Genome-wide association study of over 40,000 bipolar disorder cases provides new insights into the underlying biology

Key results

1. 01

   ShownThe meta-analysis detected 64 independent genome-wide significant loci, 33 new for BD; the major histocompatibility complex (MHC) signal, significant for the first time, is led by a variant outside classical HLA and C4 genes, and GWAS inflation mostly reflected polygenicity, not stratification.

   P < 5E-08 in 57 cohorts (effective N = 101,962); 28/30 PGC2 loci replicated; liability SNP-heritability 18.6% (SE=0.008, 2% prevalence); genomic inflation λGC 1.38 but attenuation ratio 0.06 (SE=0.02); 17 loci shared with schizophrenia and seven with major depression GWAS; MHC lead SNP rs13195402 is a BTN2A1 missense variant; imputed C4 alleles were not associated with BD. · GWAS results; MHC locus; Discussion

2. 02

   ShownBD signal concentrated in synaptic signaling gene sets, in genes specifically expressed by excitatory and inhibitory neurons of cortex and hippocampus, and in targets of psycholeptic, calcium channel blocker, antiepileptic and anesthetic drugs.

   MAGMA: 161 of 19,576 genes and 4 synaptic gene sets (of 11,858) significant; enrichment for neuron-specific genes (mouse cortical/subcortical; human hippocampal pyramidal neurons, prefrontal/hippocampal interneurons; 5% FDR); mibefradil and nisoldipine target sets significant among 893 drug sets; four of 63 Anatomical Therapeutic Chemical drug classes enriched. · Enrichment analyses; Enrichment analyses; Discussion

3. 03

   SuggestedIntegrating brain and blood expression quantitative trait locus (eQTL) data prioritized 15 genes whose BD association is consistent with mediation through expression, including druggable HTR6, MCHR1, DCLK3 and FURIN; many are not the nearest gene to the lead variant.

   FUSION TWAS with PsychENCODE weights (1,321 brain samples): 77 genes in 40 regions; FOCUS posterior inclusion probability (PIP) > 0.9 for 22 genes, > 0.7 for 32; 15 of those 32 significant in SMR (PsychENCODE or eQTLGen, 31,684 blood samples) and passing HEIDI; DCLK3 lies 87 kb upstream of top-locus gene TRANK1; HTR6/MCHR1 top eQTLs act oppositely in brain and blood. · eQTL integrative analyses; eQTL integrative analysis; Discussion

4. 04

   ShownBD overlaps extensively with schizophrenia and major depression at the variant level, while its subtypes diverge: BD I shows higher SNP-heritability and closer genetic ties to schizophrenia, BD II to major depression, and their correlation falls short of 1.

   Genetic correlation (rg) 0.68 with schizophrenia (SCZ), 0.44 with major depression (MD); MiXeR: ~8.6 k BD-influencing variants, 97% shared with MD and 96% with SCZ, sign concordance 80% vs 69% (SCZ vs MD); BD I 44 loci, BD II one locus; liability SNP-heritability 20.9% vs 11.6%; BD I-BD II rg 0.85 (P=1.6E-03 vs 1); SCZ rg 0.66 vs 0.54, MD rg 0.34 vs 0.66 (BD I vs BD II). · Genetic architecture of BD and other traits; BD subtypes; Discussion

5. 05

   ShownEuropean polygenic risk scores (PRS), built by leaving each target cohort out of discovery, explained modest liability-scale variance; full meta-analysis PRS explained less in East Asian and admixed African American samples, rising with European discovery size; authors deem PRS not yet clinically useful.

   European weighted mean liability R^2 4.57% (GWAS P threshold < 0.1, 2% prevalence; ~4% for PGC2-based PRS), weighted AUC 65%; top vs middle decile OR 3.5 (95% CI 1.7-7.3), vs lowest decile OR 9.3 (95% CI 1.7-49.3); R^2 up to 2.3% and 1.9% (two East Asian samples) vs 1.2% and 0.4% (two admixed African American samples). · Polygenic risk scoring; Polygenic risk scoring; Discussion; Introduction