therefore clear that miR-128a increases the steady-state level of superoxide in medulloblastoma cells. To further confirm that the increase in ROS is due to the inhibition of Bmi-1, Daoy cells were co-transfected with Bmi-1 and miR-128a. This resulted in a decreased ROS level in cells compared to that of miR-128a alone (Figure 6A-iv). Rescue experiments with Bmi-1 lacking its 3′UTR also revealed that miR-128a targets Bmi-1 and thus leads to increase in ROS level in cells (Figure S5)

15Reactive oxygen species, a byproduct of oxidative stress can induce irreversible cell growth arrest and senescence. Furthermore senescing cells have significantly more ROS when compared to apoptotic cells. In addition p16 expression is one of the hall marks of cellular senescence. Given that we found miR-128a transfected Daoy cells showed increased in the steady state level of ROS, increased in p16 protein level and demonstrated a cell growth arrest, we performed an assay for senescence. Five days after transfection with miR-128a, an increased number of senescence-associated β-galactosidase (SA-β-Gal) positive cells were observed (Figure 7A). There was a five-fold increase in senescent cells after miR-128a transfection of cells when compared to the empty vector transfected cells (Figure 7B) Similar data was observed in ONS 76 medulloblastoma cells (Figure S6).

16To further support our finding that miR-128a transfection leads to cellular senescence, we examined the methylation levels of histone 3, lysine 9 (H3K9) by western blotting. Senescence-associated heterochromatin foci are associated with methylation of histone 3 lysine 9 (H3K9me2). Interestingly, re-expression of miR-128a resulted in increased methylation of histone 3 lysine 9 (H3K9me2), another mark of repressed gene expression mediated by the Bmi-1 polycomb repressor complex (Figure 7C). All these results strongly suggest that the growth inhibitory effect of miR-128a is due to the senescence-signaling pathway as triggered by the increase in ROS

17Here we report the novel regulation of reactive oxygen species by microRNA 128a via the specific inhibition of the Bmi-1 oncogene. The functional role of microRNA 128a in medulloblastoma has not previously been described.

18We found significant down regulation of several microRNAs known to be involved in CNS development in medulloblastoma,. Our data is in line with previous studies that have described decreased expression of microRNAs in medulloblastoma,. Ferretti et al found 55 microRNAs to be down regulated in medulloblastoma with differences in expression among the major histological subtypes. Similarly Northcott et al found 61 microRNAs to be decreased in medulloblastoma when compared to normal cerebellum. Interestingly when further classified into 4 molecular groups there were significant differences in microRNA expression. For example the Sonic Hedge Hog (SHH) driven tumors had decreased expression of 10 microRNAs but none that overlapped with microRNAs identified by us or by Ferretti et al. Indeed comparison analysis revealed significant differences between the microRNAs detected by the three groups (Figure S2). These differences are likely related to the platforms used by each group, tumor source and the source and age of the normal cerebellum used. We used primary cell explants in culture at passage 2 while the other two groups used biopsy material. Furthermore our patient tissues were all of the classical histology, and did not have a gene expression signature for SHH. Of note both our study and the report by Ferretti et al. found that miR-128a was significantly decreased in expression in medulloblastoma when compared to normal cerebellum.

19We show that this decreased expression is a property of primary tumor samples as well as a panel of commonly used medulloblastoma cell lines. This data will be of particular use as microRNA function is further probed in medulloblastoma. Of note the miR17-5p cluster was over expressed in our samples, which is consistent with previous reports.

20The analysis of re-expressing miR-128a in medulloblastoma cells showed that miR-128a inhibited the growth of medulloblastoma cells most likely by decreasing proliferation. MiR-128a demonstrated tumor suppressive effects by potently inhibiting the colony formation of medulloblastoma cells. Further analysis of possible mechanisms revealed that the Bmi-1 oncogene was a putative target of miR-128a. We demonstrated that Bmi-1 protein is down regulated by miR-128a, which in turn leads to an increase in p16 a cell cycle inhibitor. Regulation of Bmi-1 by microRNAs is exciting given that Bmi-1 is over-expressed in medulloblastoma and critical for normal cerebellar development. Bmi-1 is also critical for neural stem cell self-renewal. Bmi-1 was also recently shown to be a target of miR-128a in glioblastoma. Our data extends this observation by demonstrating that inhibition of Bmi-1 is functionally critical for the miR-128a growth suppressive function. Restoring Bmi-1 expression in miR-128a expressing cells rescues the medulloblastoma cells from growth arrest. We then sought to further explore the mechanism of the miR-128a-Bmi-1 pathway in medulloblastoma.

21Recent data suggests that Bmi-1 regulates reactive oxygen species. We show that microRNA 128a induces intracellular superoxide generation, possibly by regulating Bmi-1 levels and that Bmi-1 re-expression reverses the superoxide generation. Recent evidence suggests that cancer stem cells are more resistant to therapy due to a lower overall redox state. Thus modulating regulatory mechanisms of ROS generation in cancer stem cells maybe a useful strategy to destroy these cells. We envision a scenario in which microRNA 128a can be used to induce ROS in medulloblastoma stem cells and thereby make them more radiosensitive. The key is to modulate the homeostasis of ROS. At normal concentrations, ROS play a role in cellular functions involving signal transduction. However, an imbalance between generation of ROS and capacity of antioxidants to neutralize ROS can result in a disruption of cellular redox status, leading to oxidative stress. Our observation of an increased steady-state level of ROS, and induction in p16 may contribute to premature-senescence in cells expressing miR-128a (Figure S7).

22In summary, we describe the functional role of microRNA 128a in medulloblastoma. We demonstrate that microRNA 128a decreases medulloblastoma cell growth through mechanisms involving ROS and senescence. We are now investigating the role of microRNA 128a in radio-sensitizing medulloblastoma using in vivo orthotopic xenograft models.

23Daoy and D283 medulloblastoma cells were obtained from American Type Culture Collection (Rockville, Md.) and cultured in DMEM medium (Gibco, Grand Island, N.Y.) supplemented with 10% fetal bovine serum (Gibco) according to the supplier's recommendations. Cell line ONS76 was kindly provided by Dr. James T. Rutka (University of Toronto, Canada) and was cultured in DMEM containing 10% FBS. D341 was obtained from ATCC and cultured in DMEM with 20% FBS and 1% of sodium pyruvate and L-glutamine. Primary cell cultures were derived from biopsy specimens of medulloblastoma patients under a protocol approved by the Institutional Review Board at the University of Iowa Hospitals and Clinics. Written consent from all patients was obtained as mandated by the University of Iowa IRB. Cultures were maintained at low passage numbers (p2–p4) in DMEM medium supplemented with 20% fetal bovine serum. Normal human cerebellum and medulloblastoma patient samples were obtained from the Pediatric Co-operative Human Tissue Network (Columbus, Ohio) under a University of Iowa IRB approved protocol. All samples were obtained in an anonymous fashion as mandated by the IRB. Samples we Normal cerebellar samples were from nonmalignant pediatric and adult brain. All medulloblastoma samples were from pediatric patients. Details of patient samples have been previously described. Briefly all samples were from non-metastatic (M0) tumors with classical medulloblastoma histology.

24Small RNAs were isolated using the mirVANA RNA isolation kit (Ambion).

25MicroRNA expression was measured by quantitative PCR using an ABI 7700 thermal-cycler. MicroRNA primers and probes were purchased from Applied Biosystems (Foster City, CA) and the assays performed per the manufacturer's recommendations with several modifications to enable us to detect low abundance microRNAs. For the reverse transcription (RT) reaction 20 nanograms of total RNA was used. For the qPCR reaction the resultant cDNA was diluted 1∶2. Each RT step was performed in duplicate and the qPCR in triplicate for each RT reaction. U6b and U66 small RNAs were used as endogenous controls and relative microRNA quantity calculated by the ΔΔCT method.

26Cell proliferation was measured by MTT assay and cell viability was determined by trypan blue dye exclusion assay. To assay for cell proliferation, medulloblastoma cells were transfected with control miR, miR-9 or miR-128a oligonucleotide in 24 well plates using LT1 reagent (Mirus, Madison WI). 6, 000 cells/well were then seeded in triplicate in to 96 well plates in 100 ul of medium. After 2 days, 100 ul of Cell Titer AQeous (Promega, Madison, WI) was added and cells incubated for 1 hour and the absorbance measured at 490 nm using a microplate reader per the manufacturers recommendations. Relative cell number was calculated by normalizing the absorbance to untreated cells.

27To assay for cell viability, Daoy cells were seeded in a 6 well plate and 24 h later cells were transfected with control miR or with two different concentrations of miR-128a oligonucleotide. The cell viability was measured every day for upto 5 days after the transfection. Day 0 is the day of transfection.

28The ability of Daoy cells to clonally expand and create colonies with and without either miR-128a or Bmi-1 transfection or co- transfection with both miR-128a and Bmi-1 was measured. Daoy cells were seeded for colony formation 48 h after transfection. After 7 days of incubation at 37°C, colonies were stained with 0.5% crystal violet. Individual colonies (>50 cells) were counted using an Olympus SZ51 (Leeds Instruments) inverted microscope.

29Luciferase experiments were done using the siCHECK dual luciferase system (Promega, Madison, WI). The putative miR-128a target sites from the BMI-1 3′ UTR were cloned into the 3′UTR of the Renilla luciferase gene. Cells were transfected with 0.5 ug of the siCHECK plasmid using LT1(Mirus, Madison, WI) and Renilla luciferase activity was measured 48 hours after transfection and normalized to firefly luciferase. Some cells were co-transfected with 50 or 100 nM miR-128a microRNA oligonucleotide (Ambion, Texas). Western blotting was performed per standard methods. The primary antibodies used were rabbit polyclonal anti-Bmi-1 (Cell Signaling Technology, Danvers, MA); mouse monoclonal anti-P16 (Santa Cruz Biotechnology Inc., Santa Cruz, CA.) and rabbit polyclonal anti-H3K9 (Cell Signaling Technology, Danvers, MA.). The primary antibody dilutions were 1∶1000. Horseradish peroxidase secondary antibodies were detected by chemiluminescence (Supersignal® West Dura extended duration substrate from Thermo Scientific)

30E2F activity was measured using luciferase reporters (Cignal Finder Reporter, SA Biosciences) per the manufacturers instructions. Cells were seeded in 6 well plates containing luciferase reporters to E2F with surefect transfection reagent. Luciferase activity was measured using the Dual Luciferase Assay system (Promega) on a Glomax multi luminometer (Promega). Firefly luciferase was experimental E2F reporter and Renilla luciferase was the normalizing reporter.

31The formation of superoxide in Daoy cells transfected with either miR-128a, Bmi-1, control vector or with both miR-128a and Bmi-1 was followed by electron paramagnetic resonance (EPR). EPR spin trapping technique with DMPO was used to detect the level of superoxide (O2•-). This technique involves an addition reaction of a short-lived radical to a diamagnetic compound (spin trap, here DMPO) to form a more stable free radical product (spin adduct), which can be studied by EPR. The intensity of the spin adduct signal corresponds to the amount of short-lived radicals trapped; the hyperfine couplings of the spin adduct are characteristic of the original trapped radical. In brief, after 48 h of transfection, cell monolayers were washed with PBS and incubated with 100 mM DMPO in chelated- PBS (pH 7.4) for 20 min. The specificity of the superoxide signal was determined by pre-treating the cells with CuZnSOD (1000 U/ml) for 30 minutes. The cells were then transferred to a TM quartz flat cell and EPR spectra were recorded using a Bruker EMX spectrometer equipped with a TM cavity. EPR spectra were obtained as an average of 15 scans with a modulation amplitude of 1 G; scan rate 80 G/81 s; receiver gain 104–106; microwave power, 40 mW; and modulation frequency of 100 kHz. The EPR peak heights are in arbitrary units.

32At day 5 after transfection with miR-128a or control miRNA, Daoy cells were fixed and stained by using an SA-beta-gal kit (Cell Signaling Technology) according to the manufacturer's instructions. The blue senescent cells were distinguished from viable cells under the microscope. Four high power fields per sample were counted in three independent samples to quantify the number of senescent cells.

33P values were calculated using a student's t-test. Error bars are depicted as mean ± standard error.