Proneural and mesenchymal glioma stem cells display major differences in splicing and lncRNA profiles

Abstract

Therapy resistance and recurrence in high-grade gliomas are driven by their populations of glioma stem cells (GSCs). Thus, detailed molecular characterization of GSCs is needed to develop more effective therapies. We conducted a study to identify differences in the splicing profile and expression of long non-coding RNAs in proneural and mesenchymal GSC cell lines. Genes related to cell cycle, DNA repair, cilium assembly, and splicing showed the most differences between GSC subgroups. We also identified genes distinctly associated with survival among patients of mesenchymal or proneural subgroups. We determined that multiple long non-coding RNAs with increased expression in mesenchymal GSCs are associated with poor survival of glioblastoma patients. In summary, our study established critical differences between proneural and mesenchymal GSCs in splicing profiles and expression of long non-coding RNA. These splicing isoforms and lncRNA signatures may contribute to the uniqueness of GSC subgroups, thus contributing to cancer phenotypes and explaining differences in therapeutic responses.

Keywords: Cancer genomics; Cancer stem cells; Gene expression analysis; Gene regulatory networks.

Fig. 1. Mesenchymal and proneural GSCs have distinct alternative splicing profiles, but a similar number of genes harboring alternative splicing events.

a Hierarchical cluster showing that GSCs can be grouped based on their alternative splicing profile (PSI values). Bootstrap (bp) and approximately unbiased (AU) probability from pvclust. b Percentage of events and number of genes harboring exon skipping (ES), mutually exclusive exons (MXE), alternative 3′ or 3′ splice site (ASS), and retained introns (RI) events in MES or PN GSCs. c Quantitative RT-PCR (qRT-PCR) validation of four exon-skipping events in PN and MES GSCs. ELMOD3, FDFT1, OSBPL6, and RAB18 are official gene names. d An exon- skipping event (gene OSBPL6; second exon) in PN and MES GSCs. e qRT-PCR validation of three exon-skipping events in MES and PN GSCs. GSN, SLC9A5, and WASF3 are official gene names. d An exon-skipping event (gene GSN; second exon) in MES and PN GSCs.

Fig. 2. Proneural and mesenchymal GSCs share most of their AS events with GBM samples.

a Percentage of alternative splicing events from GSCs in GBM samples. GSC-MES & GBM-MES contains AS events found in GSC MES and GBM (molecular subtype) mesenchymal. GSC-PN & GBM-PN contains AS events found in GSC PN and GBM (molecular subtype) proneural. GSC-MES & GBM and GSC-PN & GBM contains AS events found in GSC-MES and GSC-PN and GBM samples (all molecular subtypes), respectively. GSC-MES and GSC-PN contain AS events not found in GBM samples. b Intersection between differentially spliced and differentially expressed genes (up- and downregulated) in the two GSC subtypes.

Fig. 3. Differentially expressed genes in mesenchymal or proneural GSCs are related to GBM patient prognosis of patients with GBM.

a Protein-coding genes with higher expression in proneural GSCs (showed in blue) or mesenchymal GSCs (showed in red) are exclusively associated with better versus poor prognosis in GBM patients with tumors of molecular subtypes proneural or mesenchymal subtypes of GBM, respectively. b Genes associated with survival in GBM patients are listed by type of alternative splicing event detected in the GSCs: ES, exon skipping; IR, intron retention; MXE, mutually exclusive exons; ASS, alternative splice sites 3′ or 5′. c Survival curves of four genes (presented in a) exclusively affecting patients with GBM proneural or mesenchymal GBM. Significant log-rank p-values and multivariate Cox p-values are shown in blue.

Fig. 4. Proneural and mesenchymal GSCs have genes with alternative splicing and differential expression related to key biological processes in tumorigenesis and sets of highly connected networks of genes.

a Gene Ontology (Biological Process) enrichment analysis of genes harboring alternative splicing events and genes differentially expressed genes in proneural versus mesenchymal types of GSCs. The networks based on protein-protein interactions display genes associated with b DNA repair; b cell cycle; c cilium assembly; and d mRNA splicing.

Fig. 5. Mesenchymal and proneural GSCs have splicing-related genes with distinct profile of alternative splicing and/or differentially expressed.

a Splicing-related genes differentially expressed and/or harboring alternative splicing events in mesenchymal versus proneural GSCs. Mutated genes are shown in bold. b Differentially expressed splicing-related genes in MES and PN GSCs. Genes reported to harbor mutations are shown in bold. c Number of splicing-related genes in our study that harbor mutations according to studies by Seiler and colleagues and Jayasinghe and colleagues.

Fig. 6. RBPs co-expressed in mesenchymal and proneural GBM samples.

Correlation plots display RBPs positively correlated in (a, b) mesenchymal GBM; and (c, d) proneural GBM (Spearman’s significance: **p-value < 0.01; ***p-value < 0.001). Interaction networks show RBPs functionally related from (e) mesenchymal GBM and (f) proneural GBM. Blue nodes represent selected RBPs, while white nodes were additionally included in the network to give a broader context of associations between the RBPs. Only experimentally validated associations are shown in gray.

Fig. 7. Mesenchymal and proneural GSCs present lncRNAs differentially expressed, harboring alternative splicing events, and associated with prognosis in GBM patients.

a Number of lncRNAs presenting alternative splicing events or differentially expressed (left side) and their intersections (right side) in MES or PN GSCs. b Differentially expressed lncRNAs in GSCs with prognostic value to GBM and other cancer patients (“TCGA tumors” follows TCGA nomenclature for cancer types (e.g., LGG low grade glioma)). PO Proto-oncogenes; TSG tumor-suppressor genes. c Survival curves based on median expression values of lncRNAs in GBM patients. Corresponding log-rank p-values and multivariate Cox p-values are also shown. d Representation of lncRNA GAS5 and its retained introns (in red) in MES GSC. Retained introns host three small nucleolar RNAs: SNORD79, SNORD80, and SNORD57.