Transcriptome-wide association study of schizophrenia and chromatin activity yields mechanistic disease insights

Abstract

Genome-wide association studies (GWAS) have identified over 100 risk loci for schizophrenia, but the causal mechanisms remain largely unknown. We performed a transcriptome-wide association study (TWAS) integrating expression data from brain, blood, and adipose tissues across 3,693 individuals with schizophrenia GWAS of 79,845 individuals from the Psychiatric Genomics Consortium. We identified 157 genes with a transcriptome-wide significant association, of which 35 did not overlap a known GWAS locus; the largest number involved alternative splicing in brain. 42/157 genes were also associated to specific chromatin phenotypes measured in 121 independent samples (a 4-fold enrichment over background genes). This highthroughput connection of GWAS findings to specific genes, tissues, and regulatory mechanisms is an essential step toward understanding the biology of schizophrenia and moving towards therapeutic interventions.introduction Genome-wide association studies (GWAS) have yielded thousands of robustly associated variants for schizophre-1 nia (SCZ) and many other complex traits, but relatively few of these associations have implicated specific 2 biological mechanisms 1,2 , as GWAS association signals often span many putative target genes, may affect 3 gene expression through regulatory 3 or structural elements 4 , and may affect genes at considerable genomic 4 distances via chromatin looping 5,6 . A growing body of research has demonstrated the enrichment of SCZ 5 GWAS risk variants and heritability within regulatory elements identified through maps of chromatin mod-6 ifications and accessibility 1,7-13 . Since chromatin modifications are themselves under genetic control 6,14-19 , 7 a causal mechanism for SCZ loci could lead from genetic variation to chromatin modifiers to gene expression 8 and finally to disease risk. Indeed, QTLs for chromatin (and other molecular phenotypes) are enriched within 9 GWAS associations, further supporting this hypothesis 6,18,20,21 . 10 In this work, we leveraged large gene expression cohorts from multiple tissues, as well as splice variants in 11 brain, to perform a transcriptome-wide association study (TWAS) 22-24 in a large SCZ GWAS data set 1 12 to identify genes whose expression is associated with SCZ and mediated by genetics. We subsequently 13 performed a TWAS for a diverse set of chromatin phenotypes to identify SCZ susceptibility genes that are 14 also associated with specific regulatory elements. To our knowledge, this is the first TWAS to integrate 15 analysis of gene expression, differential splicing, and chromatin variation, moving beyond top SNPs to 16 implicate SCZ-associated molecular features across the regulatory cascade ( Figure 1A). 17 results TWAS for SCZ identifies new susceptibility genes 18 We analyzed gene expression and genome-wide SNP array data in 3,693 individuals across four expression 19 reference panels spanning three tissues: RNA-seq from the dorsolateral prefrontal cortex (PFC) of 621 20 individuals -including SCZ and bipolar (BIP) cas...

Integrating eQTL data with GWAS summary statistics in pathway‐based analysis with application to schizophrenia

“…Because not all SNPs with non-zero weights (derived from the reference eQTL data set) were presented in the GWAS summary data, we used the ImpG-Summary software (Pasaniuc et al, 2014) to impute missing z-scores to the 1000 Genomes Project reference panel accordingly. Because the correlations among can be approximated by LD among the SNPs (Gusev et al, 2016b;Kwak & Pan, 2016), we used the 1000 Genomes Project reference panel (European ancestry) (or other panels for other ethnic/racial groups) to estimate the LD and thus the correlation matrix for . In this study, we used five sets of gene expression reference weights that were based on the following four eQTL data sets: microarray gene expression data measured in peripheral blood from 1,245 unrelated subjects from the Netherlands Twin Registry (NTR), microarray expression array data measured in blood from 1,264 individuals from the Young Finns Study (YFS), RNA-seq measured in adipose tissue from 563 individuals from the Metabolic Syndrome in Men study (METSIM), and RNA-seq measured in the dorsolateral prefrontal cortex from 621 individuals from CommonMind Consortium (CMC) (Gusev et al, 2016b).…”

“…In this study, we used five sets of gene expression reference weights that were based on the following four eQTL data sets: microarray gene expression data measured in peripheral blood from 1,245 unrelated subjects from the Netherlands Twin Registry (NTR), microarray expression array data measured in blood from 1,264 individuals from the Young Finns Study (YFS), RNA-seq measured in adipose tissue from 563 individuals from the Metabolic Syndrome in Men study (METSIM), and RNA-seq measured in the dorsolateral prefrontal cortex from 621 individuals from CommonMind Consortium (CMC) (Gusev et al, 2016b). The weights for differentially spliced introns were further constructed by analyzing CMC data (CMC-introns) (Gusev et al, 2016b).…”

Transcriptome‐wide association study identifies multiple genes and pathways associated with pancreatic cancer

“…TWAS analysis uses pre-computed gene expression weights together with disease GWAS summary statistics to calculate the association of every gene to disease. 14 The genetic values of expression were computed one probe set at a time using SNP genotyping data located 500 kb on either sides of the gene boundary. For this study, the gene expression weights of whole blood, peripheral blood, adipose, and pancreas were driven from the FUSION website (https://gusevlab.org/projects/fusion/).…”

“…For this study, the gene expression weights of whole blood, peripheral blood, adipose, and pancreas were driven from the FUSION website (https://gusevlab.org/projects/fusion/). 14 The genes with significant and suggestive association signals were identified at P value <3.73 × 10 -6 after strict Bonferroni correcting and P value <0.05, respectively.…”

The genetic architecture of psychiatric disorders

“…That is, the same SNPs could independently lead to expression changes in one gene and via a different route have an effect on the phenotype. The summary statistics based Mendelian randomization (SMR) method [106] [108].…”