Pheochromocytoma (PCC) and Paraganglioma (PGL) are tumours arising from chromaffin cells of the neuroendocrine lineage. They can occur within the adrenal medulla in the case of PCC, where as PGL occurs along the sympathetic or parasympathetic ganglia. Although most often presenting as a benign disease, a subset of patients develop aggressive malignant disease. In such cases, long-term survival is dismal and treatment options are limited. SDHB-mutations have displayed association with aggressive disease, however early markers of malignant disease are still lacking.
In this study we performed ARACHNE analysis in combination with gene-based T-SNE dimension reduction on RNA profiling datasets of PCC/PGL in order to identify key transcriptional networks underlying different molecular subtypes of the disease. Interestingly, we identify 3 key dimensions of the disease, consisting of gene-networks regulating epithelial to mesenchymal transition (EMT), xenobiotic metabolism/drug metabolism and finally inflammatory response. These root networks were validated and remained stable across 5 independent PCC/PGL datasets consisting of a total of 538 tumours. Clustering of patients based on these 3 root networks displayed 3 main groups of tumours; (1) those with elevated xenobiotic metabolism, EMT and inflammatory characteristics, (2) those with elevated EMT characteristics alone, and (3) those with low expression of all three of the identified root networks. Group (1) of these classified tumours, displayed a significant association with malignant disease, independent of underlying genetic-mutations including SDHB.
Using stable molecular networks to identify new subgroups of PCC/PGLs with varying prognoses may aid not only diagnosing early malignant disease, but also identifying key transcriptional networks that could serve as potential therapeutic targets. Preliminary network analyses reveal YAP1 as the main central node in regulating the mesenchymal root network in PCC/PGL. Further investigation into the role of YAP1 in functional models could be of biological and clinical interest.