Heterogeneity of radiation response in mesenchymal subtype glioblastoma: molecular profiling and reactive oxygen species generation

Abstract

Background: Radiotherapy-induced tumor death remains critical in the successful first-line management of glioblastoma, whereas resistance to radiation serves as a major factor in disease progression. Mesenchymal shift has been identified as a driver in GBM recurrence, with gene expression associated with enhanced repair of macromolecular damage caused by radiation.

Methods: Using distinct mesenchymal subtype GBM cells lines, radiation response was assessed by clonogenic assay and orthotopic mouse tumor model. RNA-sequencing was performed in the setting of increasing radiation dosing while real-time assessment of ROS generation was achieved by the measurement of hydroxyl spin trap adducts via electron paramagnetic resonance.

Results: Radiation-induced cell death determined by clonogenic assay was significantly different at low dose (4-8 Gy) between the resistant U3035 cells and the sensitive U3020 cells. Similar trends were present in the in vivo NSG mouse model following radiation dosing on post-implantation day 7-10, with the rate of reduction in tumor bioluminescence reversing between the U3020 and U3035 cells after the third dose of radiation. Changes in gene expression following radiation determined by RNA-sequencing indicate both U3035 and U3020 cells demonstrate a shift toward more mesenchymal profiles, with concurrent shift away from pro-neural subtype gene expression in the U3020 cells that appeared to develop resistance to radiation in vivo. Persistence of ROS generated following radiation was greater in U3020 cells shown to be more sensitive to radiation.

Conclusions: Despite the same molecular classification, distinct GBM cell lines can demonstrate differential response to radiation and potential for mesenchymal shift associated with radiation resistance.

Keywords: Glioblastoma; Mesenchymal shift; PCA; Radiation; Reactive oxygen species.

Fig. 1

Clonogenic survival differences among mesenchymal GBM (ms-GBM) cell lines. Using the traditional approach to clonogenic survival, sub-confluent cultures of U3035 and U3020 cells were treated with single fractions of 4 Gy, 8 Gy, or 20 Gy, immediately dissociated, and plated at 1000 cells/well. Cultures were maintained for 10 days. Staining/fixation was performed using 0.5% ethidium bromide in 50% ethanol. Colony images were captured under UV light and quantified using ImageJ (b; lower panel). At 4 Gy and 8 Gy doses, U3020 and U3035 cells demonstrated significant differences in surviving fractions (SF) (a; upper panel). Each radiation dose was performed in replicates of four over three independent experiments

Fig. 2

In vivo response to fractionated radiation over 3 weeks following orthotopic implantation. Following successful implantation of U3035-Luc2 or U3020-Luc2 cells, as demonstrated by post-implantation Day 7 (D7) bioluminescence imaging (BLI), tumor-bearing mice were administered 4 Gy whole brain fractions on D7, D10, D14, and D17. Imaging performed immediately prior to each radiation dose was analyzed based on BLI (photons/sec) at 10 min post luciferin substrate (CycLuc1) intraperitoneal injection (bottom). The change in BLI was expressed as the percent change from the previous imaging study for each individual mouse and plotted over time (upper panels). For D7–10 and D10–14 interval, U3020-Luc2 bearing mice exhibited a greater mean decrease in BLI than the U3035 cells, a trend that reverses in the D14–17 and D17–21 intervals. Biological replicates of five were performed in each group

Fig. 3

DMPO-hydroxyl adduct formation and decay following exposure to ionizing radiation. Using fixed number aliquots of dissociated U3020 or U3035 cells in the presence of 200 mM DMPO spin trapping solvent, electron paramagnetic resonance (EPR) analysis was performed. DMPO-OH generation was dose-dependent based on the radiation dose (a) with the signal peaks represented −OH and −H adducts (b). Baseline generation of DMPO-OH as determined immediate after completion of radiation exposure were the same in U3035 and U3020 cells (c), while the decay of the DMPO-OH adducts was significantly greater in the U3035 cells than the U3020 cells (p = 0.017) and DMPO only controls (d). Experiments were performed in triplicate

Fig. 4

Construction of principal components based on functional gene expression sets. Eigenvalues for the mesenchymal and proneural gene sets were determined and graphically represented on both scree plots (left) and a two-dimensional system using the first and second principal components (middle). The top 10 genes from each PC are listed in tabular format (right) with color coding for genes associated with DNA repair processes (green) or free radical reduction (red)

Fig. 5

Principal component analysis (PCA) of RNA-seq data in U3035 and U3020 cells following radiation exposure. Using RNA isolated 6 h post-radiation exposure to sub-confluent cultures of U3035 or U3020 cells, next-generation sequencing was performed. PCA was performed across both cell lines and all radiation doses (0, 4, 8 Gy) using functional gene expression sets for mesenchymal and proneural GBM subtypes. Positive shifts within the mesenchymal gene set was present in the U3020 cells along the first principal component (PC1) and in both cell lines for PC2 for both cell lines (upper left). Within the proneural gene set, negative shifts were present in the U3020 cells in PC1 and PC2, whereas U3035 cells only demonstrated change along PC2 (upper right). The top 10 genes contributing to PC1 and PC2 in each analysis demonstrated seven genes (green) involved in DNA damage repair and two genes (red) related to oxidative stress/free radical mitigation (lower)