A Study of the Behavioral Variant of Frontotemporal Dementia and Bipolar Disorder: a Neuroimaging and Epigenetics Integrated Approach

Overview

The aim of this study is to investigate the selective epigenetic modifications and their effect on brain's morphology and functionality in the frontotemporal dementia behavioral variant and bipolar disorder. The open-label, multicentric, interventional case-control study involves the analysis of 3 separate cohorts of patients, partly selected over the course of the past 10 years. More specifically, 80 behavioral variant Frontotemporal Dementia (bvFTD) patients (40, of whom 20 carry G4C2 expansion in the C9orf72 gene, are already available, while 40 will be prospectively recruited), 80 Bipolar Disorder (BD) patients (40, including 20 with early onset and 20 with late onset, are already available, while 40 will be prospectively recruited) and 50 healthy control (HC) subjects (20 of whom are already available from other previously approved studies), will be enrolled in this study. For each participant a blood sample will be collected, processed, and studied in order analyze the expression of miRNA. Every participant will also undergo Nuclear Magnetic Resonance Imaging (NMR), Nuclear Magnetic Resonance Spectroscopy (1H-MRS), and Positron Emission Tomography (PET) and, lastly, a battery of behavioral scales to explore different cognitive domains will be administered to all participants by a team of psychologists and physicians. The overall estimated duration of the study is 36 months.

Following the inclusion and exclusion criteria, the participants will be recruited and subdivided into three groups: * bvFTD patients, followed prospectively for at least 2 years. These include bvFTD carriers of G4C2 expansion in the C9orf72 gene (best model of the two diseases because such expansion is widely associated with the development of psychosis) * BD patients who are part of a cohort of multi affected families (MAF) followed since January 2017 at the Psychiatry Operating Unit of the Policlinico di Milano and who have positive family history of neurodegenerative diseases. * Healthy subjects who underwent neurocognitive tests at baseline that excluded the presence of dementia and psychiatric diseases. The investigators will be collecting participants blood samples, which will be processed and analyzed. More specifically, total exosomes will be isolated from 500 microliters of plasma by ExoQuick precipitation solution (SBI). Isolation and purification of NDEs will be performed by ExoFlow purification kit using biotinylated anti-human CD171 (L1CAM) antibody (clone 5G3, Ebiosciences). Total RNA contained in NDEs will be extracted by Total Exosome RNA and Protein Isolation Kit and miRNA expression analysis by TaqMan OpenArray Human Advanced MicroRNA Panel (Thermo Fisher Scientific). Expression analysis of lncRNAs will be conducted by LincFinder Array and inflammatory and autoimmunity arrays (Qiagen). The participants will also undergo a neuroimaging session, where structural MRI and 1H-MRS sessions will be performed using a 3T MRI scanner available at the Neuroradiology Unit. The 1H-MRS will provide sensitive and reliable assessment of neurochemical changes in specific brain areas. The acquisition voxels will be palced in the dorso- and ventrolateral prefrontal cortex (DLPFC/VLPFC), amygdala, and hippocampus. Finally, an FDG-PET scan will be performed with a Biograph Truepoint 64 PET/TC scanner. T1-weighted and FDG-PET images will be used to explore brain morphological/metabolic differences between the groups. Gray matter and white matter volumes will be estimated locally and compared between groups using voxel-based morphometry. A parallel region of interest comparison (ROIs) will be performed to estimate regional volumes using the Automated Anatomical Labelling (AAL) atlas as a reference, again focusing on the DLPFC/VLPFC, amygdala and hippocampus. Finally, an additional regional analysis based on Freesurfer software will allow the investigators to estimate the cortical thickness, cortical surface area, and cortical gyrification of the Desikan-Killiany atlas regions. All MRI and PET analyses will be performed in the context of a general linear model using a specific software implementation in MATLAB called Statistical Parametric Mapping (SPM). Lastly, neuroimaging data and ncRNA expression profiles will be used as predictors in the ML analysis. Given the small sample size, initially linear models (e.g., Support Vector Machine) will be used. Later, to improve the predictive power of our models, the investigators will apply the ML method called random forest, a more versatile and powerful classification algorithm, and the XGBoost (eXtreme Gradient Boosting) algorithm. To avoid overfitting of the learning process, a "grid" search of model hyperparameters will be performed. A 5-fold cross-validation will be used to validate the results, with a split between training-tests of 80%-20%.