Genome wide association studies (GWAS) have identified hundreds of loci contributing to bipolar disorder (BD) risk. However, translating genome-wide significant (GWS) loci into causal genes and mechanisms for BD is challenging due to linkage disequilibrium (LD) between risk variants, and incomplete understanding of the non-coding regulatory mechanisms in the brain. Recently, the Psychiatric Genomics Consortium Bipolar Disorder Working Group has performed GWAS meta-analyses of BD in cohorts of European (N cases = 131,969), East Asian (N cases = 5,969), African American (N cases = 7,076) and Latino (N cases = 13,022) ancestries, as well as a multi-ancestry meta-analysis (Total N = 158,036 cases, N= 2,796,499 controls) by including datasets with different ascertainment strategies. These analyses led to the identification of 298 GWS risk loci for BD, further emphasizing the need to identify the true causal variants and elucidate their biological mechanisms at the cellular level.
Here, we implemented SuSiEx, a statistical fine-mapping method leveraging differences in the LD architecture among different genetic ancestries, to prioritize likely causal SNPs, within these 298 GWS risk loci for BD. Then, we mapped these SNPs to their relevant gene(s), and investigated their likely functional consequences by aggregating multiple lines of evidence: (i) integration of variant annotation and brain cell-type epigenomic data (PLAC-seq data), (ii) implementation of Summary data-based Mendelian Randomization (SMR) to functionally interpret the likely causal SNPs in the context of brain bulk tissue quantitative trait loci (QTLs) (expression, splicing and methylation QTLs), and (iii) refining the cell-type specific context of likely causal SNPs via SMR, by leveraging a novel (unpublished) resource of brain single nuclei eQTLs.
Our comprehensive fine-mapping analysis prioritized 113 likely causal SNPs, from 298 GWS loci for BD using LD estimates from all 4 represented populations in the multi-ancestry GWAS. By integrating expression, splicing or methylation QTLs, preliminary results based on a previous BD GWAS indicated that the following genes, among others, are strongly implicated in BD: FURIN, FADS1, DCC, MED24, TTC12, SP4, POU6F2, TRANK1, and DDRD2. Additionally, our preliminary results showed that fine-mapped SNPs for BD can mediate their likely causal effect in specific brain cell-types, specifically inhibitory and excitatory neurons. Taken together, the abovementioned genes represent promising candidates for functional experiments to understand biological mechanisms and therapeutic potential. Finally, we demonstrated that fine-mapping effect sizes can improve performance and transferability of BD polygenic risk scores across ancestrally diverse populations, thus highlighting the potential clinical utility of fine-mapping.