Cascade Tf-Mirna-Mrna Regulations Identified By Co-Expression Network Modules In Brains Of Patients With Schizophrenia And Bipolar Disorder

Background Schizophrenia and bipolar disorder are complex mental disorders, with risks contributed by multiple genes. Recently, genome-wide systemic approaches have been used to reveal the associations of hundreds of SNPs with those disorders. Dysregulation of gene expression has been implied, but little is known about such regulation systems in the human brain. Several methods have been developed to achieve measurement of large-scale or multi-level data on interactions in biological systems. Additional biological regulators, such as transcriptional factors (TFs), enhancers, and microRNAs (miRNAs) binding information may help us to unveil the underlying regulatory mechanisms in the networks and suggest causal relationships. Methods We analyzed two data sets using brain tissues from 51 patients with schizophrenia or bipolar disorder and 24 healthy controls. We applied whole genome transcriptome profiling using Affymetrix Human Gene 1.0 ST Array (Affymetrix, Santa Clara, CA) and miRNA sequencing on the Illumina 1G Genome Analyzer. Using weighted gene co-expression network analysis, we integrated gene expression data of mRNA and miRNA from the same brain collections, built mRNA-miRNA co-expression networks and detected differentially expressed modules. In silico predicted binding relationships of TFs and miRNAs were used to validate the putative regulations suggested by co-expression patterns. Using SNP genotype data, expression quantitive trait loci (eQTL) and Network Edge Orienting (NEO) analysis, we resolved causal relationships among the regulators and their targets. We further validated the predicted regulations using RNA interference knockdown experimentally. Results We identified a module differentially expressed between cases and controls (p = 7.6e-4, FDR q = 0.01). This module were enriched for neuron differentiation (p = 4.8e-7, FDR q = 8.5e-4) and neuron development (p = 2.3e-5, FDR q = 4.0e-2), and contained five miRNAs and 501 mRNA genes. Six TFs also served as hub genes in these modules (POU2F1, EPAS1, PAX6, ZNF423, SOX5 and SOX9). The co-expression-suggested regulatory relationships were consistent with the binding relationships predicted by databases. Focusing on those regulations containing TFs and miRNAs, we resolved a regulation cascade from SNP variants (rs16853832) to TF (POU2F1) to miRNA (hsa-mir-320e) to target genes (NR2E1) and ultimately, to disease risks. Discussion We revealed POU2F1 as a key regulator in a neuron differentiation/developmental module associated with disease, and revealed a putative cascade regulation effect. This study showed that we can utilize multi-dimensional data to construct co-expression networks. Causal relationships can be resolved among SNPs, regulatory molecules and their downstream target genes through data integration. Novel genes and their corresponding regulations underlying the disease risks could be revealed.