Rna sequencing analyses in bipolar multiplex families point to novel genes and dysregulated pathways

Bipolar disorder (BD) is characterized by fluctuations between mania and depression with a strong genetic component. The genetic architecture of BD comprises multiple loci, including common risk variants of small effect and rare risk variants of higher penetrance. Whole exome or genome sequencing in multiplex families have rapidly accelerated the identification of candidate genes for the disorder while RNA sequencing (RNAseq) studies in BD are still limited. We performed RNAseq analyses in a cohort of multiplex bipolar families aiming to identify candidate genes and pathways involved in the disorder. RNA-seq was performed on total mRNA extracted from lymphoblastoid cell lines of 16 BD and 15 unaffected individuals from eight bipolar multiplex families. We performed differential gene expression analysis (DGE) and weighted gene co-expression network analysis (WGCNA) using DESeq2 and WGCNA software, respectively. For both analyses we accounted for ‘age’, ‘sex’ and ‘pedigree’ as covariates. Differentially expressed genes (DEGs) and co-expressed modules were examined for enriched categories via gene set enrichment analysis (GSEA) and over-representation analysis (ORA), respectively. The DEGs and enriched categories were validated using MAGMA and PoPS software, using as input file the GWAS summary statistics of BD (41,917 BD cases; 371,549 controls). Validation of DEGs using RNAseq data derived from human brain will be an important next step, which is currently underway in an independent cohort from the PsychENCODE dataset (RNAseq of 73 BD; 259 controls). Sixty genes were found to be significant for differential expression between affected and unaffected relatives, the most significant being LINC01237. High correlation (r=0.83) of this set of DEGs was found between the lymphoblastoid cell lines and the average of gene expression across 13 brain tissues; 56/60 DEGs showed basal expression in brain (>0.1 TPM; https://gtexportal.org). Four DEGs, CEBPZ, GNB2, PPDPF and RC3H1, were pinpointed by both MAGMA and PoPS analyses for genetic association to BD. GSEA identified 500 significant enriched categories from 7 curated databases, and MAGMA gene-set analysis validated 68 of them. The top validated categories were related to signalling in the nervous system and ion transmembrane transport. WGCNA built 47 modules and 9 of them were significantly associated with BD. The ORA analysis found 281 enriched GO terms in 5 of the associated modules: 11 categories were validated via MAGMA gene-set analysis, “hypoplasia on corpus callosum” being the most significant category. The long non-coding RNA LINC01237 was previously reported as top differentially methylated locus in BD. We found also another long non-coding RNA, LINC01829, in our DEGs, suggesting that this class of regulatory modulators may be key elements in the pathogenesis of BD. Our validated DEGs – CEBPZ, GNB2, PPDPF and RC3H1 – are genes related to signal transduction, response to stimuli and inflammasome. Their function in the brain and previous genomic evidence make them valid candidate genes for BD. Enriched categories from GSEA and ORA after validation pointed to: i) signalling crucial for brain maturation and maintenance, including Wnt, hedgehog, NOTCH4, and ROBO receptors pathways; ii) ion channel transport, such as calcium channels, of which dysregulation may affect the action potential in neurons; iii) ‘hypoplasia of corpus callosum’, which has been reported to be altered in the largest imaging study in BD.