ADDRESSING THE IMPACT OF RARE GENETIC VARIANTS IN EXTENDED/MULTIPLEX BIPOLAR FAMILIES

Background Bipolar Disorder (BD) is a complex psychiatric condition with high heritability, the genetic architecture of which likely comprises both common variants of small effect and rare variants of higher penetrance. Multiplex families with high density of illness provide an opportunity to map novel risk genes or consolidate evidence for existing candidates, by identifying genes carrying pathogenic rare variants. Methods We performed Whole Exome Sequencing (WES) in 15 BD families (117 subjects: 72 affected, 28 unaffected relatives), augmented with Copy Number Variant (CNV) microarray data, to examine contributions of: i) predicted pathogenic SNVs and likely-gene-disruptive variants shared in affected versus unaffected relatives; ii) genome-wide burden of likely-gene-disruptive variants; iii) de novo variants; iv) rare CNVs; and v) rare highly penetrant variants under family-specific linkage peaks. Linkage analysis and haplotype reconstruction using WES-derived genotypes enabled elimination of false-positive SNVs, CNV inheritance, and candidate gene prioritisation. Gene Ontology and Gene-sets enrichment analyses were performed after correction for WES-covered ORF length and genic constraint using a null distribution of overlap with randomly drawn gene sets matching on these characteristics. Spearman's correlation assessed relationships between genome-wide burden of likely-gene-disruptive variants in brain-expressed genes and age of onset. Results Rare predicted pathogenic variants shared amongst ≥3 affected relatives were over-represented in Postsynaptic Density (PSD) genes (P=0.002, PBonf=0.024), with no enrichment in unaffected relatives. Burden of likely-gene-disruptive variants was no different in affected versus unaffected relatives (P=0.24), but correlated significantly with age of BD onset (P=0.017). The number of de novo variants in affected versus unaffected offspring was no different (P=0.23), although we note a higher overall rate (1.44/individual) than previous reports. We identified novel BD candidate genes, including the X-linked IRS4 with a stop mutation in all 5 affected siblings which mapped within a family-specific linkage peak. IRS4 displays a restricted pattern of expression in the amygdala, and IRS4-/- mice show compromised maternal nurturing behaviours. Further, our study implicates the protocadherin family of genes, which act to mediate neuronal connectivity. We observed heterogeneity within and between families, with alleles of modest effect and reduced penetrance being the most likely genetic model. Discussion Genetic approaches that combine WES, CNV and linkage analyses in extended families is an effective method for detection of potential pathogenic variation, pinpointing genes and pathways that may contribute to the pathophysiology of BD. A high burden of brain-expressed loss-of-function variants may expedite symptom onset in BD individuals, and further investigation of this relationship with other measures of BD severity are warranted.