Functionally-instructed modifiers of response to ATR inhibition in experimental glioma

Abstract

Background: The DNA damage response (DDR) is a physiological network preventing malignant transformation, e.g. by halting cell cycle progression upon DNA damage detection and promoting DNA repair. Glioblastoma are incurable primary tumors of the nervous system and DDR dysregulation contributes to acquired treatment resistance. Therefore, DDR targeting is a promising therapeutic anti-glioma strategy. Here, we investigated Ataxia telangiectasia and Rad3 related (ATR) inhibition (ATRi) and functionally-instructed combination therapies involving ATRi in experimental glioma.

Methods: We used acute cytotoxicity to identify treatment efficacy as well as RNAseq and DigiWest protein profiling to characterize ATRi-induced modulations within the molecular network in glioma cells. Genome-wide CRISPR/Cas9 functional genomic screens and subsequent validation with functionally-instructed compounds and selected shRNA-based silencing were employed to discover and investigate molecular targets modifying response to ATRi in glioma cell lines in vitro, in primary cultures ex vivo and in zebrafish and murine models in vivo.

Results: ATRi monotherapy displays anti-glioma efficacy in vitro and ex vivo and modulates the molecular network. We discovered molecular targets by genome-wide CRISPR/Cas9 loss-of-function and activation screens that enhance therapeutic ATRi effects. We validated selected druggable targets by a customized drug library and functional assays in vitro, ex vivo and in vivo.

Conclusion: In conclusion, our study leads to the identification of novel combination therapies involving ATRi that could inform future preclinical studies and early phase clinical trials.

Keywords: Combination therapies; DNA damage response pathway; DigiWest; Functional genomics.

Fig. 1

Transcriptomic and proteomic profiling before and after ATRi treatment. a, Venn diagram of upregulated differentially expressed genes (DEGs) in LN229 (n = 3) and LNZ308 (n = 3) cells treated with AZD6738 for 72 h. 341 upregulated genes are identified to overlap in both cell lines upon treatment. Lower panel depicts KEGG pathway analysis of identified overlapping genes. Red dashed line indicates significance level of p < 0.05. b, Based on the likelihood ratio test (LRT), genes identified to be differentially affected upon ATRi treatment in between cell lines are analyzed for KEGG pathway affiliation. p53 signaling is strongly upregulated in LN229 cells, PI3K-Akt signaling is downregulated in LN229 and upregulated in LNZ308 cells. Red dashed line indicates significance level of p < 0.05. c, DigiWest protein profiling heatmap depicting treatment- specific effects across both cell lines confirming target engagement (pATR), apoptosis induction (cleaved PARP), NFκB activation (NF-κB p100) and cell-cycle regulation (Chk2). Statistical analysis of significance for heatmap using Wilcoxon test (non-parametric, p < 0.05), for bar graphs Mann-Whitney test (non-parametric, rank comparison, p < 0.05. DMSO vs. AZD6738, LN229 n = 2, LNZ308 n = 2). d, Left, heatmap depicting indicated analytes separated by cell line. Right, bar graph depicting analytes differentially regulated in both cell lines upon treatment. In line with transcriptomic data, p53 is upregulated in LN229 cells while pAkt is downregulated in LN229 cells and trends towards upregulation in LNZ308 cells upon treatment

Fig. 2

CRISPR-Cas9 genome-wide knockout and activation screens identify novel combination partners for ATRi. a, b Upper half, CRISPR-Cas9 screen analyses using a knockout (Brunello) library in LN229 (a) and LNZ308 cells (b) using 750 nM of AZD6738. Left, 9-square plot of MAGeCK MLE results comparing Brunello sgRNA distributions from DMSO or AZD6738-treated cells to the plasmid library pool. Right, rankview plot illustrating MAGeCK MLE results comparing Brunello sgRNA distributions from AZD6738-treated cells to the corresponding DMSO control. c, d Lower half, CRISPR-Cas9 screen analyses using an activation (Calabrese) library in LN229 (c) or LNZ308 cells (d), respectively, using 1.5 µM AZD6738 for LN229 and 1.4 µM AZD6738 for LNZ308 cells. Left, 9-square plot of MAGeCK MLE results comparing Calabrese sgRNA distributions from DMSO or AZD6738-treated cells to the plasmid library pool. Right, rankview plot illustrating MAGeCK MLE results comparing Calabrese sgRNA distributions from AZD6738-treated cells to the corresponding DMSO control. The respective experimental set-up was used to prioritize genetic vulnerabilities (Brunello library) and resistance mechanisms (Calabrese library) upon ATR inhibition

Fig. 3

Functionally-instructed combination therapies in vitro. a, Schematic workflow of the functionally-instructed drug screen. For all selected drugs IC₅₀ values were determined. Then, a pre-selection screen, i.e. 1 × 1 (IC₅₀xIC₅₀) combination of AZD6738 plus drug of interest, was set-up. All combinations with AZD6738 resulting in higher efficacy than additive interaction were then included in 4 × 4 synergy map analyses. b, Analysis of the 1 × 1 pre-selection screen (n = 1 with 8 technical replicates per sample). The heatmap depicts the delta value between prediction of additive drug-drug interaction and measured viability (Bliss synergy score). Positive values (blue) indicate a higher efficacy of the drug combination than predicted, negative values (red) indicate a lower efficacy of the drug combination than predicted. Tideglusib, harmine, everolimus, hydroxyurea, olaparib, temozolomide, vorinostat, cisplatin, etoposide and fludarabine phosphate were selected as top candidates. c, Analysis of the 4 × 4 synergy map experiments. Heatmap depicts the average ZIP synergy score across tested combinations (n = 2). Green coloring indicates high synergism scores, brown coloring negative synergism scores. Hydroxyurea, cisplatin and fludarabine phosphate show positive synergism values across all four cell lines tested

Fig. 4

Combination of AZD6738 with Cisplatin or Fludarabine Phosphate show synergistic efficacy ex vivo and in vivo. a, Evaluation of combination treatment efficacy and synergy of AZD6738 with cisplatin in primary glioma cultures (PC). Shown as black bar is the predicted value for additive combination effects based on the Bliss Independence Criterion as outlined in Methods. Observed measurements (purple) of combination treatments are depicted in purple. Lower values than predicted indicate a synergistic effect of the combination. Shown are means ± SD (n = 1, 3 technical replicates per sample). Representative bar graphs for one AZD6738 treatment in combination with cisplatin in each PC. b, Tumor areas of control (untreated (n = 16), DMSO (n = 16)), AZD6738 [50 µM] (n = 12), cisplatin [150 µM] (n = 19) and the combination of both (n = 12) treated wildtype zebrafish embryos xenotransplanted with LN229-GFP cells. Data was collected in two independent experiments. Tumor surface areas are measured using Imaris (version 9.2.0) after 48 h of treatment. Measurements were normalized to untreated control, means ± SD of the respective groups are indicated, each dot represents one embryo. Statistical analysis using one-way ANOVA (all to all comparison of means), Sidak correction for multiple testing, shown are only comparisons with corrected p-values < 0.5. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001. Right panel, exemplary pictures of zebrafish embryos of each group. Left, untreated zebrafish embryo with anatomical features “eye”, “yolk sac” and “LN229-GFP cells in midbrain region” highlighted by red arrows. Scale bars: 500 μm. c, Waterfall graph (left) and Kaplan-Meier curves (right) of untreated, control treated, AZD6738 (50 mg/kg), cisplatin (1 mg/kg), AZD6738/cisplatin-treated nude mice transplanted with LN229 cells. Median survival for each group listed below waterfall plot. Statistical analysis using log-rank (Mantel Cox) test, p-value below 0.05 considered significant, n.s. “not significant”. Combination of AZD6738 plus cisplatin significantly prolongs survival compared to control treated mice