Long non-coding RNAs in brain tumors

Abstract Long non-coding RNAs (lncRNAs) have been found to be central players in the epigenetic, transcriptional and post-transcriptional regulation of gene expression. There is an accumulation of evidence on newly discovered lncRNAs, their molecular interactions and their roles in the development and progression of human brain tumors. LncRNAs can have either tumor suppressive or oncogenic functions in different brain cancers, making them attractive therapeutic targets and biomarkers for personalized therapy and precision diagnostics. Here, we summarize the current state of knowledge of the lncRNAs that have been implicated in brain cancer pathogenesis, particularly in gliomas and medulloblastomas. We discuss their epigenetic regulation as well as the prospects of using lncRNAs as diagnostic biomarkers and therapeutic targets in patients with brain tumors.


Central nervous system tumors
Gliomas represent about a third of all central nervous system (CNS) tumors and include astrocytomas, oligodendrogliomas, ependymomas and glioblastomas. Despite progress in surgery, radiotherapy and chemotherapy, many glioma patients have a poor prognosis (1); for example, the median survival for patients with high-grade glioma is only 15 months (2). A better understanding of the characteristics and pathobiology of gliomas is necessary to improve tumor classification and patient stratification for more targeted treatment. There have been substantial efforts to characterize the gene expression, mutational and epigenetic landscapes of gliomas, not least for the purposes of biomarker identification (3). As a direct result of these efforts, a new and improved glioma classification system that includes molecular biomarkers was approved in 2016 (4). Diffuse astrocytic and oligodendroglial tumors (WHO grade II and III) are now separated according to IDH mutations and 1p/19q-codeletion status, while grade IV gliomas are now divided into IDH wild-type, IDHmutant, glioblastoma not otherwise specified and diffuse midline glioma, H3 K27M-mutant. In adults, about 32% of brain tumors are malignant, of which about half are glioblastomas, about one-fifth are diffuse astrocytomas and about one in twenty are oligodendrogliomas and ependymal tumors (5).
In children, medulloblastomas (MBs) are the most common brain tumors (6)(7)(8), representing ∼20% of pediatric brain tumors but only 1% of adult cases (6,9). MBs are poorly differentiated 'embryonal' tumors arising in in the cerebellum and are highly malignant; they commonly metastasize to other parts of the brain and to the spinal cord and, occasionally, to extraneural sites (10,11). Transcriptional programs in MBs have been shown to mimic developmental cerebellar lineages, highlighting their embryonic origin (12). The clinical management of MBs depends on several factors including molecular and histopathological tumor subtype, stage and extent of resection, location and overall patient health. Treatment strategies tend to be aggressive, consisting of a mixture of surgical resection, radiotherapy, chemotherapy and stem cell/bone marrow transplantation. Despite advances in their diagnosis and treatment, MB remains deadly in 35-40% of cases and those patients that do survive often suffer long-term sideeffects including organ dysfunction, neurocognitive impairment, endocrine disabilities and secondary tumors (13)(14)(15)(16). In the major 2016 WHO re-classification of CNS tumors, MBs were subdivided into following major molecular subgroups (8): wingless-type (WNT)-activated MBs (10%; children and adults, associated with very good prognosis); sonic hedgehog (SHH)-activated MBs (30%; intermediate prognosis, infants and adults; further characterized as TP53 mutant or TP53 wild-type); and two provisional non-WNT/SHH subgroups: group 3 MBs (25%; poor prognosis, infants and children) and group 4 MBs (35%; intermediate prognosis, children and adults) (reviewed in (17,18)). Despite this progress, the underlying subtype-specific pathobiology and molecular processes in MB are not fully defined, hampering the exploitation of subgrouping to personalize therapy for individual patients.

Long non-coding RNAs
Given rapid developments in genome and transcriptome sequencing technologies and the implementation of genomics consortia such as ENCODE and FANTOM, the classical, mRNA-centric view of the transcriptomic landscape has undergone fundamental changes (19)(20)(21). The genome serves as a template not only for coding RNAs but for also a far greater number of non-coding RNAs (ncRNAs). Of the non-coding RNAs, long non-coding RNAs (lncRNAs), which are >200 nucleotides in length, have been widely investigated and identified as key regulators of various biological processes, including cellular proliferation, differentiation, apoptosis, migration and invasion (22)(23)(24)(25). Recent studies have shown that lncRNAs play key roles in a wide range of cellular processes by regulating gene expression at the epigenetic, transcriptional and post-transcriptional levels (26)(27)(28). LncRNAs may be classified into several archetypes based on their mode of action and functions: (i) Molecular signals. LncRNAs can function as molecular signals to induce transcriptional activity. LncRNAs display tissue-specific expression and respond to varied cellular stimuli, highlighting their tight transcriptional control. They can act as molecular signals through exquisite orchestration of their transcription and subcellular location, allowing the integration of responses to different stimuli (29 (v) Competing endogenous RNAs (ceRNAs). Some evidence shows that lncRNAs may be potent natural microRNA (miRNA) sponges. LncRNAs containing miRNA binding sites de-repress miRNA target genes through complementary binding to miRNAs (35). (vi) Enhancers. Enhancers are a signature feature of the regulation of gene expression during development and play a key role in specifying cell identity. There are up to 400 000 enhancers in the human genome, many clustered as so-called 'super-enhancers' (36). Enhancers transcribe to produce noncoding RNAs specifically in the cells in which they are active and appear to be integral to enhancer action by looping of regulatory elements to affect target gene transcription and the formation of topologically associated domains in the process (36,37).
Dysregulated lncRNA expression could potentially alter basic cell biology and contribute to tumorigenesis of all the tumors listed above (38)(39)(40). LncRNAs are widely expressed in the brain and during neurological development, mostly in highly specific patterns and cell types (41,42) and play important roles in brain tumor pathogenesis via multiple gene expression control mechanisms ( Figure 1). In this review, we explore the expression profiles and functions of lncRNAs in brain tumors, especially gliomas and MBs, and discuss their potential use as diagnostic and prognostic biomarkers and therapeutic targets.

Oncogenic lncRNAs in gliomas
LncRNA-ATB is a lncRNA induced by TGF-␤ (ATB) that is widely expressed in different types of cancer (43). It functions in part through upregulation the zinc finger and homeodomain-containing transcription factors ZEB1 and ZEB2 by competitively binding the miR-200 family and inducing endothelial-mesenchymal transition and invasion. It also promotes organ colonization of disseminated tumor cells by binding IL-11 mRNA, autocrine induction of IL-11 and triggering STAT3 signaling (44). The expression of lncRNA-ATB positively correlates with glioma grade and is negatively associated with survival in glioma patients. LncRNA-ATB is targeted by miR-200a, which affects TGF-β2 (transforming growth factor-beta 2) expression. In turn, TGF-␤2 activates the MAPK (mitogen-activated protein kinase) signaling pathway (45). Downregulation of lncRNA-ATB suppresses tumor growth in vivo (46).
The lncRNA CCAT1 is involved in cellular proliferation and is a miR-410 target (54). miR-410 expression was found to be suppressed in glioma tissue, leading to MET/Akt pathway activation (55). CCAT1 enhanced FGFR3 (fibroblast growth factor receptor 3) and PDGFRA (plateletderived growth factor receptor A) expression by targeting miR-181b; both growth factors activate MAPK signaling (55). CCAT1 also promoted tumor growth in vivo (56).
CRNDE lncRNA was found to be highly upregulated in glioma tissue and it promoted tumor proliferation, migration and invasion (57). CRNDE induced gliomagenesis in vivo. CRNDE is a miR-384 target, and both regulate PIWIL4 (piwi-like RNA-mediated gene silencing 4) protein, which promotes progression by inducing STAT3 (signal transducer and activator of transcription 3) phosphorylation (58). Upregulated expression of CRNDE induces EGFR (epidermal growth factor receptor) activation, resulting in apoptosis inhibition through an increased Bcl2 (B-cell CLL/lymphoma 2)/Bax (BCL2-associated X) ratio and an association with poor survival in glioma patients (59).
NEAT1 lncRNA is a widely expressed lncRNA that is essential for the formation of paraspeckles, enigmatic mammalian-specific subnuclear organelles that are important for placental biology and may act as reservoirs of edited transcripts and/or transcripts that are important for stress responses (75). NEAT1 has associated with epilepsy (76), and paraspeckle components are implicated in motor neuron disease (75). NEAT1 overexpression in glioma tissue positively correlates with glioma grade (77), and re-duced NEAT1 expression is associated with longer survival in glioma patients (78). NEAT1 activates proliferation, migration and invasion and deregulates apoptosis, possibly by interacting with miR-449b-5p, which activated the MET oncogene (77). In addition, NEAT1 is a direct target of the miRNA let-7e and activates cellular proliferation via PI3K/AKT/mTOR (mammalian target of rapamycin) and mitogen-activated protein kinase (MAPK)/extracellularsignal-regulated kinase (ERK) signaling (79). NEAT1 also activates MET by silencing miR-449b-5p (77). Recently, Yang et al. discovered that NEAT1 increased CDK6 (cyclindependent kinase 6) expression via miR-107 in glioma cell lines (80). On the other hand, PVT1 lncRNA expression is increased in glioma tissue, is positively associated with poor outcomes and alters EZH2 expression. PVT1 also regulates cell cycle and apoptosis in vitro and in vivo (81,82). PVT1 regulates the expression of MEF2C (myocyte enhancer factor 2C) and JAG1 (Jagged1) by binding to miR-190a-5p and miR-488-3p (83).
TUG1, a lncRNA first identified as being involved in retinal development, is widely expressed in the brain. It is also expressed in T cells (84) and is required for male fertility (85). TUG1 also upregulation activates caspases by inhibiting Bcl-2 expression, suggesting involvement in the regulation of programmed cell death and immune responses (86). Furthermore, TUG1 maintained glioma stem cells through interactions with polycomb repressive protein 2 (PRC2) components (EZH2 and SUZ12 (SUZ12 polycomb repressive complex 2 subunit)) and transcription factor YY1 (Yin and Yang 1 protein), thereby epigenetically suppressing multiple neuronal differentiation-associated genes (87).
FOXD3-AS1 lncRNA has been shown to be involved in cell proliferation, migration and invasion, associated with poor prognosis and correlated with glioma grade. FOXD3-AS1 overexpression reduces transcription factor FOXD3 (forkhead box D3) levels to participate in differentiation, proliferation, migration and apoptosis (97). ZFAS1 lncRNA activates cell proliferation, invasion and migration by activating EMT and Notch signaling pathways (98). Linc-OIP5 lncRNA is upregulated in glioma tissue and positively correlates with glioma grade. Linc-OIP5 induces proliferation and migration through NOTCH1 (Notch homolog 1, translocation-associated), YAP (yes-associated protein 1), JAG1 and HES1 (hairy and enhancer of split-1) and downregulating its expression reduced tumor growth in vivo (99).

Tumor suppressive and context-dependent lncRNAs in gliomas
The majority of available evidence shows that the expression of the lncRNA HOTTIP is mainly downregulated in gliomas. However, its overexpression inhibited proliferation and induced apoptosis in vitro. Also, there was a negative correlation between HOTTIP and BRE (BRCA1-A complex subunit) (100). BRE plays an essential role in preventing replicative and DNA damage-induced premature senescence (101). In addition, reduced BRE expression downregulated CCNA (cyclin A) and CDK2 expression and increased p53 expression (100). Interestingly, another study showed that HOTTIP was overexpressed under hypoxic conditions, and HOTTIP expression in metastatic glioma samples was higher than in glioma samples without metastasis. Zhang et al. demonstrated that HOTTIP contributed to EMT and metastasis in glioma by targeting miR-101 and elevating ZEB1 transcription, itself a target of miR-101 (102).
MALAT1 lncRNA has been shown to be downregulated in glioma tissue and cell lines, and high MALAT1 levels are associated with improved survival in glioma patients. MALAT1 also negatively regulates miR-155 expression, and FBXW7 (F-box and WD repeat domain containing 7) is a validated target of miR-155. MALAT1 suppresses glioma cell viability by downregulating miR-155 and promoting FBXW7 expression (103). Han et al. also measured MALAT1 expression in glioma specimens and cell lines and found that MALAT1 expression is higher in non-cancerous brain tissues, albeit with no differences between grades. Overexpression of MALAT1 is associated with MAPK pathway suppression (104).
In apparent contradiction, MALAT1 expression has been shown by others to be highly expressed in glioma tissue and cell lines (105). MALAT1 knockdown suppresses glioma cell proliferation and induces apoptosis, together with downregulated CCND1 and MYC expression (105). Fu et al. also detected an increase in MALAT1 expression in gliomas, concluding that MALAT1 can induce proliferation and autophagy and target miR-101, which regulates not only ZEB1, but also STMN1 (stathmin 1), RAB5A (RAS-associated protein RAB5A) and ATG4D (autophagy-related 4D cysteine peptidase) expression (106). The variable reports of the expression patterns of MALAT1 in gliomas may be due to the complexity of ncRNA biology and glioma heterogeneity or the vagaries of assessing differential gene expression, which may apply more broadly.
TUSC7 lncRNA is reported to be downregulated in glioma tissue and negatively associated with overall survival and miR-23b expression (107). In addition, Jiang et al. showed that miR-23b regulated the PI3K/Akt signaling pathway (108). Several studies have implicated MEG3 lncRNA in glioma pathogenesis (109)(110)(111). MEG3 is downregulated in glioma cell lines, and its overexpression increased cell death and inhibited proliferation (110). Other studies have shown that the MEG3 promoter can be hypermethylated by DNA methyltransferase 1 (DNMT1) in glioblastoma tissue, thereby completely silencing the gene (112). Additionally, MEG3 acted as a ceRNA to miR-19a, which repressed PTEN (phosphatase and tensin homolog) expression, leading to glioma cell proliferation, migration and invasion (113).
RAMP2-AS1 lncRNA is downregulated in glioblastoma tissues, and its ectopic overexpression reduces tumor growth in a subcutaneous mouse model (114). This lncRNA interacts with the DHC10/NOTCH3/HES1 signaling pathway and plays a tumor-suppressive role in glioblastoma progression through inhibition of NOTCH3 (114). CASC2 lncRNA is downregulated in glioma tissues and glioblastoma cell lines (115). CASC2 overexpression inhibits glioblastoma cell proliferation, migration and invasion via miR-21, and miR-21 overexpression abrogated CASC2-induced inhibitory effects (115). Mechanistically, miR-21 bound to CASC2 in a sequence-specific manner, leading to the transcriptional repression of CASC2. Similarly, CASC2 overexpression suppresses the Wnt/␤-catenin signaling pathway in glioblastoma cells (116). Glioblastoma cells display resistance to temozolomide (TMZ), at least in part, through increased autophagy. Interestingly, TMZ sensitivity was CASC2-dependent in glioma stem cells. In one study, CASC2 overexpression inhibited autophagy and increased glioma stem cell susceptibility to TMZ by the accumulation of lipid peroxides, leading to cell death (117). Autophagy inhibition mediated by CASC2 upregulation may be due to sponging miR-193a-5p and regulating mTOR expression (117). Table 1 summarizes the lncRNAs implicated in glioma and their molecular partners or genomic targets that mediate glioma progression.

Dysregulated lncRNAs in medulloblastoma
Knowledge of lncRNA function in MB is still fragmentary, and there are only a few articles on the topic. Here we discuss recently reported mechanistic insights into how lncR-NAs regulate gene expression and contribute to MB formation.
PVT1 is a non-coding host gene for four miRNAs (miR-1204, miR-1205, miR-1206 and miR-1207) and is amplified together with MYC in group 3 MBs. PVT1 fusion genes are highly recurrent, restricted to group 3 MBs, arise through a chromothripsis-like process and were the first recurrent translocation reported in MB (119). The PVT1 locus is thought to be genomically fragile, as the majority of MYCamplified group 3 MBs harbor PVT1 fusions. The identified PVT1 fusions involve only PVT1 exon 1 and miR-1204, and only miR-1204 and not the adjacent miR-1205, miR-1206 and miR-1207 are expressed at higher levels in PVT1-MYC fusion group 3 MBs compared to non-fusion cases. miR-1204 inhibition reduced proliferation of group 3 MB cells at a level comparable to MYC knockdown, while an MB cell line with neither MYC amplification nor detectable PVT1-MYC fusion gene was unaffected by miR-1204 knockdown (119). PVT1 stabilizes MYC protein to promote tumorigenesis, and the PVT1 locus is often amplified in breast cancer (120). However, PVT1 is also frequently disrupted by recurrent translocations or deletions in many cancer types, including MB, suggesting that it may have additional regulatory functions. Of note, a recent study showed that PVT1 lncRNA expression was not required to inhibit MYC transcription; instead, the PVT1 promoter competed with MYC for enhancer binding at the PVT1 locus, preventing MYC promoter firing and suppressing transcriptional elongation of the MYC oncogene and reducing cancer cell growth (121). This might indicate a lncRNAindependent tumor-suppressive role for the PVT1 promoter in MB and suggest that regulatory sequences in lncRNA genes may contribute to tumorigenesis.
Zhang et al. measured HOTIAR expression in human clinical MBs samples and cell lines and found that HOTAIR expression is higher in MB tissues and cell lines than normal samples. HOTAIR can promote MB cell growth, migration, invasion, EMT and inhibit cell apoptosis by negatively regulating miR-1 and miR-206 and activating YY1 expression (122). CRNDE lncRNA was found to be highly upregulated in cisplatin-treated MB cells, and it promoted tumor proliferation, apoptosis, migration, invasion and resistance to chemotherapeutics. Mechanistically, CRNDE functions

Interacts with EZH2
YAP-NOTCH pathway (197,198) Linc-ROR Inhibition of glioma cell growth Inhibition of self-renewal of glioma cell Unknown (215,216) as a ceRNA that directly binds to and negatively regulates miR-29c-3p, thereby mediating MB growth and drug resistance (123). The recent oncogenic lncRNA characterized in MB was TP73-AS1 (124), a transcript antisense to p73 mRNA encodes for a p53 TF family member. p73 is implicated in brain development (125) and cancer (126), including MB (127). Comparative expression analyses of TP73-AS1, using the R2 platform1 and the Cavalli cohort (128), revealed its upregulation in SHH-MB, where TP73-AS1 promoter is significantly hypomethylated (125). In SHH-MB cells, Varon et al. tested the role of TP73-AS1 and found that knocking down TP73-AS1 inhibited cell proliferation, cell migration and increased apoptosis, irrespective of TP53 mutant status. Supporting the genetic independence of TP73-AS1 and P53 family, the TP73 knockdown did not impact the levels of TP73-AS1. In vivo experiments showed that knocking down TP73-AS1 reduced tumor growth, leading to longer survival time in tumor-carrying mice (124). Li et al., found TP73-AS1 acts as a potential oncogene and regulates proliferation, migration, invasion and apoptosis by targeting miR-494-3p in MB cells (129).
Our group has recently analyzed publicly available microarray and RNA-seq datasets by using machine-learning statistical algorithms (random-forest and Lasso) to identify a group of putative lncRNA signatures that may be able to differentiate medulloblastoma subgroups accurately (130). Among those, we found lnc-HLX-2-7 lncRNA was highly upregulated in group 3 MBs compared to other groups. Knockdown of lnc-HLX-2-7 significantly reduced cell growth and induced apoptosis in vitro and in vivo. lnc-HLX-2-7 regulates modulation of cell viability, cell death and energy metabolism signaling pathways (131). Table  2 summarizes the lncRNAs implicated in MB and their molecular partners or genomic targets that mediate MB phenotypes of proliferation, growth suppression, migration, invasion and metastasis.

Epigenetic control of lncRNA expression in brain tumors
Overexpression or repression of key lncRNAs can be mediated by aberrant DNA methylation or histone modifications at their gene promoters. Dysregulated expression of such lncRNAs may further contribute to tumorigenesis through aberrant epigenetic changes in gene regulatory mechanisms. Emerging evidence indicates that lncRNAs constitute a network of epigenetic modulators by forming platforms for chromatin-remodeling complexes and transcription factors capable of regulating the transcriptional state of lncRNA-controlled genomic loci (132,133). Several studies have demonstrated that lncRNA expression can be controlled by DNA methylation at promoter regions and by direct interaction with histone modifiers and chromatinremodeling complexes (33,134,135). MEG3 downregulation in gliomas was recently found to be due to hypermethylation of its promoter. Treatment of glioma cells with the DNA methylation inhibitor 5-AZA-2'-deoxycytidine decreased aberrant promoter hypermethylation and prevented the loss of MEG3 expression. DNMT1 expression is inversely correlated with the MEG3 transcript, supporting DNMT1's involvement in MEG3 dysregulation in gliomas  (112). Importantly, the same association between MEG3 expression and promoter methylation has been observed in hepatocellular carcinoma, and direct involvement of miR-29a, a DNMT1 and DNMT3B epigenetic modulator, was also demonstrated (136). The activity of the tumor suppressor ADAMTS9-AS2 is also controlled by DNA methylation in gliomas, and its expression is positively correlated with ADAMTS9 and DNMT3A expression (137). A recent study showed regulation of LOC285758 expression by DNA methylation that differed within glioma subtypes, with overexpression and higher promoter hypomethylation in glioblastomas (138). Histone modifications play an essential role in the regulation of lncRNA expression. LncRNA promoters exhibit specific histone marks, including trimethylated H3K4 (H3K4me3), H3K27 (H3K27me3), H3K36 (H3K36me3), and acetylated H3K9 (H3K9ac) and H3K27 (H3K27ac), suggesting that they undergo epigenetic regulation similar to that of protein-coding genes (139). Genomewide analysis in different cell and tissue types has revealed that highly expressed lncRNAs were enriched in active H3K4me3 and H3K36me3 sites, whereas those with low expression were highly marked by H3K27me3. Zhao et al. proposed an integrative method to identify epigenetically altered lncRNAs and their associated histone modifications. By combining RNA-and ChIP-seq data for H3K4me3 and H3K27me3 marks, the authors found that many disease-related lncRNAs were associated with aberrant epigenetic modifications and that their neighboring genes were enriched for PRC2-binding sites (140). For example, fetal tissues and glioblastoma expressed high levels of HOTAIRM1 (141) and showed a gain of H3K4me3 in glioma stem cells compared to in glioblastoma multiforme tissue. Epigenetically activated HOTAIRM1 can also negatively or positively affect the expression of its neighboring genes. Indeed, HOTAIRM1 plays key roles in myeloid differentiation by cis-regulating the expression of neighboring HOXA genes (142). In this way, H3K4me3-activated HO-TAIRM1 could contribute to glioblastoma development by enhancing the expression of neighboring HOXA1, an oncogene present in several cancers. Hyperacetylation of H3K9 in the promoter region may account for the upregulation of CRNDE, a key lncRNA involved in neurogenesis and probably also in gliomagenesis. H3K9ac is a known mark for actively transcribed promoters, and the acetyltransferase P300/CBP protein responsible for its acetylation was shown to be recruited to the CRNDE promotor in glioblastoma cells but not in normal glial cells (143).
Epigenetic modulators such as BRD4 (member of BETbromodomain and extra terminal domain) proteins with histone acetyltransferase and chromatin remodeling functions upregulate HOTAIR expression in glioma cells. ChIP analysis demonstrated that BRD4 binds to the HO-TAIR promoter, suggesting that BET proteins can directly regulate HOTAIR expression. Moreover, the treatment of glioblastoma cells with the BET inhibitor I-BET151 restored the expression of several other glioblastomadownregulated lncRNAs. Conversely, HOTAIR overexpression in conjunction with I-BET151 treatment abrogated the BET bromodomain inhibitor's anti-proliferative activity in glioblastoma cells (64). Also, several lncRNAs func-tion as scaffolds for the PRC2 complex and guide these complexes to their endogenous targets. HOTAIR regulated cell cycle progression in glioma cells via interactions with EZH2, and inhibition of HOTAIR repressed glioblastoma tumor growth in vivo (144).
TUG1 represents another example of complex epigenetic gene regulation by lncRNAs. In brain development, by binding to PRC2 components, TUG1 selectively regulates neuronal differentiation-associated genes. In gliomas, TUG1 promotes locus-specific methylation of histone H3K27 via YY1 binding (87). The YY1 and PRC2 binding regions of TUG1 are evolutionarily conserved in mice and humans, suggesting their importance for TUG1 function in tumorigenesis (145). The large number of epigenetically dysregulated lncRNAs reveals that epigenetic inhibitors may induce anti-tumor effects in brain cancers by modulating lncRNA networks. Thus, the expression of lncRNAs is mediated by DNA methylation and histone modifications, while lncRNAs themselves dynamically regulate epigenetic modifications, both normally and in brain tumors ( Figure 2).

Therapeutic potential of lncRNAs in brain tumors
New findings on lncRNAs in brain tumors are beginning to be incorporated into clinical trial data. LncRNAs may affect clinical research and patient management in several ways. Some data associate the expression of certain lncR-NAs with patient prognosis (146). Genome-wide lncRNA expression profiles in glioma patients have revealed that dysregulated lncRNA expression plays important role in tumorigenesis and malignant progression, and several lncR-NAs are related to prognosis patients with glioblastoma (147).
For glioma patients, surgery followed by chemotherapy is an effective treatment. However, the efficacy of chemotherapy remains poor due to the development of drug resistance (148). Suppressing drug resistance in glioma cells is thus a challenge for the treatment of glioblastoma (149,150). Accumulating evidence confirms that dysregulated lncRNAs play important roles in the drug resistance of many cancer types, including glioma.
Liu et al. reported changes in the expression levels of several lncRNAs in response to genotoxic stress. By detecting the expression patterns of lncRNAs in human glioma cell lines treated with doxorubicin and resveratrol, they found that GAS5 was upregulated during doxorubicininduced apoptosis, NEAT1 and MIR155HG were upregulated in response to resveratrol-induced apoptosis, and MEG3 and ST7OT1 are upregulated in response to apoptosis induced by both agents. TUG1, BC200 and MIR155HG were downregulated during necrosis induced by high-dose doxorubicin. These results may indicate that cellular defenses against genotoxic agents are controlled by distinct lncRNAs (151).
TMZ-based chemotherapy is one of the most widely used treatments for glioma and its use has contributed to overall survival gains in glioblastoma patients (152,153). Several lncRNAs have been found to be involved in chemoresistance to TMZ in glioma cells. H19 is significantly upregulated in TMZ-resistant glioma, and H19 knockdown decreased the IC50 values for TMZ and increased rates of apoptosis in TMZ-resistant glioma cells by altering the expression of major drug resistance genes (154). MALAT1 reduced the sensitivity of glioma cell lines to TMZ by enhancing EMT and upregulating multidrug resistance-associated protein expression (155). CASC2 sensitized glioma cells to TMZ and amplified the TMZ-induced suppression of cell proliferation by upregulating PTEN protein expression and downregulating phosphorylated AKT through miR-181a (156).
The DNA repair proteins MGMT and SP1 play important roles in TMZ resistance and are upregulated in TMZresistant glioma cell lines (157). MiR-29c regulated SP1 and MGMT expression to enhance glioma cell chemosensitivity to TMZ. XIST directly targeted miR-29c and inhibited miR-29c expression from amplifying glioma cell line chemoresistance to TMZ (158). These findings indicate that lncRNAs have potential as therapeutic targets in TMZresistant glioma. Circulating levels of HOTAIR are significantly correlated with high-grade brain tumors, and one study demonstrated that this lncRNA could be considered a novel prognostic and diagnostic biomarker for glioblastoma (159). Moreover, aberrant lncRNA expression is correlated with response to therapy in glioma in vitro and in vivo. For example, MALAT1 knockdown increased the permeability of the blood-tumor barrier, which might help to improve tumor penetration with therapeutics (160), while restoration of CASC2 expression upregulated PTEN and increased glioma sensitivity to TMZ-based chemotherapy (161).
There are also examples of lncRNA in MB, leading us to distinguish subgroups of this tumor more precisely. The two most heterogeneous, and otherwise closely related and difficult to distinguish, MB subgroups are groups 3 and 4. It has been reported that some lncRNAs, such as ARHGEF7-AS2, lnc-HLX-1, lnc-EXPH5-2, lnc-CH25H-2 and lnc-TDRP-3, demonstrate differential expression in these two groups, and that other lncRNAs were subgroup-specific: lnc-CCL2-2 in the WNT subgroup, lnc-ABCE1-5 in the SHH subgroup, USP2-AS1 in group 3 and lnc-TBC1D16-3 in group 4 (130). Another lncRNA, lnc-FAM84B-15 (CCAT1), was upregulated in WNT and group 3 of MB, activating the MAPK pathway and significantly stimulating proliferation and metastasis (162).
Compared to protein-coding transcripts, targeting lncR-NAs is challenging. Lack of protein products means that only RNA-based tools are usable. Unlike proteins with specific domains that are easy to target with small molecule drugs, lncRNA conformations are poorly understood, making structure-based strategies difficult to develop. However, several approaches have been proposed to reduce the expression of lncRNAs or inhibit their function (Figure 3). One of the most explored methods to inhibit lncRNAs is the delivery of synthetic oligonucleotide-based molecular products (163). Small interfering RNAs (siRNAs) are double-stranded RNA oligonucleotides, antisense and complementary to target lncRNAs. They induce the degradation of their target by recruiting the RISC (RNA-induced silencing) complex. SiRNA targeting HOTAIR has been shown to suppress the progression of endometrial carcinoma in vivo, demonstrating that targeting HOTAIR can be a novel therapeutic strategy for endometrial cancer (164). The delivery of nanoparticles containing siRNAs against lncRNA DANCR prolonged the survival of AML (Acute Myeloid Leukemia) mouse model (165). Antisense oligonucleotide (ASO) are single-stranded DNA oligonucleotides duplexing by base complementarity their target lncRNAs to promote their degradation by RNase H (166,167). For example, a study showed that using ASOs targeting TUG1 coupled with a drug delivery system induced glioma cell differentiation and repressed tumor growth in vivo (87). Aptamers could also provide greater specificity. The chemical structure of aptamers could be modified to enhance their stability and half-life. Aptamer-based -therapeutics are undergoing clinical trials (non-small cell lung cancer, renal cell carcinoma and AML) (168). Fatemi et al. proposed innovative therapeutic strategies based on small molecule inhibitors that disrupt lncRNA-protein interactions (169,170). New genome editing strategies such as CRISPR/Cas9 seem to be powerful tools to knock-in or knock-out lncRNA candidates (171). Modified CRISPR systems could also generate cytidine substitution into uridine to correct oncogenic single nucleotide polymorphisms (SNPs), knowing that numerous SNPs have been associated with potential predictive biomarkers for the risk of cancer, including SNPs in ANRIL, MALAT1, HULC and PRNCR1 lncRNA (172,173).
The ongoing case-control observational study sponsored by Peking Union Medical College Hospital (NCT03738319) will analyze the expression of miRNAs and lncRNAs in blood samples from 160 patients with high grade serous ovarian cancer and benign gynecologic diseases. Differential expression of miRNAs and lncRNAs will be compared between cancer and control groups, and candidate ncRNAs will be validated as biomarkers for the detection and prognosis of ovarian cancer. Hopefully, completing this trial will lead to promising clinical lncRNA biomarkers that can aid in early detection of this disease, which is often not diagnosed until an advanced stage (174).
H19 is a lncRNA with several oncogenic mechanisms of action and is often upregulated in different types of cancer (175). Due to its upregulation in cancer cells, researchers have used its promoter to deliver a cancer-specific expression of downstream sequences. One compound, BC-819 (also known as DTA-H19), is a DNA plasmid expressing diphtheria toxin A (DTA) under the regulation of the H19 promoter, delivered as a complex with polyethyleneimine (PEI). BC-819 has completed phase 1/2a trials in pancreatic, ovarian and bladder cancers, with current clinical development focused on bladder cancer, where H19 is often suppressed in normal cells but upregulated in tumor cells (176). While not directly targeting an lncRNA, this is an interesting example of how ncRNAs properties, like H19 cancer-specific expression, can be harnessed for novel cancer therapeutic interventions.

CONCLUSION
Despite major efforts directed toward improving brain tumors treatment, prognosis remains poor and the median overall survival time of patients with high-grade tumors remains abysmal, even with aggressive surgery, radiotherapy and chemotherapy. A lack of available tools for early diagnosis of brain cancers and the development of resistance to chemotherapy limit current treatment efficacy. Recent rapid developments in characterizing the expression profiles and functions of lncRNAs in brain tumors have generated excitement in the field of brain tumor research and therapy over the past few years. Due to the rapid development of high-throughput RNA-seq technology, a set of lncRNAs have been identified as dysregulated in brain tumors and associated with tumor progression and poor clinical outcomes. Therefore, lncRNAs have the potential to serve as biomarkers. The material reviewed showed that lncRNAs could serve as patient biomarkers for tumor types and or grades.
Moreover, lncRNAs play key roles in brain tumor progression, and altering their expression level can affect the proliferation, apoptosis and invasion of glioma cells. A study showed that the lncRNA TUG1 responded to Notch signaling and promoted the self-renewal of glioma cells and that using ASOs targeting TUG1 coupled with a drug delivery system induced glioma cell differentiation and repressed tumor growth in vivo (87). This study indicates that lncRNAs have the potential as anti-glioma therapies. As lncRNAs also participate in drug resistance, they could potentially serve as sensitizers when combined with traditional chemotherapy. LncRNA-based cancer therapies are promising. Future work must be needed to screen for candidate lncRNAs and develop delivery strategies to translate these molecular data to the clinic.
A recent study functionalized lncRNAs in drug resistance by integrated genome-wide CRISPR activation of 14 701 lncRNAs and found that growth arrest-specific six antisense (GAS6-AS2) triggered hyperactivation of the GAS6/TAM (TYRO3-AXL-MER2K) pathway, which is a known drug resistance mechanism in multiple cancers (177). With further identification of the molecular mechanisms of lncR-NAs, especially the regulatory networks between lncRNAs and other oncogenic and tumor suppressor genes, we postulate that soon lncRNAs could be used as novel biomarkers and therapeutic targets for patients with brain tumors.