Long non-coding RNA profiling of pediatric Medulloblastoma

Background Medulloblastoma (MB) is one of the most common malignant cancers in children. MB is primarily classified into four subgroups based on molecular and clinical characteristics as (1) WNT (2) Sonic-hedgehog (SHH) (3) Group 3 (4) Group 4. Molecular characteristics used for MB classification are based on genomic and mRNAs profiles. MB subgroups share genomic and mRNA profiles and require multiple molecular markers for differentiation from each other. Long non-coding RNAs (lncRNAs) are more than 200 nucleotide long RNAs and primarily involve in gene regulation at epigenetic and post-transcriptional levels. LncRNAs have been recognized as diagnostic and prognostic markers in several cancers. However, the lncRNA expression profile of MB is unknown. Methods We used the publicly available gene expression datasets for the profiling of lncRNA expression across MB subgroups. Functional analysis of differentially expressed lncRNAs was accomplished by Ingenuity pathway analysis (IPA). Results In the current study, we have identified and validated the lncRNA expression profile across pediatric MB subgroups and associated molecular pathways. We have also identified the prognostic significance of lncRNAs and unique lncRNAs associated with each MB subgroup. Conclusions Identified lncRNAs can be used as single biomarkers for molecular identification of MB subgroups that warrant further investigation and functional validation.

Long non-coding RNAs (LncRNAs) are RNAs of more than 200 bp in length and can be transcribed from an intergenic region, genic regions or super enhancer regions in the genome. LncRNAs can modulate chromatin structure, gene regulation via interactions with epigenetic modifiers and transcriptional co-factors, and also have post-translation effects via affecting the stability of mRNA or proteins [11,12]. Deregulated lncRNA expression is associated with many cancers [13]. LncRNA signatures have been used to classify different types of cancer as biomarkers for diagnosis, prognosis and therapy [14][15][16][17][18]. LncRNAs are secreted in serum, plasma, and CSF in a stable form protected from endogenous RNAase and can be used for noninvasive analysis from patient samples [19,20]. The role of lncRNA in brain development is well studied [21][22][23][24][25][26]. However, there is not much known about role of lncRNAs in MB. LncRNA LOXL1-AS1 promotes the proliferation and metastasis of MB by activating the PI3K-AKT pathway [27]. LncRNA CCAT1 promotes cell proliferation and metastasis in human  MB by regulating the MAPK pathway [28]. Silencing of ANRIL in MB cell lines significantly lowered cell viability and migration. ANRIL promoted the apoptosis of MB cell lines through miR-323-mediated regulation of BRI3, which activates p38 MAPK, ERK, and AKT as well as the WNT signaling pathway [29]. LINC-NeD125 expression is upregulated in Group 4 MB and after interacting to miRNA-induced silencing complex(MISC), it directly binds to miR-19a-3p, miR-19b-3p and miR-106a-5p. Functionally, LINC-NeD125 acts by sequestering the three miRNAs, which leads to the de-repression of major driver genes (CDK6, MYCN, SNCAIP, and KDM6A) of Group 4 MB [30]. LncRNA CRNDE expression is elevated in MB and knockdown of CRNDE significantly reduced cell proliferation and inhibited colony formation in MB cell lines, Daoy and D341 [31].
In the current study, we have identified the lncRNAs expression profile of pediatric MB subgroups and associated molecular pathways. We have also identified the unique lncRNAs associated with each subgroup.

Methods
We searched the Gene Expression Omnibus (GEO) database for MB related microarray datasets and found two relevant studies, GSE37418 [for pediatric MB subgroups expression data] and GSM1094863, GSM1094864, GSM1094865, GSM1094866, GSM1094867 [for pediatric primary cerebellum expression data from GSE44971] for our analyses. We further used large GSE124814 datasets for the validation of lncRNAs expression profiles of MB subgroups obtained from our original analyses. We selected the age < 18 years as an inclusion criteria for selecting pediatric MB samples. We selected the datasets which  used the Affymetrix U133 Plus2 array for probe level RNA expression studies. For data analyses, we first did background correction, normalization (RMA), quality control checks, intensity and batch effect corrections of each dataset. Following that, we did probe level differential analyses of datasets using the limma package (ANOVA with eBayes) with criteria of p < 0.001 and fold change greater than two folds. We then annotated the probe sets with the Affymetrix U133 Plus2 library and filtered out lncRNA genes. The lncRNA gene database used is verified and approved by HGNC. Functional analysis of differentially expressed lncRNAs was done by Ingenuity pathway analysis (IPA) software from BioRad, Inc. We used default parameters and checked all the node types, all species (except uncharted), and all tissue types for core analysis in IPA.

Results
Differentially expressed lncRNAs in the WNT subgroup and their functional roles We found all the top 10 upregulated and downregulated lncRNAs present in validation datasets. We mostly see non-overlap in lncRNAs at lower expression values. We did functional analysis of differentially expressed (DE) lncRNAs of the WNT subgroup using IPA. We identified different functional parameters involved in this subgroup. MAX (a MYC interacting partner), miR-150, miR-133a, FOLR1, E2F NCAM1, GAS2L3 and ATF5 are the most significantly associated upstream regulators, while cancer, neurogenesis, metastasis and cellular development are the most important biological functions          fold change > 2 provided 150 differentially expressed lncRNAs with approved status. Tables 13 and 14

Prognostic significance of lncRNAs in different subgroups of MB
We used a publicly available dataset GSE85217 (Cavalli dataset) to understand the prognostic significance of DE lncRNAs of different MB subgroups. As shown in Fig. 5, high expression of HAND2-AS1 is associated with poor prognosis in WNT MB. Similarly, low expression of MEG3 in SHH, high expression of DLEU2 and DSCR8

Discussion
LncRNAs are known regulators of gene expression. Disruptions in gene regulatory pathways in cancers dictate the aberrant LncRNAs expression [11][12][13]. Notably, almost 40% of lncRNAs are aberrantly expressed in the brain-related disorders including brain tumors. However, lncRNA expression profile in MB is largely unexplored. In this study, we have identified the lncRNA expression profile of pediatric MB subgroups and associated molecular pathways. The identified key lncRNAs require further functional validation in vitro and in vivo to explore their potential role in MB subgroup-specific manner. Here, we discuss the known cancer-relevant function of the key lncRNAs identified in MB subgroups.
EMX2OS is the most differentially expressed lncRNA in the WNT subgroup. This lncRNA is known to regulate EMX gene expression in the brain development [32,33]. OTX2-AS1 (antisense strand of the OTX2 gene) is predominantly involved in eye development [34]. High PGM5-AS1 (antisense strand of the PGM5 gene) expression is associated with development and poor prognosis of colorectal cancer (CRC) [35]. Increased expression of DSCR8 is associated to malignant pathology and poor survival in hepatocellular carcinoma (HCC) patients [36]. LOXL1-AS1 (antisense strand of the LOXL1 gene) is involved in the progression and metastasis of MB by regulating the PI3K-AKT signaling [27]. In addition, it is also known to play roles in the proliferation and survival of prostate cancer (PC) cells via miR-541-3p and cell cycle gene CCND1 [37] as well as aggressive nature of glioblastoma by activating NF-kB pathway [38]. HAND2-AS1 (antisense strand of the HAND2 gene) is overexpressed in esophageal squamous cell carcinoma (ESCC) [39] while it is downregulated in non-small cell  lung cancer (NSCLC) cells [40]. TMEM51-AS1 (antisense strand of the TMEM51 gene) is associated with renal cell carcinoma (RCC) [41]. RMST acts as a tumor suppressor in triple-negative breast cancer (TNBC) by inducing apoptosis and inhibiting proliferation/invasion and migration [42]. PART1 promotes gefitinib-resistance in ESCC by regulating the miR-129/Bcl-2 pathway [43] and also associated with PC tumorigenesis [44]. LINC00461 is involved in glioma tumorigenesis via MAPK/ERK and PI3K/AKT signaling pathways [45]. Downregulation of MEG3 is involved in the proliferation and apoptosis of PC cells by regulating miR-9-5p and its target gene QKI-5 [46]. Downregulation of LINC00844 is associated with poor clinical outcomes and suppressed tumor progression/metastasis in PC [47]. SOX2-OT is overexpressed and promotes tumorigenesis by upregulating SOX2 gene and activating PI3K/AKT signaling pathway in cholangiocarcinoma (CCA) [48]. SOX2-OT is also a prognostic biomarker for osteosarcoma (OS) and involved in cell survival and cancer stem cells [49]. TUNAR plays a tumor suppressive role in glioma cells by upregulating miR-200a and inhibiting Rac1 [50]. MALAT1 promotes the chemo-resistance of cervical cancer via BRWD1-PI3K/AKT pathway [51]. MALAT1 is a well-studied lncRNA in several solid and hematological cancers [52]. NEAT1 is overexpressed in most cancer types, except leukemia and myeloma, where it is down-regulated [53][54][55]. DLEU2 exhibits role in the proliferation and survival of laryngeal cancer cells via miR-16-1 [56]. DLEU2 is also significantly overexpressed in gastric cancer and contributes to cell proliferation [57]. TPT1-AS1 (antisense strand of the TPT1 gene) expression is upregulated in cervical cancer and has influence on proliferation and migration   [58]. HCG11 is significantly overexpressed in hepatocellular carcinoma (HCC) and genetic-silencing of HCG11 in HCC cells leads to decreased proliferation [59]. HCG11 expression is downregulated in PC and associated with poor prognosis of patients [60]. CCEPR contributes significantly in promoting cell proliferation and inhibiting apoptosis in bladder cancer [61]. BLACAT1 is overexpressed in chemo-resistant NSCLC and induces autophagy by regulating miR-17 and ATG7 pathway [62]. It also triggers proliferation/ survival by regulating WNT signaling in cervical cancer [63].
XIST is elevated in bladder cancer and inhibits p53 function via binding to TET1 [64]. XIST also binds to miR-34a and elicits proliferation and tumor development in thyroid cancer [65]. XIST is an important regulator of progression and oxaliplatin-resistance in malignant melanoma [66]. MIR100HG is known to be involved in cetuximab-resistance in CRC via the β-catenin cellular pathway [67]. In addition, elevated expression of MIR100HG is correlated with poor prognosis of osteosarcoma [68]. MIAT is overexpressed in clear cell renal cell carcinoma (CCRCC) and associated with poor prognosis [69]. MIAT associates with miR-133 and contributes a role in the progression pancreatic cancer development [70]. MIAT also plays a key role in CRC tumorigenesis via miR-132/Derlin-1 axis [71]. NR2F1-AS1 (antisense strand of the NR2F1 gene) promotes chemotherapy-resistance in HCC by regulating miR-363-ABCC1 drug-transporter pathway [72].

Conclusions
We propose that the majority of DE lncRNAs in MB might have oncogenic properties as seen in other cancers (Supplementary Table S1 in Additional file 3) [73][74][75][76][77][78][79][80][81][82]. We found approximately 25% of these DE lncRNAs in MB are tumor suppressive. Also, each MB subgroup has unique and common lncRNAs in their expression   profile (Fig. 6). We performed a unique lncRNAs analysis in both original datasets and validation datasets (Additional files 1 and 2). Unique lncRNAs can be validated for differential diagnosis and prognosis of MB subgroups. Common lncRNAs and associated molecules in pathways can be important therapeutic targets. We identified important lncRNAs DELU2, CASC15, LINC01355 and GAS5 are present in each subgroup and can be further explored for functional analyses in different MB subgroups. We also found SOX2, Protein kinase C delta (PRKCD), and EZH2 associated with functional networks of each subgroup and could be important drug targets. We also identified the prognostic significance of lncRNAs in different subgroups of MB.