Identification of transcriptomic signatures associated with glioblastoma recurrence

Understanding the molecular changes in glioblastoma (GBM) as it develops from primary to recurrent disease in patients during the course of standard treatment is an area of critical importance. Although much has been done over the years to examine these responses using in vitro and mouse model systems, we have only recently begun to appreciate the real-world trajectory of responses in human patients at the molecular level. An increasing number of studies are addressing this question using omics approaches on paired samples from primary and recurrent disease. Recent papers from the Glioma Longitudinal Analysis (GLASS) consortium^1–3 and others⁴ have revealed recurrence-associated transcriptional changes with evidence for a transition to a mesenchymal transcriptional state, as well as increased neuronal signaling and invasion upon recurrence, but in a variable fashion. In a recent paper published in the journal Genome Biology, the authors use RNAseq of IDH wild-type GBM patient specimens to identify a subset of genes that is consistently transcriptionally altered at recurrence.⁵ The expression of this gene set is regulated by the polycomb repressor complex 2 (PRC2). Most intriguingly expression can go either up or down in a patient-specific fashion in recurrent tumors compared with their matched primary counterparts.

In GBM, the response to standard of care (surgery followed by concomitant radiation and chemotherapy with temozolomide) is variable and leads to a median survival of only 15–20 months.⁶ Here the authors compared transcriptomes in primary and recurrent (second surgery) tumors in a cohort of 107 IDH wild-type matched GBM specimens, specifically with local recurrence, and identify a novel subset of differentially expressed genes. Data were validated using the GLASS consortium dataset. The authors identified a subset of genes that appear to be reproducibly altered between primary and recurrent GBM but found that this group of genes is either up- or downregulated upon recurrence—a type of binary/oppositional response which is indicative of transcriptional reprogramming in opposite directions. The authors refer to these patients as “Up responders” and “Down responders.” To perform the study, first, the authors identified genes differentially regulated between primary and recurrent tumors. To understand their regulatory mechanisms—and here is the important step—the authors used published ChIPseq datasets to identify potential transcriptional drivers of the observed transcriptional alterations. This uncovered JARID2 (Jumonji and AT-Rich Interacting Domain 2) as the most prominent regulatory factor among the differentially expressed genes. JARID2 is an accessory protein in the multiprotein PRC2 complex, which also comprises EZH2, and regulates chromatin accessibility by promoting the suppressive H3K27me3 modification. Analysis of the JARID2-regulated gene set revealed 2 distinct responses, in which signature genes either increased (60% of patients) or decreased (40% of patients). Further investigation of these responder subtypes showed no major differences in overall or progression-free survival with transcriptional subtype. Also, no connection was identified with MGMT expression, and transcriptional subtypes were mixed. Patients with elevated expression of the JARID2 regulated gene set exhibited an increase in oligodendrocytes and neurons as well as increased lymphoid/NK populations and decreased monocytes as predicted by deconvolution algorithms, as opposed to a more mesenchymal and cell-cycle-driven phenotype in the downregulated patients. To investigate this further, single-cell sequencing data was obtained for 22 paired IDHwt GBMs. This supported the cancer cells as being the main drivers of the changes observed in the bulk dataset. In agreement with the bulk sequencing data, glioma cell differentiation and neuronal signaling indicators including GABA signaling components were identified as one of the key ontologies being increased in patients with higher gene signature expression. Reduction in gene expression was correlated with enrichment of cell-cycle genes on recurrence and tumors became more mesenchymal. Bioinformatic network analysis suggested the deregulated signature genes were also drivers of broader transcriptional alterations. Experiments in GBM patient-derived cell lines showed that they also exhibit a dual response, albeit to a reduced level in comparison to the real-world patient scenario. This observation provides an opportunity to interrogate mechanistically the underpinnings of this dual response in future studies.

This study provides an intriguing insight into trajectories of GBM in patients during tumor evolution, and illustrates the power of exploring this type of data in different ways. The identification of a consistently dysregulated gene set in tumor recurrence is an important observation with therapeutic implications. It provides a unifying principle that could be used to enable rapid development of new approaches. Tumor cells in GBM patients appear to have intrinsic drivers of recurrence that stem from bidirectional deregulation of PRC2 signaling. In patients with a downregulated signature, the adoption of a more mesenchymal and cell cycle is in agreement with other studies⁴ and may drive resistance through known mechanisms.⁷ Whereas, in the group with the elevated expression, there was increased differentiation and connections with normal brain cells via synapses, both of which have been reported to be associated with resistance to standard of care.⁸

It would be very compelling if data showing protein expression alterations via immunohistochemistry matched these bioinformatic findings. It would also be of interest to know if there are any elements of the primary gene signature that could be used to predict the directionality of the resistance response in these patients, or a link to underlying tumor genetics: Can we predict up or down responders based on initial signatures? What are the implications for EZH2 targeting in GBM? Overall, this data suggests that a single pathway driving a group of PRC2 regulated genes may work in different ways to promote resistance to therapy and tumor survival, and provides useful starting points for future intervention.

Conflict of interest statement

None declared.