Glioblastoma (GBM) remains the most devastating primary central nervous system malignancy with a median survival of around 15 months. The past decades of research have not yielded significant advancements in the treatment of GBM. In that same time, a novel class of molecules, long non-coding RNAs (lncRNAs), has been found to play a multitude of roles in cancer and normal biology. The increased accessibility of next generation sequencing technologies and the advent of lncRNA-specific microarrays have facilitated the study of lncRNA etiology. Molecular and computational methods can be applied to predict lncRNA function. LncRNAs can serve as molecular decoys, scaffolds, super-enhancers, or repressors. These molecules can serve as phenotypic switches for GBM cells at the expression and/or epigenetic levels. LncRNAs can affect stemness/differentiation, proliferation, invasion, survival, DNA damage response, and chromatin dynamics. Aberrant expression of these transcripts may facilitate therapy resistance, leading to tumor recurrence. LncRNAs could serve as novel theragnostic or prognostic biomarkers in GBM and other cancers. RNA-based therapeutics may also be employed to target lncRNAs as a novel route of treatment for primary or recurrent GBM. In this review, we explore the roles of lncRNAs in GBM pathophysiology and posit their novel therapeutic potential for GBM.

Key molecular regulators of acquired radiation resistance in recurrent glioblastoma (GBM) are largely unknown, with a dearth of accurate preclinical models. To address this, we generated 8 GBM patient-derived xenograft (PDX) models of acquired radiation therapy–selected (RTS) resistance compared with same-patient, treatment-naive (radiation-sensitive, unselected; RTU) PDXs. These likely unique models mimic the longitudinal evolution of patient recurrent tumors following serial radiation therapy. Indeed, while whole-exome sequencing showed retention of major genomic alterations in the RTS lines, we did detect a chromosome 12q14 amplification that was associated with clinical GBM recurrence in 2 RTS models. A potentially novel bioinformatics pipeline was applied to analyze phenotypic, transcriptomic, and kinomic alterations, which identified long noncoding RNAs (lncRNAs) and targetable, PDX-specific kinases. We observed differential transcriptional enrichment of DNA damage repair pathways in our RTS models, which correlated with several lncRNAs. Global kinomic profiling separated RTU and RTS models, but pairwise analyses indicated that there are multiple molecular routes to acquired radiation resistance. RTS model–specific kinases were identified and targeted with clinically relevant small molecule inhibitors. This cohort of in vivo RTS patient-derived models will enable future preclinical therapeutic testing to help overcome the treatment resistance seen in patients with GBM.

Accurate patient-derived models of cancer are needed for profiling the disease and for testing therapeutics. These models must not only be accurate, but also suitable for high-throughput screening and analysis. Here we compare two derivative cancer models, microtumors and spheroids, to the gold standard model of patient-derived orthotopic xenografts (PDX) in glioblastoma multiforme (GBM). To compare these models, we constructed a custom NanoString panel of 350 genes relevant to GBM biology. This custom assay includes 16 GBM-specific gene signatures including a novel GBM subtyping signature. We profiled 11 GBM-PDX with matched orthotopic cells, derived microtumors, and derived spheroids using the custom NanoString assay. In parallel, these derivative models underwent drug sensitivity screening. We found that expression of certain genes were dependent on the cancer model while others were model-independent. These model-independent genes can be used in profiling tumor-specific biology and in gauging therapeutic response. It remains to be seen whether or not cancer model-specific genes may be directly or indirectly, through changes to tumor microenvironment, manipulated to improve the concordance of in vitro derivative models with in vivo models yielding better prediction of therapeutic response.

Glioblastoma (GBM) is the most common and deadly primary brain tumor in adults. While in vitro patient-derived xenografts (PDX) lines are useful for studying GBM, they often exclude astrocytes and macrophages, which contribute significantly to tumor growth, invasion, and chemoradioresistance. Integrating these cells into tumor models is difficult due to their need for serum, which triggers GBM-PDX lines to lose their stem-like properties. The aim of this study was to develop a serum-free triculture model of GBM-PDX lines, normal human astrocytes (NHAs), and macrophages. Serum-free media alternatives were formulated for NHAs and identified for THP-1 macrophages, then combined with GBM PDX media to establish “PSX,” an experimental maintenance media. Cells were transitioned to serum-free media alternatives and functionally assessed through several parameters unique to each cell type. In addition to assessing GBM “stemness,” a custom 350-gene NanoString chip was used to assess differential gene expression in monocultured PDX cells versus PDX cells exposed to NHAs and macrophages. PSX maintained canonical function in astrocytes and macrophages while preserving the stem-like properties of GBM-PDX cells. Tri-culturing all three cells increased the expression of stemness-associated transcription factors and increased the expression of genes related to stemness and hypoxia in GBM cells. GBM PDX cells exposed to NHAs and macrophages in direct triculture exhibit increases in markers of stemness and hypoxia. These findings suggest that the serum-free triculture model presented herein may better recapitulate the tumoral heterogeneity of GBM in vitro, providing a novel model to utilize in current research.

Long non-coding RNAs (lncRNAs) are a heterogenous population of largely evolutionary preserved transcripts, over 200nt in length which play important roles in human health and disease. These versatile and numerous transcripts can possess RNA:Protein, RNA:RNA, and RNA:DNA binding potential. Some lncRNAs also encode short peptides which can have biological significance. Many, if not a plurality of known lncRNAs, are expressed in the brain, believed to help encode complex information and facilitate various neurologic processes. In fact, lncRNA expression appears to be positively correlated with organism and organ complexity in stark contrast to coding genes. Glioblastoma (GBM) is the most common primary brain tumor with a dismal prognosis. The 5-year survival rate for GBM is around 7% with a median survival of between 14-16mos. GBM standard of care has not changed much in past decades and there is a need for new therapeutic strategies. In this review we focus on six lncRNAs of significance to cancer highlighting their potential applications in GBM treatment. These lncRNAs are MALAT1/NEAT2, HOTAIR, ZFAS1, SOX2-OT, CASC19/PCAT2, and SAMMSON. Some of these lncRNAs are well characterized across numerous cancers, but most are poorly characterized in the context of GBM or glioma. Many lncRNAs, including those highlighted herein, have been implicated in therapy resistance including chemoresistance and radiation resistance. Some of these transcripts can be found in patient serum. lncRNAs then present potential as theragnostic, prognostic, and diagnostic potential for GBM and other cancers. They may also represent therapeutic opportunities directly and as concomitant therapies to address resistance mechanisms.