Downregulation of 14q32 microRNAs in Primary Human Desmoplastic Medulloblastoma

Medulloblastoma (MB) is one of the most common pediatric cancers, likely originating from abnormal development of cerebellar progenitor neurons. MicroRNA (miRNA) has been shown to play an important role in the development of the central nervous system. Microarray analysis was used to investigate miRNA expression in desmoplastic MB from patients diagnosed at a young age (1 or 2 years old). Normal fetal or newborn cerebellum was used as control. A total of 84 differentially expressed miRNAs (64 downregulated and 20 upregulated) were found. Most downregulated miRNAs (32/64) were found to belong to the cluster of miRNAs at the 14q32 locus, suggesting that this miRNA locus is regulated as a module in MB. Possible mechanisms of 14q32 miRNAs downregulation were investigated by the analysis of publicly available gene expression data sets. First, expression of estrogen-related receptor-γ (ESRRG), a reported positive transcriptional regulator of some 14q32 miRNAs, was found downregulated in desmoplastic MB. Second, expression of the parentally imprinted gene MEG3 was lower in MB in comparison to normal cerebellum, suggesting a possible epigenetic silencing of the 14q32 locus. miR-129-5p (11p11.2/7q32.1), miR-206 (6p12.2), and miR-323-3p (14q32.2), were chosen for functional studies in DAOY cells. Overexpression of miR-129-5p using mimics decreased DAOY proliferation. No effect was found with miR-206 or miR-323 mimics.


INTRODUCTION
Medulloblastoma (MB) is an embryonic tumor of the cerebellum and the most common malignant brain tumor in childhood, likely originating from abnormal development of cerebellar progenitor neurons (1,2). Transcriptional profiling of large number of MB samples unraveled the existence of at least four distinct molecular subgroups: (i) Wingless (Wnt) group, (ii) Sonic Hedgehog (SHH) group, (iii) Group 3 and (iv) Group 4 (3,4). These various subtypes of MB are suggested to arise from different populations of precursor or stem cells which form the cerebellum (5)(6)(7)(8). This transcriptome-based classification has opened new avenues for the understanding of the molecular mechanism contributing to MB.
MicroRNAs (miRNAs) are suggested to play an important role in controlling the development of the central nervous system (CNS) by regulating cell proliferation and differentiation, as well as apoptosis (9). miRNAs are small non-coding RNA molecules of ∼22-25 nucleotides that post-transcriptionally downregulate gene expression by binding the 3 -untranslated region (UTR) of protein coding transcripts, resulting in either mRNA cleavage or translational repression (10, 11). miRNA expression profiling of both mouse and human MB has led to the identification of signatures associated with the molecular subgroups of MB, tumor diagnosis, and response to treatment, as well as novel targets of potential clinical relevance (12-15). Previous studies, however, interrogated limited number of miRNAs and included adult cerebellum in the normal control group. We investigated the expression profile of 847 miRNA in primary human desmoplastic MB of younger children in comparison to normal fetuses or newborn cerebellum. Eighty-four miRNAs were found to be differential expressed in MB, most of them belonging to the cluster 14q32. Possible mechanisms of 14q32 locus downregulation were investigated by the analysis of publicly available gene expression data set. Functional studies using mimic miR-129-5p (11p11.2/7q32.1), miR-206 (6p12.2), and miR-323-3p (14q32.2) and the DAOY cell line, suggested a suppressive role for miR-129-5p in MB proliferation.

PRIMARY MEDULLOBLASTOMA TISSUE SAMPLES
Surgical specimens were obtained from 1 to 5 years old children (n = 10), with desmoplastic MB ( Table 1). Of note, microarray analyses were performed with MB samples from children with 1-2 years old. Desmoplastic MBs belong, with rare exceptions, to the SHH molecular subgroup (13-16). All MB samples used in the present study had high mRNA levels of PTCH1 and low levels of OTX2 (Figure A1 in Appendix), in comparison to normal cerebellum, which is in keeping with the differential transcriptional profile of SHH tumors (3). Normal cerebellum tissues were

APOPTOSIS ASSAY
DAOY cells transfected with miR-206, 129-5p, 323-3p, or scramble mimics were cultured for 24 h in serum-free RPMI-1640 (Cultilab), harvested and part of it was resuspended in the appropriate binding buffer, stained with FITC-conjugated Annexin V (BD Biosciences, San Jose, CA, USA) and propidium iodide at room temperature for 15 min, and subsequently analyzed by flow cytometry in a FACS Canto II (Becton Dickinson). The remaining cells were replated in six-well plates for another 24 h culture period in serumfree RPMI-1640 (Cultilab) and harvested 48 h post-transfection for Annexin V labeling.

STATISTICAL ANALYSIS AND BIOINFORMATICS METHODS TO SIGNALING PATHWAY PREDICTION
MicroRNA expression was analyzed in R environment 1 using the packages Affy and RankProd from Bioconductor (19-21). The MB miRNA profile was compared to the cerebellum profile. Differentially expressed miRNAs were selected according to the fold change ≥2.00 and p-value ≤0.05. Heat maps were created using tools of the MetaboAnalyst 2.0 2 . Signaling pathways were prospected by DIANA-miRPath (microT-v4.0, beta version) 3 . The input dataset enrichment analysis was performed by Pearson's chi-squared test and each pathway was represented by the negative natural logarithm of the P-value (−ln P). The Ingenuity Pathway Analysis (IPA) software 4 was used to identify possible pathways associated to differentially expressed miRNAs.
Comparisons of RT-qPCR values from MB versus normal cerebellum were performed by the Mann-Whitney test. Cell proliferation results, at each time point, from mimic miRNA transfections versus mimic-negative control #1 were analyzed by the two-tailed unpaired t -test. Alpha error of P = 0.05 was tolerated. The GraphPad Prism 5 software was used throughout.

IDENTIFICATION OF DIFFERENTIALLY EXPRESSED miRNAs IN DESMOPLASTIC MBs OF 1-2 YEARS OLD CHILDREN
Global miRNA profiles were generated for primary MB of the desmoplastic subtype and most likely SHH molecular subgroup (n = 6), and normal fetal/NW cerebellum (n = 6). Eighty-four miRNAs (64 miRNAs downregulated and 20 miRNAs upregulated) were considered to be differentially expressed (fold change ≥2.0, p ≤ 0.05) in MB in comparison to normal fetal/NW cerebellum (Table 3; Figure 1). Among these 84 miRNAs, 46 had been previously described as deregulated in human primary MB ( Table 3), and only 8 were previously validated by functional assays ( Table 4). Upregulation of miRNAs from the miR-17 ∼ 92 cluster (in this work miR-18a, 19a, and 92a-1) and downregulation of miR-324-5p were previously described in human MB of the SHH subgroup (12, 13). Of especial note, 32 of the 64 downregulated miRNAs belong to a large cluster on human chromosome 14q32 (Figure 1; Table 3).

SIGNALING PATHWAYS ANALYSIS BY DIANA
Signaling pathways putatively altered by MB deregulated miRNA were depicted by DIANA-miRPath. The list of the top 20 pathways is shown in Table 5. The Ribosome pathway was only pointed by the list of downregulated miRNAs. Adherens junction, oxidative phosphorylation, and TGF-beta signaling pathways showed higher enrichment when the list of downregulated miRNAs was used in the analysis. On the other hand, the MAPK pathway and genes associated to cancer showed higher enrichment when upregulated miRNAs were used in the analysis.
Interestingly, oxidative phosphorylation, TGF-beta signaling pathway, and ubiquitin mediated proteolysis were enriched in the list of 14q32 miRNAs.

INGENUITY PATHWAY ANALYSIS
Network analysis by IPA identified two networks as putative targets for 73 out of the 84 MB miRNAs. Networks were prospected considering only relationships that were experimentally observed. Interestingly, both networks were enriched with miRNAs belonging to the 14q32 cluster.

RT-qPCR VALIDATION OF SOME DEREGULATED miRNAs
miR-323-3p and 495, both belonging to cluster 14q32, were chosen for validation by RT-qPCR. In addition, miR-206 and miR-129-5p were chosen for analysis because of their high fold change (see Table 3), lack of previous functional studies and possible oncogenic role. miR-206 expression was reported to inhibit cell proliferation in breast cancer cells (22). miR-129 is reported to be significantly downregulated in pediatric brain tumors compared to normal tissues (23). Most importantly, miR-129 downregulation is associated to SOX4 overexpression in endometrial and gastric cancers (24, 25). SOX4 is upregulated and has prognostic impact in MB (26, 27).

UPREGULATION OF miR-206, 129-5p, AND 323-3p IN DAOY CELLS
As a first approach to investigate the functional significance of miRNAs downregulation in MB, DAOY cells were transiently transfected with mimics of miR-206, 129-5p, 323-3p, or negative control #1. Transfection efficiency was confirmed by RT-qPCR (Figure A2 in Appendix). Twenty hours post-transfection, cells were collected and seeded in 96-well plates in serum-free medium. DAOY cells survive and even proliferate in serum-free medium for a short period of time. As shown in Figure 4, no consistent differences in proliferation were found in transfections with miR-206 and miR-323-3p. On the contrary, transfections with miR-129-5p www.frontiersin.org resulted in a significant decrease in DAOY cell proliferation, as evaluated by the MTS assay (Figure 4). Similar experiments were conducted to evaluate cell survival and apoptosis by the Annexin V and propidium iodide staining methodology. No significant differences were found on cell viability or apoptosis after miR-206, miR-129-5p, and miR-323-3p transfections in comparison to control (Figure 4; Figure A3 in Appendix, respectively).

DISCUSSION
We investigated the expression profile of miRNAs in primary human MB of desmoplastic histology and SHH molecular subgroup, in comparison to normal fetal/newborn cerebellum. Eighty-four miRNAs were found to be differentially expressed in MB. The majority of these differentially expressed miRNAs were downregulated in comparison to normal cerebellum, corroborating previous studies. Most upregulated miR-NAs identified in our study (12 out of 20) had been previously described in MB (12-14, 23, 28-32). On the contrary, 31 out of the 64 downregulated miRNAs are here described for the first time in association to MB ( Table 3). Differences may be explained by the fact that a more comprehensive version of Affymetrix miRNA microarray was used in the present study. Moreover, previous studies included different subtypes of MB and a mix of children and adults cerebellum samples in their analysis (12, 13, 29). We believe that analysis on more uniform groups of both cancer and control samples may have helped us in detecting some smaller but consistent differences between groups. A computational analysis was performed to predict the network and signaling pathways collectively targeted by the 64 downregulated and 20 upregulated miRNAs. Downregulated miRNAs in MB were predicted to target genes related to the ribosome, adherens junction, oxidative phosphorylation, metabolism of xenobiotics by cytochrome P450, and transforming growth factor-beta (TGF-β) signaling pathways. Axon guidance, TGF-β,WNT, insulin signaling pathways are known to play an important role in neurulation, CNS developmental, and/or MB pathogenesis (3,33,34). Since miRNA act as negative regulators of gene expression, a simpler interpretation of these findings is that MB has increased activation of these pathways in comparison to normal cerebellum.

Frontiers in Oncology | Pediatric Oncology
Most importantly, half (32/64) of downregulated miRNAs reported in our study were found to belong to the cluster at 14q32 locus (also known as miR-154 cluster). This is in keeping with a previous study in a mouse model of MB, reporting that activation of SHH signaling leads to downregulation of the miR-154 cluster (35). Moreover, previous publications with primary MB found downregulation of some 14q32 miRNAs in MBs of the molecular subgroups WNT, SHH, and C as compared to normal cerebellum and MBs of subgroup D (13, 29). However, this is the first time that so many 14q32 miRNAs are shown to be downregulated in MB, thus suggesting a co-regulatory control of this cluster's expression.
Deletions at locus 14q32 would be one possible explanation to the decreased 14q32 miRNAs expression. The recent analysis of somatic copy number aberrations in 1,087 MB samples report significant losses of chromosome arm 14q in the SHH subgroup of MBs, though not restricted to 14q32 (36). Alternative explanations are discussed below. Our Ingenuity pathways analysis pointed to nuclear orphan receptor NR0B2 (also known as Small Heterodimer Partner, SHP) and (ESRRG) as possible controllers of 14q32 miRNA cluster expression. There is indeed experimental evidence in mouse showing that NR0B2 is a repressor while ESRRG is and activator of a miRNA cluster in chromosome 12, which is ortholog to the 14q32 cluster in humans (37). Our analysis of microarray mRNA expression data for 64 primary human MB samples, accessible through GEO Series accession number GSE28245 (38) in NCBI's Gene Expression Omnibus (39) revealed that NR0B2 is not expressed in MB. Interestingly, ESRRG expression was found to be relatively high in MBs of the molecular subgroup D, intermediate in MBs of the WNT and C subgroups, and very low or absent in MBs of the SHH subgroup ( Figure A4A in Appendix), thus reflecting 14q32 miRNAs abundance in each of the MB subgroup. These findings were confirmed by the analysis of gene expression data of an independent cohort of 90 primary MB samples (accession number GSE21166) deposited by Northcott et al. (13) (data not shown). ESRRG suppress cell proliferation in prostate cancer cells (40) and the estrogen receptor beta agonist diarylpropionitrile (DPN ) exhibit a pro-apoptotic and anti-proliferative effect on MB (41). Experiments are warranted to investigate a possible causal connection between ESRRG and 14q32 miRNA cluster expression in MB.
The miRNA cluster at 14q32 lies within a parentally imprinted chromosomal area spanning genes Dlk1, Meg3, Rtl1, Meg8, and Dio3 (42). Dlk1, Rtl1, and Dio3 are paternally-, whereas Meg3 and Meg8 are maternally expressed transcripts (43). Imprinting of 14q32 is regulated, to some extent, by two intergenic differentially methylated regions known as IG-DMR and MEG3-DMR (44,45). www.frontiersin.org Deletions of the regulatory regions and/or epigenetic modifications may in theory cause aberrant 14q32 silencing in cancer. The recent 1,000 genome study of somatic copy number aberrations shows no recurrent focal deletions at locus 14q32 in MB (36).
However, our analysis of public mRNA microarray expression data GSE28245 (38)    www.frontiersin.org levels while group D have MEG3 levels closer to normal cerebellum (Figure A4B in Appendix). Thus MEG3 expression seems to correlate with the expression of 14q32 miRNAs among the different MB molecular groups, suggesting that the 14q32 miRNA locus may be under epigenetic regulation in MB. However, a genome wide analysis of promoter methylation on four primary MB samples showed no consistent methylation of 14q32 gene promoters (46). Although higher number of samples should be analyzed, this result corroborates findings in osteosarcoma, a tumor also presenting with downregulated 14q32 miRNAs expression and with no consistent changes in the methylation patterns at 14q32. Instead, silencing of 14q32 miRNA in osteosarcoma seems to be mediated by histone modification(s) (47).
Mimics for miR-206 and 323-3p had no significant effect on DAOY cells. miR-129-5p overexpression resulted in decreased cell proliferation, which may suggest a tumor suppressor role in MB.