Preclinical exploration of the DNA damage response pathway using the interactive neuroblastoma cell line explorer CLEAN

Abstract

Neuroblastoma (NB) is the most common cancer in infancy with an urgent need for more efficient targeted therapies. The development of novel (combinatorial) treatment strategies relies on extensive explorations of signaling perturbations in neuroblastoma cell lines, using RNA-Seq or other high throughput technologies (e.g. phosphoproteomics). This typically requires dedicated bioinformatics support, which is not always available. Additionally, while data from published studies are highly valuable and raw data (e.g. fastq files) are nowadays released in public repositories, data processing is time-consuming and again difficult without bioinformatics support. To facilitate NB research, more user-friendly and immediately accessible platforms are needed to explore newly generated as well as existing high throughput data. To make this possible, we developed an interactive data centralization and visualization web application, called CLEAN (the Cell Line Explorer web Application of Neuroblastoma data; https://ccgg.ugent.be/shiny/clean/). By focusing on the regulation of the DNA damage response, a therapeutic target of major interest in neuroblastoma, we demonstrate how CLEAN can be used to gain novel mechanistic insights and identify putative drug targets in neuroblastoma.

Graphical Abstract

Figure 1.

Overview of data processing and availability in CLEAN. Overview of default data processing in CLEAN (left). Pie chart showing the different experimental perturbations that are currently implemented in CLEAN with precise number of studies indicated between parentheses (top-mid). CLEAN screenshot of the default RNA-Seq tab, indicating the 3 main components: (1) the sidebar takes user input and updates what's in the main panel; (2) a searchable table with detailed differential gene expression or related information and (3) an interactive plot (bottom-right). External data sources that are directly (left) or indirectly (via hyperlinks) (top-right), integrated in CLEAN.

Figure 2.

CLEAN-based identification of previously studied drugs that result in a reduced DNA Damage Response transcriptomic signature in neuroblastoma. CLEAN was used to search for previous studies that resulted in significantly downregulated genes (default settings; i.e. logFC < –1.5 at 5% FDR) that were enriched (at 5% FDR) for a custom DNA Damage Response (DDR) gene set (276 genes, as described by Knijnenburg et al. (34)). Four retrieved studies were then further explored in the RNA-Seq tab. Running score (A) and volcano plot (B) for the DDR gene set in the study of Chen et al, 2019 (35). DDR genes indicated in blue in the volcano plot with labelling of genes that are discussed in the main text. (C) Venn diagram showing the number of unique and intersecting genes between the 4 selected studies. (D) A GSEA using canonical pathways was then performed in the 109 common genes. Bar plots indicate –log₁₀(P_adj) values for a selection of enriched pathways as indicated. All plots were generated based on downloaded data from CLEAN. See Supplementary file 1 for an illustration of this use case.

Figure 3.

Comparison of phosphomotif-based protein kinase enrichments after ATR and ALK inhibition and/or stimulation. Four studies were selected in CLEAN that had available phosphoproteomics data after drug treatment with elimusertib, lorlatinib or the ALK ligand ALKAL2. The preranked GSEA functionality was then used to search for protein kinase enrichments and results were downloaded. (A) Volcano plot showing protein kinase enrichment results after elimusertib treatment in the study of Szydzik et al. (3). Protein kinases that are discussed in the main text are labelled. (B) Heatmap showing normalized enrichment scores (color scale) and P values (size) of the four selected studies as indicated on the bottom. See Supplementary file 2 for an illustration of this use case. NES, normalized enrichment scores.