Circular RNAs Associated with Idiopathic Parkinson’s Disease: Case/Control Study Results on PPMI Dataset

Abstract Background PD is the second most common neurodegenerative disease after AD, for which biomarkers do not exist. We aimed identify circular RNAs associated with Parkinson’s Disease (PD) in a longitudinal dataset and test their potential as Bioamrkers. Method We included 2507 European ancestry blood samples from PPMI participants, with transcriptomic data available and distributed across five time‐points. We aligned reads to the reference genome (GRCh38) using STAR in chimeric alignment mode. Annotation and quantification of circRNAs were performed using DCC, followed by a cross‐sectional differential expression analyses for each visit using DESeq2. Significance was assessed and corrected by FDR at the last visit, with nominal p‐value for the remaining visits. We used Circular RNA Interactome web tool to predict microRNA (miRNA) target sites within significantly dysregulated circRNAs. Subsequently, we performed pathway analyses using DIANA tools. Result Two circRNAs were significantly dysregulated in PD across all visits: circRHBDD1 (p = 3.34´10 −4 , Log2FC = ‐0.13) and circDHRSX (p = 1.03´10 −3 , Log2FC = 0.4). We identified twenty miRNAs that putatively interact with circRHBDD1. Pathway analyses showed that fourteen of them are involved in the Hippo signaling pathway, fifteen in oxytocin signaling pathway, and sixteen in the ErbB1 signaling pathway. We also identified one hundred and twenty‐eight miRNAs interacting with circDHRSX that are also part of the Hippo signaling pathway. Conclusion We identified circRHBDD1 and circDHRSX to be dysregulated in PD across time. Both of them are enriched in binding sites for miRNAs that target the Hippo and ErbB1 signaling pathways. ErbB1 has already been associated with PD in brain studies, whereas the Hippo pathway is associated to neuronal cell death. We are conducting additional analyses in independent datasets and using mixed models to replicate the results and obtain the best predictive models.