Genomic view of bipolar disorder revealed by whole genome sequencing in a genetic isolate

Abstract

Bipolar disorder is a common, heritable mental illness characterized by recurrent episodes of mania and depression. Despite considerable effort to elucidate the genetic underpinnings of bipolar disorder, causative genetic risk factors remain elusive. We conducted a comprehensive genomic analysis of bipolar disorder in a large Old Order Amish pedigree. Microsatellite genotypes and high-density SNP-array genotypes of 388 family members were combined with whole genome sequence data for 50 of these subjects, comprising 18 parent-child trios. This study design permitted evaluation of candidate variants within the context of haplotype structure by resolving the phase in sequenced parent-child trios and by imputation of variants into multiple unsequenced siblings. Non-parametric and parametric linkage analysis of the entire pedigree as well as on smaller clusters of families identified several nominally significant linkage peaks, each of which included dozens of predicted deleterious variants. Close inspection of exonic and regulatory variants in genes under the linkage peaks using family-based association tests revealed additional credible candidate genes for functional studies and further replication in population-based cohorts. However, despite the in-depth genomic characterization of this unique, large and multigenerational pedigree from a genetic isolate, there was no convergence of evidence implicating a particular set of risk loci or common pathways. The striking haplotype and locus heterogeneity we observed has profound implications for the design of studies of bipolar and other related disorders.

Figure 1. Amish pedigree sub-structure.

A) Nuclear family graph representation of the entire pedigree. Each node represents a nuclear family and families are connected by parent-child relationships. The pedigree includes families with genealogical information but without DNA available (grey), families with SNP genotype data available (white and red) and families with at least one WGS subject (red). B) Visualization by multidimensional scaling (MDS) of population substructure and distribution of affection status in the extended pedigree based on microsatellite genotypes. Individuals with Major Affective disorder (BPI or BPII) are shown in red, unaffected individuals in blue. Subjects with other mental illness are depicted as having unknown phenotype (black). C) Pair wise kinship coefficients for known family relationships in the Amish pedigree and an independent set of unrelated parents from the Autism Genetics Research Exchange (AGRE) . As reported in _ENREF_34 , expected kinship coefficients are >0.354 for duplicate samples/monozygotic twins, [0.177–0.354] for 1^st degree relatives, [0.0884–0.177] for 2^nd degree relative, [0.0442–0.0884] for 3^rd degree relatives and <0.0442 for unrelated subjects The expected range of kinship coefficients is shown to the left, the observed kinship coefficients to the right.

Figure 2. Comprehensive genetic analysis of bipolar disorder in a genetic isolate.

Pipeline for the integrative analysis of microsatellite, SNP genotype and whole-genome sequence data. Linkage analysis of the ∼2K microsatellite markers is performed to identify genomic regions with evidence for linkage and to identify the nuclear families and specific haplotypes responsible for each linkage signal. Variants segregating on these haplotypes were then identified by phasing and imputation of variants found by whole-genome sequencing. In the final step, family-based association analysis is performed specifically on these variants.

Figure 3. Linkage peak and haplotype analysis at the 4p16.3 locus.

A) Parametric linkage analysis of chromosome 4 under a recessive model with penetrance parameter 0.85. B) Haplotype frequency distribution in affected and unaffected individuals within the three linked families at 4p16.3.

Figure 4. Top linkage results from non-parametric and parametric linkage analysis.

Top linkage LOD scores for the three diagnostic schemes BPI, Narrow and BPS from the analysis of the extended pedigree and defined subpedigrees. Suggestive linkage (LOD>2.5) was observed for regions on A) 2p25.3-p25.1, B) 7q21.11-q31.33, C) 16p13.13-13.12 and D) 18p11.22-q12.1. The subset of nuclear families used in the analysis for each peak is shown in red in the nuclear family graph insets. Linkage on chromosomes 7 and 18 were observed using all 49 nuclear families, the peaks on chromosomes 2 and 16 were observed in subpedigrees NB4 and NB6 respectively.