Abstract Genome-wide association studies (GWAS) of schizophrenia (SCZ) have revealed over 100 risk loci. We investigated whether these SCZ-associated variants regulate gene expression by cell type. Using a fully unsupervised deconvolution method, we calculated gene expression by clusters of estimated cell types (cell-groups, CGs). Five CGs emerged in the dorsolateral prefrontal cortices (DLPFC) of 341 donors with and without SCZ. By mapping expression quantitative trait loci (eQTL) per CG, we partitioned the heritability of SCZ risk in GWAS by CGs. CG-specific expressions and eQTLs were replicated in both a deconvoluted bulk tissue data set with a different method and also in sorted-cell expression data. Further, we characterized CG-specific gene differential expression and cell proportion changes in SCZ brains. We found upper-layer neurons in the DLPFC to be associated with SCZ based on enrichment of SCZ heritability in eQTLs, disease-related transcriptional signatures, and decreased cell proportion. Our study suggests that neurons and related anomalous circuits in the upper layers of the DLPFC may have a major contribution to SCZ risk.

Cell group analysis reveals changes in upper-layer neurons associated with schizophrenia