Abstract

   Gliomas are the most aggressive type of malignant brain tumors. Recent
   studies have demonstrated that the existence of glioma stem cells
   (GSCs) is critical for glioma recurrence, metastasis, and chemo- or
   radio-therapy resistance. Temozolomide (TMZ) has been used as an
   initial therapy for gliomas. However, the overall survival time is
   still limiting due to the lack of effective targets and treatment
   options. Therefore, identifying novel biomarkers for gliomas,
   especially for GSCs, is important to improve the clinical outcome in
   the future. In this study, we identify a human-specific long non-coding
   RNA (lncRNA, ENSG00000250377), termed GSCAR (glioma stem cell
   associated lncRNA), which is highly expressed in glioma cancerous
   tissues and cell lines. We reveal that GSCAR positively correlates with
   tumor grade. Glioma patients with GSCAR high expression exhibit
   shortened overall survival time, compared to patients with GSCAR low
   expression. Furthermore, we show that GSCAR knockdown by shRNAs or
   antisense oligonucleotide (ASO) reduces tumor cell proliferation,
   migration and xenograft tumor formation abilities. Mechanistic study
   shows that GSCAR acts as a ceRNA (competing endogenous RNA) for
   miR-6760-5p to promote the expression of oncogene SRSF1 (serine and
   arginine rich splicing factor 1). In addition, GSCAR mediates the
   protein complex formation between DHX9 (DExH-Box helicase 9) and
   IGF2BP2 (insulin-like growth factor 2 mRNA-binding protein 2), leading
   to the stabilization of SOX2 (sex-determining region Y-box 2) mRNA and
   then the transcriptional activation of GSCAR. Depleting GSCAR reduces
   SOX2 expression and GSC self-renewal ability, but promotes tumor cell
   responses to TMZ. These findings uncover that GSCAR/miR-6760-5p/SRSF1
   axis and GSCAR/DHX9-IGF2BP2/SOX2 positive feedback loop are critical
   for glioma progression, which could be used as prognostic biomarkers
   and therapeutic targets in the future.

   Keywords: GSCAR, miR-6760-5p, SRSF1, SOX2, glioma stem cells (GSCs)

Introduction

   Gliomas account for approximately 30% of all brain tumors, and
   pilocytic astrocytoma (WHO grade I) is the least malignant subtype,
   which can progress to most malignant glioblastoma (GBM, WHO grade IV),
   and the average survival time for GBM patients is approximately 15
   months after diagnosis [51]^1. Gliomas are characterized by intense
   neovascularization with unusual vessel-like structures and are commonly
   resistant to radio- or/and chemotherapies, which leads to tumor
   relapses and poor prognosis. During the past decades, dedicated studies
   in gliomas have resulted in the identification of multiple key genetic
   and molecular underpinnings, which contribute to the new classification
   for gliomas [52]^2. Mutations in the IDH1/2 have been identified in
   gliomas, and IDH-mutant low-grade gliomas (LGGs) may develop malignant
   transformation after further genetic alterations, such as Myc, PTEN,
   KRAS, PIK3CA, and MET, are acquired. However, the pathological
   consequences resulting from IDH mutation remain elusive [53]^3. To
   date, surgical resection, temozolomide (TMZ)-dependent chemotherapy,
   radiotherapy, and bevacizumab treatment are the conventional therapies
   for gliomas, which are still far from sufficient in combating tumor
   progression [54]^4.

   The different locations in the brain and the regulatory molecular
   events may generate various types of neural stem and progenitor cell
   (NSPC) pools, and glioma stem cells (GSCs) with self-renewing and
   tumorigenic abilities have also been identified, which are resistant to
   standard chemo- and radio-therapies, indicating their critical role in
   tumor recurrence and metastasis [55]^5. Our group and others have
   recently identified that GSCs develop multiple molecular mechanisms to
   mediate therapeutic resistance, including hypoxia, Notch, EZH2, and DNA
   damage checkpoint-related signaling pathways [56]^6^, [57]^7. Multiple
   biomarkers for GSCs, including SOX2, CD133 and CD44, have been
   documented in recent years, although the underlying mechanisms by which
   these biomarkers are specifically induced in GSCs need to be unraveled
   [58]^7^, [59]^8.

   Recently, an increasing number of findings have shown that noncoding
   RNAs may serve as valuable therapeutic targets for glioma patients
   [60]^9. LncRNA-HOTAIR was highly expressed in high-grade gliomas
   (HGGs), which correlates with a poor survival rate [61]^10. The tumor
   suppressive lncRNA GAS5 was described to inhibit GSC maintenance via a
   miR-196a-5p/FOXO1 feedback loop [62]^11, while FOXM1-AS was found to
   facilitate the interaction of ALKBH5 with FOXM1 nascent transcript,
   leading to GSC activation and glioma progression [63]^12. The Sox2 gene
   has been well documented as a pluripotent factor essential for stem
   cell self-renewal and differentiation [64]^13. Furthermore, increased
   SOX2 related to adverse clinical outcomes in glioma patients,
   suggesting that depleting of SOX2 may be a novel therapeutic approach
   to combat glioma [65]^14. However, the posttranscriptional regulation
   of SOX2 by long noncoding RNAs in gliomas remains unclear.

   Here, we identified a 676-bp lncRNA, termed glioma stem cell
   association long noncoding RNA (GSCAR; ENSG00000250377), that is
   upregulated in glioma cancerous tissues and cell lines, especially in
   GSCs, and is correlated with worse clinical outcomes. We demonstrated
   that GSCAR promotes the growth, migration, and invasion of glioma tumor
   cells by competing for endogenous miR-6760-5p to induce the expression
   of the oncogene SRSF1. In addition, we showed that GSCAR activates the
   self-renewal ability of GSCs by mediating DHX9 and IGF2BP2 complex
   formation, leading to the stabilization of the SOX2 transcript and
   tumor growth. Therefore, we decided to decipher the potential mechanism
   by which the GSCAR/miR-6760-5p/SRSF1 axis and GSCAR/DHX9-IGF2BP2/SOX2
   feedback loop promote glioma progression, which may provide new
   therapeutic targets for glioma in the future.

Methods and Materials

Constructs

   Independent GSCAR, SRSF1, DHX9, SOX2 and IGF2BP2-targeting shRNAs were
   connected to the pLKO.1 vector [66]^15, and all the oligonucleotides
   are indicated in [67]Table S1. Human GSCAR and SRSF1 cDNA was amplified
   by PCR and subcloned into the pCDH-MCSV-E2F-eGFP vector. Lentiviral
   vectors expressing Ctrl shRNA, GSCAR shRNA#1, GSCAR shRNA#2, SRSF1
   shRNA#1, and SRSF1 shRNA#2 were cotransfected into HEK-293T cells with
   pMD2.G and psPAX2 plasmids (Addgene), lentiviruses were packaged. ASOs
   targeting GSCAR, control ASO, miR-6760-5p mimics, and inhibitors were
   ordered from Ruibo and transfected into cells using Lipofectamine 3000.

Chromatin immunoprecipitation (ChIP) assay

   ChIP assay was performed as previously documented [68]^16. Briefly, 9x
   10^6 cells were harvested, and 5 μg of preimmune mouse IgG and
   anti-SOX2 antibodies were used for the ChIP reaction [69]^16. The oligo
   sequences are provided in [70]Table S1.

Tissue microarrays

   Glioma tissue microarrays comprised of 10 normal brain tissues and 60
   glioma tissues annotated with clinical and pathological information
   (Wuhan Servicebio, IWLT-C-70GL61, China) were used to verify GSCAR
   expression via RNA in situ hybridization (ISH). All specimens were
   graded by the pathological and clinical stages ([71]Table S2).

RNA pull-down assay

   For in vitro RNA synthesis, the GSCAR fragment was connected to
   pcDNA3.1, the construct was then linearized, and the RNAs were
   transcribed with T7 RNA polymerase. The Pierce™ RNA 3' End
   Desthiobiotinylation Kit was used to biotinylate sense and antisense
   GSCAR RNAs. These RNAs were then incubated with GSCs cell extracts at 4
   °C. Then using Elution Buffer to elute potential proteins. The obtained
   proteins were then examined by SDS-PAGE followed by immunoblot and mass
   spectrometry detection.

RNA immunoprecipitation assay

   9 x 10^6 GSC cells were lysed in 1 ml of RIP lysis buffer supplemented
   with RNase inhibitors. The GSCs cell lysates incubated with beads
   coated with IgG, anti-IGF2BP2, or anti-DHX9 antibodies on a rotator at
   4 °C overnight. The RNA-protein complexes were washed and then
   incubated with the Proteinase K digestion system. Protein-bound RNAs
   were finally extracted by RNA extraction methods and performed RT-PCR
   examination.

RNA decay assay

   3X10^4 GSC cells were seeded in 6-well plates and treated with
   actinomycin D (5 μg/mL) at 0, 4, and 8 h time points, respectively.
   Total RNAs were then isolated by TRIzol and subjected to RT-PCR.

Mass spectrometry analysis

   Proteins bound on the streptavidin magnetic beads were eluted with the
   Elution Buffer of the Pierce™ Magnetic RNA-Protein Pull-Down Kit
   (17-700, Millipore). The recovered proteins were then examined by mass
   spectrometry detection. All experiments were performed on a Q-Exactive
   mass spectrometer with an ancillary EASY-nLC 1000 HPLC system (Thermo
   Fisher Scientific). The mass spectrometry instrument parameters were:
   MS1 full scan resolution, 70000 at m/z 200; automatic gain control
   target, 3 × 10^6; maximum injection time, 120 ms. The candidate GSCAR
   interacting proteins were indicated in [72]Table S3.

Tumorsphere formation assay

   Briefly, 3x10^4 GSCs cells were transferred to 6-well plates, and the
   spheres were cultured for approximately 14 days, white-field images
   were collected, and the sphere numbers were quantified.

Bioinformatics assay

   Most statistical assays were examined using GraphPad Software 7
   (GraphPad Software Inc., CA, USA). The expression of lncRNAs,
   microRNAs, and mRNAs in Gene Expression Omnibus (GEO), The Cancer
   Genome Atlas (TCGA), and the Genotype-Tissue Expression (GTEx)
   [73]^17^, [74]^18, the survival curves for the prognostic analysis were
   generated via the Kaplan-Meier method [75]^19. The KEGG pathway
   enrichment analysis was performed using the GSEA software [76]^20. The
   specificity and sensitivity of GSCAR, SRSF1 and miR-6760-5p were
   assessed via receiver operating characteristic (ROC) curves, and the
   area under the curve (AUC) was quantified using the pROC R package. The
   correlation was analyzed by Pearson's correlation analysis. The
   significance of the data between two experimental groups was determined
   by Student's t-test, and multiple-group comparisons were analyzed by
   one-way ANOVA. P < 0.05 (*), P < 0.01 (**), and P < 0.001 (***), were
   significant.

Results

GSCAR was highly expressed in gliomas

   To identify the potential oncogenic lncRNAs in gliomas, we
   characterized the lncRNAs located in SCNAs in gliomas using the
   TCGA-LGG dataset, and 24 candidate lncRNAs were selected according to
   the criteria (Relative CNAs in >40% glioma samples; occurring in the
   amplification CNA area; prior to long intergenic non-coding RNA; Log
   FC>4, P<0.0001). To narrow down the potent candidate involved in glioma
   stem cells, we further examined the deregulated lncRNAs in U251/TMZ
   (TMZ resistant cell line) and cancer stem cells ([77]Table S4). We
   uncovered that 4 lncRNAs were unanimously upregulated, including
   ENSG00000250377 (named GSCAR based on its functional role), LINC01060,
   PVT1, and CRNDE. Importantly, GSCAR, but not the other 3 lncRNAs, was
   identified as the only candidate whose functional role in gliomas
   remains elusive (Figure [78]1A and [79]Table S4) [80]^21^-[81]^26.

Figure 1.

   [82]Figure 1
   [83]Open in a new tab

   GSCAR was highly expressed in gliomas. (A) LncRNA GSCAR was identified
   by integrative omics analysis using GEO datasets, TCGA-LGG (blue): data
   generated from low-grade glioma tissue samples, [84]GSE146698 (red):
   data generated from a TMZ-resistant cell line, [85]GSE131744 (green):
   data generated from a glioma stem cell line, and [86]GSE188256
   (yellow): data generated from glioma tissue samples. (B) The relative
   expression levels of GSCAR in TCGA-LGG/GTEx datasets (Normal: 1152,
   Tumor: 523). (C) The relative expression levels of GSCAR in grade II,
   III, and IV glioma patients in the TCGA database (II: 224, III: 243,
   and IV: 168). (D) GSCAR expression was higher in IDH1 wild-type (WT:
   246) patients than in IDH1 mutant (MUT: 440) patients. (E) High GSCAR
   expression correlates with a worse survival rate. OS: overall survival,
   DSS: disease-specific survival, and PFS: progression-free survival. (F)
   The expression of GSCAR in normal brain tissues and glioma tissues was
   examined by ISH assay (Normal: 10, Tumor: 60). Quantification results
   are shown. (G) The ROC curve for GSCAR (AUC=0.971) was examined by the
   TCGA glioma dataset. (H) The relative expression level of GSCAR in
   glioma cancerous cell lines (U87, U251, A172) and glioma stem cell
   (GSC) lines (GBM1, GBM2, GSC11). Fetal normal human astrocytes (NHAs)
   were used as controls. (I) GSCAR was primarily localized in the
   cytoplasm of U251 and A172 cells using the nuclear and cytoplasmic RNA
   fractionation assay followed by RT-PCR examination. ACTB (β-actin:
   cytoplasmic control), U1 (nuclear control). (J) The subcellular
   localization of GSCAR was examined by FISH assay. Scale bar=50 μm. (K)
   The coding probability of GSCAR was predicted by CPAT. (L-M)
   Full-length GSCAR was cloned into an eukaryotic expression vector
   pcDNA3.1 vector/myc with an N-terminal codon ATG in the three
   expression patterns. The blue arrowhead pointing to NCAPH-Myc proteins
   was used as a control. * P < 0.05, ** P < 0.01, *** P < 0.001.

   We first found that GSCAR is specifically expressed in humans
   ([87]Figure S1A) [88]^27. We then confirmed that GSCAR expression
   positively correlated with SCNAs, which resulted in poor clinical
   outcomes ([89]Figure S1B-S1C). The increased expression of GSCAR in
   gliomas was validated in web-available datasets [90]^28, and a
   significant correlation between high GSCAR expression and higher-grade
   tumors was detected (Figure [91]1B-[92]1C). Consistently, we found that
   GSCAR expression was higher in IDH1 wild-type (WT) gliomas than in IDH1
   mutant (MUT) gliomas, and glioma patients with higher GSCAR expression
   exhibited worse clinical outcomes (Figure [93]1D-[94]1E). As expected,
   the increased expression of GSCAR was confirmed in glioma tissue
   microarray examined by ISH assay (RNA in situ hybridization) (Figure
   [95]1F and [96]Figure S1D). The ROC curve was applied to examine the
   diagnostic value of GSCAR in gliomas, which showed that the AUC value
   was 0.971, indicating that GSCAR may serve as an independent prognostic
   biomarker in gliomas (Figure [97]1G). To corroborate the bioinformatics
   results, we then examined GSCAR expression in glioma cancerous cell
   lines and glioma stem cell lines GSC11, GBM1, and GBM2, and fetal
   normal human astrocytes (NHAs) were used as controls [98]^7. We
   revealed that GSCAR was markedly upregulated in glioma cancerous cell
   lines and preferentially higher in GSCs (Figure [99]1H). Consistent
   with the web-source dataset, we revealed that GSCAR was mainly located
   in the cytoplasm, which was further confirmed by the RNA FISH assay and
   the RT-PCR analysis after nuclear and cytosolic fractionation according
   to the documented references [100]29, [101]30 (Figure [102]1I-[103]1J