The long noncoding RNA linc-NeD125 controls the expression of medulloblastoma driver genes by microRNA sponge activity

Long noncoding RNAs (lncRNAs) are major regulators of physiological and disease-related gene expression, particularly in the central nervous system. Dysregulated lncRNA expression has been documented in several human cancers, and their tissue-specificity makes them attractive candidates as diagnostic/prognostic biomarkers and/or therapeutic agents. Here we show that linc-NeD125, which we previously characterized as a neuronal-induced lncRNA, is significantly overexpressed in Group 4 medulloblastomas (G4 MBs), the largest and least well characterized molecular MB subgroup. Mechanistically, linc-NeD125 is able to recruit the miRNA-induced silencing complex (miRISC) and to directly bind the microRNAs miR-19a-3p, miR-19b-3p and miR-106a-5p. Functionally, linc-NeD125 acts as a competing endogenous RNA (ceRNA) that, sequestering the three miRNAs, leads to de-repression of their targets CDK6, MYCN, SNCAIP, and KDM6A, which are major driver genes of G4 MB. Accordingly, linc-NeD125 downregulation reduces G4 cell proliferation. Moreover, we also provide evidence that linc-NeD125 ectopic expression in the aggressive Group 3 MB cells attenuates their proliferation, migration and invasion. This study unveils the first lncRNA-based ceRNA network in central nervous system tumours and provides a novel molecular circuit underlying the enigmatic Group 4 medulloblastoma.


INTRODUCTION
Medulloblastoma is the most common malignant paediatric brain tumour [1]. Recent transcriptomics and genomics analyses of large human primary tumour cohorts assigned MBs to four molecularly distinct subgroups, each characterized by specific developmental origins, molecular features, and prognoses [1][2][3][4]. The best characterized WNT and SHH subgroups have been causally linked to altered Wingless and Sonic Hedgehog developmental cascades, respectively [1]. Larger gaps remain in our understanding of the signalling pathways underlying Group 3 (G3) and Group 4 (G4) MBs, which account for 60% of all prognoses and present the greatest clinical challenges [4]. G3, the subgroup with the worst outcomes and the highest metastasis rates (75%), is characterized by a MYC-activation signature [1]. G4 tumours have a better prognosis, and are the most common MBs characterized by a neuronal signature with over-representation of genes involved in neuronal differentiation and development [1][2][3][4]. Genomic analyses highlighted subgroup-enriched dysregulated genes, the so-called driver genes, altered by single nucleotide variants (SNVs) or somatic copy number aberrations (SCNAs) [1,2]. However, driver gene expression might also be altered by defects in transcriptional [5] or post-transcriptional regulatory mechanisms, as those involving microRNAs (miRNAs) [6].
Increasing emphasis has been recently placed on the potential roles of long noncoding RNAs (lncRNAs) as gene expression regulators in human nervous system physiology [7]. Aberrant lncRNA expression has been documented in neurodevelopmental, neurodegenerative, and neuro-oncological disorders [8]. In the latter setting, lncRNAs are emerging as critical players: their differential expression in gliomas and neuroblastomas has been actively investigated [9,10], but very little is known about their roles in MB.
We recently identified a novel human long intergenic noncoding RNA (lincRNA), as the host gene for miR-125b-1. It is specifically induced during in vitro differentiation of neuronal tumour cell lines-hence its name: Neuronal Differentiation lncRNA hosting miR-125 (linc-NeD125) [11].
In this study, we explored the roles it plays in brain cancer and discover that linc-NeD125 is an essential node in a novel regulatory network in G4 MB, the most prevalent and pathogenetically enigmatic class of MBs. We demonstrate that, when expressed at the high levels found in G4 MBs, linc-NeD125 functions as a competing endogenous RNA (ceRNA) that, sequestering miR-19a-3p, miR-19b-3p, and mir-106a-5p, de-represses the expression of their targets CDK6, MYCN, SNCAIP and KDM6A, major driver genes of G4 MB. Remarkably, we revealed a role for linc-NeD125 in reducing G4 MB cell proliferation and G3 MB cell aggressiveness in vitro.

Linc-NeD125 is targeted by specific miRNAs
In human BE(2)-C neuroblastoma cells, linc-NeD125 is localized in the cytoplasm [11]. This finding, together with preliminary bioinformatics analyses revealing multiple miRNA response elements (MREs) throughout its length (Supplementary Table 1), suggested that linc-NeD125 might act as a ceRNA. To explore this hypothesis, we performed RNA pull-down assays in BE(2)-C cells treated with retinoic acid (RA), which triggers linc-NeD125 expression approximately 6-fold compared to untreated cells [11]. These assays showed that linc-NeD125 was associated with Argonaute 2 (AGO2), a major component of the miRNA-induced silencing complex (miRISC) where miRNAs interact with their mRNA targets ( Figure 1A). Linc-NeD125-AGO2 interaction was confirmed by crosslinking immunoprecipitation (CLIP) experiments ( Figure 1B). The two approaches demonstrate that linc-NeD125 is able to recruit the miRISC, a pre-requisite to function as a ceRNA.
To identify the miRNAs possibly associated with linc-NeD125 in the miRISC, high-throughput qRT-PCR analysis was performed on complexes precipitated from two distinct linc-NeD125 pull-down assays. 15 miRNAs were found in both experiments ( Figure 1C), 6 of which were predicted to target linc-NeD125 according to the miRanda algorithm ( Figure 1D, left panel, and Supplementary Table 1). The same tool was used to eliminate 2 of the 6 miRNAs that could bind the pull-down bait, leaving a short list of 4 miRNAs-namely miR-19a-3p, miR-19b-3p, miR-106a-5p and miR-191-5p-which are specifically bound by linc-NeD125 ( Figure 1D, right panel).

Linc-NeD125 is expressed in MBs and upregulated in G4 subgroup
The experiments in tumour-derived neuronal cells provided evidence that linc-NeD125 is a potential ceRNA. Given the increasing evidence for the involvement of lncRNAs as ceRNAs in neuronal cancer-associated networks [12], we asked whether linc-NeD125 may play this role in MBs. Taking advantage of a large number of available human specimens, we evaluated linc-NeD125 expression in a cohort of 51 primary tumours (Supplementary Table 2), representing all four MB subgroups in proportions reflecting their incidence in the population [1]. As shown in Figure 2A, linc-NeD125 was expressed in all subgroups and significantly upregulated (20-fold increase on average) in G4 MB, compared to normal cerebellum. Levels found in G4 tumors were approximately twice as high as those in WNT MBs and roughly 20 times those of the SHH and G3 tumours. www.impactjournals.com/oncotarget Oncotarget 31006 www.impactjournals.com/oncotarget miR-19a-3p, miR-19b-3p and miR-106a-5p repress G4 MB driver gene expression To determine linc-NeD125 contribution to G4 MB, we checked whether genes relevant for this tumour subgroup were specific targets for the bound miRNAs. We initially assessed miRNA occurrence in G4 primary tumors. As shown in Figure 2B, only miR-19a-3p, miR-19b-3p, and miR-106a-5p were significantly overexpressed in tumour specimens, with 2-to 4-fold increases over control levels.
To validate this prediction we used, as an in vitro model system, the D283 Med cells. They are G3-derived MB cells [13] that show low levels of linc-NeD125 ( Figure 2D) allowing us to upregulate its expression as observed in G4 tumours.
Given the high expression levels of miR-19a-3p, miR-19b-3p and miR-106a-5p in D283 Med cells ( Figure 2E), miRNA loss-of-function experiments were performed. Co-transfection of Locked Nucleic Acids (LNAs) against the three miRNAs did not alter linc-NeD125b levels ( Figure 3A, left panel), while caused an increase of CDK6, MYCN, SNCAIP and KDM6A protein levels (ranging from ~1.4 to ~2-fold over LNAcontrol treated cells) ( Figure 3A, middle and right panels), demonstrating that the corresponding genes are targets of the three miRNAs bound by linc-NeD125.
Further demonstration that the three miRNAs together are able to specifically interact with CDK6, MYCN, SNCAIP and KDM6A mRNAs derived from luciferase reporter assays. As schematized in Figure 3B (left panel), portions of their 3′UTRs were cloned downstream of the r-luc ORF and expressed in D283 Med cells together with scramble or miRNA targeting LNAs. Histograms in Figure 3B (right panels) show an increase of luciferase activity indicating the specificity of the three miRNA interaction.

Linc-Ned125 de-represses G4 MB driver gene expression by sequestering miR-19a-3p, miR-19b-3p, and miR-106a-5p
To explore linc-NeD125 involvement in the control of G4 MB driver genes and the possibility that it is mediated by its miRNA sponge activity, we designed constructs expressing mature wild type linc-NeD125 (Linc-125) or a mutant derivative defective in miRNA binding activity (mLinc-125) (Supplementary Figure 1). The latter was planned by introducing single point mutations in MRE sites corresponding to miRNA seed positions 4 and 5 [14]. Importantly, we ensured that none of the mutations created new miRNA-binding sites.
To verify the specificity of miRNA-linc-NeD125 interaction, we cloned the wild type and mutant linc-NeD125 into luciferase reporter vectors ( Figure 4A, left panel) and transfected them into D283 Med cells, along with the LNAs complementary to miR-19a-3p, miR-19b-3p, and miR-106a-5p. As shown in Figure 4A (right panel), cells transfected with the LNAs and wild type linc-NeD125 (Luc/Linc-125) exhibited increased luciferase activity (1.5-fold over controls), but no change was observed in cells transfected with the mutant transcript (Luc/mLinc-125). These results confirm the specificity of the miRNA/linc-NeD125 interaction and the inability of the mutant transcript to bind the three miRNAs.
For functional analyses, the wild type and mutant linc-NeD125 sequences were cloned into an expression vector ( Figure 4B, left panel) and transfected in D283 Med cells. As shown in Figure 4B (right panel) the two transcripts were overexpressed at levels comparable to those characterizing primary G4 MBs. Their ectopic expression had no effect on the levels of miR-19a-3p, miR-19b-3p, or miR-106a-5p (Supplementary Figure 2), indicating that linc-NeD125 does not regulate their abundance [15]. G4 driver gene mRNA levels were also unaltered ( Figure 4C, left panel), suggesting that their regulation by the three miRNAs occurs mainly at the translational level (See Figure 3A, middle panel). In contrast, driver gene protein products were significantly increased by ectopic expression of wild type linc-NeD125 (1.5-to 2-fold increases), as compared with untreated controls ( Figure 4C, middle and right panels). No any effect was produced by overexpression of the mutant linc-NeD125. Complementary loss-of-function experiments were carried out in a recently reported G4 in vitro model system, the CHLA-01-MED cell line [16], where linc-NeD125 is expressed at higher levels compared to D283 Med cells ( Figure 4D, left panel). We found that knockdown of the endogenous linc-NeD125 in CHLA-01-MED cells ( Figure 4D, right panel) caused a significant decrease of the four driver gene protein products ( Figure 4E), and that this is accompanied by a significant reduction of cell proliferation ( Figure 5A, upper panel) and of the proliferation marker KI-67 ( Figure 5A, lower panel).

Linc-NeD125 impairs in vitro cell proliferation, migration and invasion of G3 MB cells
We asked whether the observed linc-NeD125mediated derepression of specific G4 MB driver genes Oncotarget 31007 www.impactjournals.com/oncotarget in a G3 MB genetic background led to any phenotypic consequence.
We assessed G3-derived D283 Med cells' capacities for proliferation, migration, and invasion after overexpression of wild type or mutant linc-NeD125 ( Figure 5B), at the high levels observed in primary G4 tumors. Unexpectedly, the results of BrdU labelling assay, which measures DNA replication rates ( Figure 5C), and the analysis of the proliferation marker KI-67 ( Figure 5D) showed that cell proliferation was significantly reduced by wild type linc-NeD125 (a decrease of approximately 30%) but was unaffected by the mutant. Migration ( Figure 5E) and invasion ( Figure 5F) were also reduced (by approximately 40%) in linc-NeD125 overexpressing cells, but no effects were seen in cells expressing the mutant transcript. To confirm that linc-NeD125 biological function depends on its ability to inhibit miRNA activity, we analyzed D283 Med cell proliferation ( Figure 6A and 6B), migration ( Figure 6C) and invasion ( Figure 6D) upon LNA-mediated miRNA sequestration. A decrease Furthermore, to demonstrate that linc-NeD125 action is directed towards the G4 driver genes CDK6, MYCN, SNCAIP, and KDM6A, linc-NeD125 was ectopically expressed in D283 Med cells while repressing their expression. Figure 7 shows that siRNA-mediated silencing of the driver genes prevents linc-NeD125 from affecting cell proliferation ( Figure 7A), migration ( Figure 7B) and invasion ( Figure 7C).
These results demonstrate that linc-NeD125 ectopic expression in D283 Med cells can effectively attenuate the G3 MB cell capacities for proliferation, migration, and invasiveness, and that these effects are mediated by linc-NeD125 sponge activity.

DISCUSSION
Roughly 60% of MBs belong to G3 and G4. Both are associated with relatively unfavourable outcomes  Oncotarget 31010 www.impactjournals.com/oncotarget (particularly G3), and targeted therapies are at their infancy. The pathogenesis of G3 and G4 is still poorly defined: their development cannot be adequately explained by SCNAs and SNVs alone, and it could depend on additional mechanisms operating widely at the epigenetic and/or post-transcriptional levels.
Our findings delineate a novel ceRNA network headed by linc-NeD125 (Figure 8), which significantly contributes to the dysregulation of four critical G4 MB driver genes. Linc-NeD125 functions as a natural miRNA sponge, competitively binding and sequestering three endogenous miRNAs-miR-19a-3p, miR-19b-3p and miR-106a-5p-whose targets include CDK6, MYCN, SNCAIP, and KDM6A transcripts (Figure 8). This is the first report of a lncRNA-based ceRNA network associated with neuronal cancer [12], and a new example of a single RNA sponge simultaneously controlling a large number of co-regulated targets in cancer [17,18].
CeRNA activity has been described in different experimental systems and attributed to several RNA species, including lncRNAs, pseudogenes, circular RNAs, and mRNAs [19]. Different models predict the requirement of near-equimolarity among ceRNA network components for regulation to occur [20]. However, natural regulatory networks are likely to be more complex owing to physiological and pathological variations (transient enrichment, local concentration) in RNA levels. Our data indicate that linc-NeD125 behaves as a miRNA sponge in G4 MBs, where its expression is approximately 20-fold higher than that found in normal cerebellar tissues. This Oncotarget 31011 www.impactjournals.com/oncotarget conclusion derives from gain-of-function experiments in D283 Med cells, which exhibit the molecular features of primary G3 MB [13], including linc-NeD125 expression levels lower than those found in G4 tumors. This makes them a highly suitable system for mechanistic studies of linc-NeD125 function. Remarkably, linc-NeD125 overexpression in D283 Med cells at levels comparable to G4 tumors elicited the de-repression of multiple G4 MB driver genes, targets of the sponged miRNAs. Linc-NeD125 sponge activity is confirmed by the evidence that overexpression of its mutant derivative, carrying microRNA target sites disrupted at specific positions in Oncotarget 31012 www.impactjournals.com/oncotarget the seed, fails to alter G4 MB driver gene expression. Unlike commonly used deletion mutants, this designed site-specific mutant created no novel miRNA binding sites and minimized changes in linc-NeD125 sequence and structure, two features crucial for noncoding RNA activity. Notably, the complementary linc-NeD125 lossof-function experiment carried out in the G4 MB cell line CHLA-01-MED confirmed its role as a ceRNA regulating G4 driver genes. This, together with the decrease of G4 cell proliferation upon linc-NeD125 knockdown, indicates it may function as an oncogene in G4 MB tumours. In vivo studies are required to address this point.
The phenotypic upshot of linc-NeD125 overexpression in a G3 cell model is the acquisition of specific G4 molecular features, namely an increase of the G4 driver gene protein products. Unexpectedly, this was accompanied by a significant attenuation of the high capacity for proliferation, migration and invasion that characterises the high metastatic G3 cells, making them G4-like cells for these capacities. Although the underlying mechanism, dependent on linc-NeD125 ceRNA function, is still unknown, we speculate that it might be related to the neuronal signature that specifically marks G4 MB [1] and that is possibly responsible for the less aggressiveness of this tumour subgroup. In fact, while group 3 has the worst prognosis with less than 50% 5-year survival, group 4 has an about 75% 5-year overall survival [1]. In light of its ability to attenuate the in vitro aggressiveness of the metastatic G3 MB cell line and the proliferation of G4 MB cells, linc-NeD125 mode of action merits deeper study for possible therapeutic applications.
In conclusion, post-transcriptional miRNA-mediated crosstalk between linc-NeD125 and protein-coding RNAs appears to play a significant role in G4 MB tumorigenesis. Linc-NeD125 may be a novel driver gene operating upstream of the already identified G4 protein-coding driver genes, and/or it might keep high the levels of driver gene products participating in the maintenance of cancer cell identity.
The results of our study highlight linc-NeD125 as a key player in G4 MB driver gene network and as a novel member of the small but growing list of lncRNAs implicated in human cancerogenesis via the ceRNA mechanism.

MB molecular subgrouping
Molecular characterization of the 51 tumours was performed as previously described [21]. Details of molecular characterization are shown in Supplementary Table 4

Plasmids
pcDNA3.1-Linc-NeD125: the wild-type version was PCR-amplified from cDNA generated from 4 day RAtreated BE(2)-C cells with the oligolucleotides UpBam and DownNot1, and inserted between the BamHI and NotI restriction sites of the pcDNA3.1+ vector. The mutant version was derived by the QuikChange II Site-Directed Mutagenesis Kit (Agilent), using the oligonucleotides Mut1-Mut4. psiCheck2-Linc-Ned125: wild type version was PCR-amplified from pcDNA3.1-Linc-NeD125 with the oligonucleotides UpXho and DownNot1. The mutant was derived from the wild type, as described above. pEZX-MYCN and pEZX-SNCAIP 3'UTR luciferase reporter constructs were purchased from GeneCopoeia. psiCheck2-KDM6A and psiCheck2-CDK6: KDM6A and CDK6 3′UTRs were PCR amplified from genomic DNA with the oligonucleotides UpXho2 and DownNot2 or UpXho3 and DownNot3, respectively. Oligonucleotides are listed in Supplementary Table 5.