Differential Gene Expression to Characterize Spontaneously Regressing Metastatic Neuroblastoma

Metastatic neuroblastoma (NB) is the leading cause of cancer death in children under the age of 5. Interestingly, infants develop a special pattern of metastasis to skin, liver and bone marrow (stage 4s) and typically undergo spontaneous remission with very low mortality. There is accumulating evidence that patients with intermediate risk group, e.g. infants with bone metastasis, have the same natural history of spontaneous regression and they do not require intensive therapy [1]. This highlights the importance of characterizing this group of patients at molecular level. We searched open repositories for quantitative gene expression data and identified GSE49711 as a suitable data source. This series includes data on 498 patients with NB including extensive phenotypic data and RNA-seq counts generated by Illumina Hiseq 2000 platform. First, we applied exploratory analysis to check the quality of data and the need for normalization before further analysis. Also, patients were categorized into 3 groups: Very low risk (VLR, N= 45), very high risk (VHR, N=92) and others (N= 361). The first 2 groups were further analyzed for differential gene expression (DGE) to identify genes responsible for their contrasting outcome. The edgeR package was used to define the top 100 differentially expressed genes with lowest adjusted p-values. These genes were applied to segregate samples using principal component analysis (PC). Interestingly, PC2 identified a special subset of patients who belong the VLR but have a unique gene expression pattern. Indeed, a small subset of VLR patients (N=6) clustered away from the main group and at extreme PC2 values. When the rotation of PC2 was analyzed, a group of 20 genes were found to have contributed to its values. The selected top 100 genes were then used to build a model of hierarchal clustering and constructing a heatmap. The same subset of samples appeared on a distinct limp of the dendogram with remarkable increased expression in the same exact group of 20 genes mentioned above. Gene ontology (GO) enrichment analysis of these specific genes was performed using an online tool developed by the GO Consortium [2], [3] revealed involvement of the following processes: Acute inflammatory response, regulation of blood coagulation, regulation of hydrolase activity and platelet degranulation. These processes indicate the role of immune system in the process, possibly combined by platelet activation. Interestingly, recent literature is highlighting the role of platelets in initiating inflammatory processes in neural tissue [4]. The model (top 100 genes selected by previous analysis) was used to test the original dataset (N=498) and segregated patients into 3 groups of patients based on outcome. Differential expression of MYCN, NME2(C-Myc transcription factor) and NCAN (neuronal adhesion) were most significantly associated with variable mortality. In conclusion, when 2 extreme groups of NB patients were defined and DGE was applied, a model was developed that showed similarity between the groups with categorization of a small subset of VLR with unique pattern of gene expression. The model worked well in defining risk of mortality and identified certain genes that can serve for prognostication.