14 August 2012. A new study takes a novel systems biology approach to microarray transcriptome profiling of schizophrenia postmortem tissue by looking at correlation patterns among genes, rather than single genes alone. Led by Vahram Haroutunian of New York’s Mount Sinai School of Medicine, and published online August 6 in the Archives of General Psychiatry, the study finds that certain groups of genes whose expression covaries, called modules, are altered in schizophrenia. Three modules in particular—the glutamatergic, GABAergic, and oligodendrocyte ones—may be directly involved in schizophrenia pathogenesis, as suggested by their enrichment for genes fingered by a genomewide association study (GWAS). Though the results are based on samples from aged, chronically ill patients, they highlight a way to define groups of genes that work together without relying on preconceived notions about their function.
Expression profiling studies that use microarrays to assess the schizophrenia transcriptome have been plentiful, and changes in functional gene groups, such as oligodendrocyte and metabolic-associated genes, have been identified (Mirnics et al., 2006). However, microarray studies have been plagued by a lack of agreement about the specific genes that are altered, and by only moderate changes in gene expression that are often not statistically significant after correcting for multiple comparisons.
In the current study, Haroutunian’s group has taken a different approach to the analysis of the schizophrenia transcriptome, using a technique called weighted gene coexpression network analysis (WGCNA) (Langfelder and Horvath, 2008). This approach has previously been used to identify perturbed molecular pathways in autism brain tissue (see SRF related news story) and recently in schizophrenia whole blood (de Jong et al., 2012). With the current study, WGCNA comes to postmortem schizophrenia brain tissue for the first time.
WGCNA looks beyond individual genes to identify higher-order relationships among modules. The most highly connected genes within a module form hubs, and the first principal component, termed the module eigengene, serves as a summary of the expression levels for a given module. The module eigengene and the expression levels of hub genes are then used to determine whether a given module is associated with any one of a number of variables, such as a disease, brain region, or genetic variant.
By looking at networks rather than individual genes, WGCNA offers an unbiased way to identify groups of genes altered in illnesses like schizophrenia. As noted by the authors, “such a data reduction strategy has the potential to alleviate the multiple testing correction uncertainties that are encountered in standard gene-centric methodological approaches for microarray data analysis and, hence, decrease the potential for type I and type II statistical errors.”
Making modules from microarrays
First author Panos Roussos and colleagues examined postmortem tissue from four cortical regions (dorsolateral prefrontal, middle temporal, temporopolar, and anterior cingulate) in 21 schizophrenia subjects and 19 controls. Notably, the subjects included in this study were of an advanced age, an average of 81 years and 74 years for control and schizophrenia subjects, respectively.
Microarrays revealed 611 genes that were differentially expressed in schizophrenia, 432 of which were downregulated in the illness. Upon breaking out the expression data by region, only three—middle temporal area, temporopolar area, and anterior cingulate—were significantly altered in schizophrenia. The lack of gene expression changes in the dorsolateral prefrontal cortex, a frequent site of postmortem schizophrenia studies, is in line with a previous report from Haroutunian’s group (Katsel et al., 2005). However, these data are inconsistent with studies from other labs (Fillman et al., 2012), and the authors note that larger pH differences between diagnostic categories in this brain area may be a potential confound.
The researchers identified five network modules—mitochondrial, microglial, oligodendrocyte, glutamatergic, and GABAergic—whose eigengenes were significantly associated with schizophrenia, showing distinct expression profiles from controls. This provides evidence of convergent molecular abnormalities in the illness, and suggests that multiple aspects of cellular functioning are disturbed. In fact, many genes within these modules are already thought to be players in schizophrenia pathology.
Next, they used GWAS data (Ripke et al., 2011) to separate primary transcriptional changes from those due to secondary factors not directly related to disease pathogenesis, with the idea that the modules enriched with genes associated with schizophrenia risk are likely to represent primary etiological changes. Only three of the modules—oligodendrocyte, GABAergic, and glutamatergic—showed a significant association with the schizophrenia GWAS data, suggesting that networks related to these genes may underlie schizophrenia pathophysiology.
When comparing gene network differences among cortical regions, Roussos and colleagues found that substantially fewer genes were differentially expressed among the four brain areas in schizophrenia subjects relative to controls (82 vs. 681 genes), indicating that the gene expression patterns that normally differentiate these cortical areas in healthy subjects are greatly reduced in schizophrenia.
Given the advanced age of the subjects used in this study, whether these findings extend to earlier stages of schizophrenia remains to be seen. By providing a system-level examination of the transcriptome, however, WGCNA offers a new alternative to traditional single gene-based microarray analysis.—Allison A. Curley.
Roussos P, Katsel P, Davis KL, Siever LJ, Haroutunian V. A System-Level Transcriptomic Analysis of Schizophrenia Using Postmortem Brain Tissue Samples. Arch Gen Psychiatry . 2012 Aug 6:1-11. Abstract