Regional Heterogeneity in Gene Expression, Regulation, and Coherence in the Frontal Cortex and Hippocampus across Development and Schizophrenia

Abstract

Highlights d Dorsolateral prefrontal cortex and hippocampus gene expression across development d Novel region-specific schizophrenia genetic risk features d Decreased regional functional coherence in schizophrenia d Public brain gene expression and eQTL resource at http:// eqtl.brainseq.org/phase2

Transcriptome-scale spatial gene expression in the human dorsolateral prefrontal cortex

“…Interestingly these same two layers -L2 and L5 -showed preferential enrichment of genes implicated in common variation for SCZD (Pardiñas et al, 2018) , and to a lesser extent, BPD (Stahl et al, 2019) . These results were in contrast to differential expression analyses from postmortem studies of brain tissue from patients with SCZD compared to neurotypical controls (Collado-Torres et al, 2019;Gandal et al, 2018) , which showed increased expression of upper layer genes and decreased expression of deep layer and WM genes. Further, we show that the heritability of schizophrenia is enriched for L2, a finding that implicates intracortical information processing as the focus of genetic risk mechanisms.…”

“…Enrichment analyses of DEGs identified in two large SCZD postmortem brain datasets (Collado-Torres et al, 2019;Gandal et al, 2018) , while highly convergent across studies, showed extensive enrichment across all layers, with increased expression of L1, L2, and L3 genes and decreased expression of WM, L4, L5 and L6 genes in patients compared to controls ( Figure 6 B ). As secondary analyses, we performed heritability partitioning analysis for layer-enriched gene sets, which again identified significant heritability enrichment exclusively for L2 enriched-genes, specifically for SCZD, BPD, and educational attainment ( Table S8 , Method Details: Clinical gene set enrichment analyses).…”

“…Second, we overlaid recent large-scale snRNA-seq data from several cohorts to both confirm our layer-enriched expression signatures and further annotate gene expression-driven clusters to individual cortical layers. The shift from homogenate sequencing studies of brain tissue (Collado-Torres et al, 2019;Fromer et al, 2016;Jaffe et al, 2018) to large-scale snRNA-seq has already begun, with increasing sample sizes and numbers of nuclei (Mathys et al, 2019;Velmeshev et al, 2019) , and will only continue to grow. Our strategy of "spatial registration" using individual gene-level statistics from both layer-specific versus cell type-specific expression profiles from hundreds or thousands of genes is likely more powerful than table-based enrichment analyses using small subsets of previously-defined marker genes.…”

“…We also downloaded and reprocessed RNA-seq data from He et al (He et al, 2017) from SRA accession SRP199498 using our previously-described RNA-seq processing pipeline (Collado-Torres et al, 2019) . These data consisted of homogenate RNA-seq data from 18 serial sections across 4 unique donors.…”

Strategies for cellular deconvolution in human brain RNA sequencing data

“…While recent work has extended the number of cell populations that can be isolated by antibodies to separate neurons into their excitatory and inhibitory subclasses and oligodendrocytes from other glia 12 , there are likely very few additional cell types that are possible to isolate using nuclear antibodies for DNAm samples. Researchers have therefore turned to using cell type-specific RNA microarray and sequencing datasets to adapt these reference-based deconvolution algorithms to homogenate RNA-seq samples 7,[13][14][15][16][17][18][19][20][21][22][23] . The majority of these studies have focused on tissues other than the brain, which can be freshly obtained and dissociated into individual cells for single cell RNA-seq (scRNA-seq) or be sorted into specific cell populations using flow cytometry for cell type-specific expression profiling.…”