1 June 2011. The brain displays consistent but abnormal patterns of gene expression in autism, according to a study appearing in Nature on May 25. By analyzing the expression profiles of thousands of genes, Daniel Geschwind of the University of California, Los Angeles, and colleagues found that typical differences seen between frontal and temporal cortex in controls were less pronounced in autism. Also, the researchers identified two sets of genes ("modules") that were misregulated in autism: one comprises neuronal genes that include several known autism susceptibility factors, and another contains immune system and glia-related genes.
The study reveals common molecular pathways perturbed in autism—a heterogeneous disorder that, like schizophrenia, seems to stem from a variety of risk factors. "At least two-thirds, maybe up to three-quarters, of the autism cases shared a common molecular pathology," Geschwind told SRF. "That amount of convergence was very surprising."
Detecting these distinctive molecular abnormalities required expansive thinking and a stomach for big matrices. While standard expression profiling looks for individual genes that differ in expression between cases and controls, Geschwind's team wanted to know what the relationships were among the genes themselves—was up- or downregulation of one gene tracked by others? Making pairwise comparisons across thousands of genes in a systematic, unbiased way revealed groups of genes whose expression is highly correlated (the so-called modules), a hallmark of genes whose function is intimately linked.
Geschwind's group has used this kind of "network transcriptome analysis" to identify basic functional units in healthy human brain, including modules related to cell types, synaptic function, and specific organelles (Oldham et al., 2008). The method has also narrowed in on gene expression differences in the brain that distinguish humans from chimpanzees (Oldham et al., 2006). The new study is the first time the method has been used to identify molecular pathology in a brain disorder.
Clusters and modules
First author Irina Voineagu and colleagues started by extracting RNA from postmortem brain samples from 19 people with autism and 17 healthy controls. The samples came from frontal cortex, temporal cortex, and cerebellum—regions associated with autism. Using a microarray to measure the transcript abundance of thousands of genes, they identified 444 genes in the cortical samples whose expression level differed significantly by a factor of at least 1.3 from that observed in control samples. Things looked fairly normal in the cerebellum, however, with only two genes differing between autism and control subjects. When the researchers grouped the expression profiles according to their similarity, they found that the autism cortex samples formed a cluster distinct from that of the controls—a finding validated in a second group of samples. This indicates that autism cortical samples can be consistently distinguished from controls based on their gene expression profiles.
To try to reduce the complexity of the transcriptome to its functional parts, Voineagu and colleagues delineated the sets of genes, or modules, whose expression levels were highly correlated with each other. Overall, 87 percent of the modules extracted from the autism samples resembled those in controls—indicating that the general organization of the brain transcriptome was fairly normal. But the autism modules deviated in other ways. For example, the distinct differences in expression normally observed between frontal and temporal cortex was not apparent for the corresponding module in the autism samples. This suggests that the two regions are less differentiated in autism, a difference that may reflect aberrant development.
Hints from hubs
Two other autism modules stood out from controls: one enriched for genes involved in neurons, synapses, and neurotransmitters was underexpressed in autism, whereas another consisting of genes related to the immune system, inflammatory responses, and glial markers was overexpressed. While these patterns distinguished autism samples from control samples, they could reflect either causal pathology or reaction to it. To get at this question, the researchers turned to association signals uncovered by genomewide association studies (GWAS) of autism, and found that their neuronal module—but not the immune one—was enriched for these signals more than expected by chance, consistent with it containing causal genetic factors.
The neuronal module was also enriched for known autism susceptibility genes, providing independent support for their role in the disorder, according to the authors. Some of these also constitute "hubs" within the module—genes whose expression co-varied the most with other genes in the module, displaying greater connectedness. These hubs identify functionally important genes, and, indeed, one was A2BP1, an alternative splicing regulator previously implicated in autism. Because A2BP1 expression was downregulated in autism, the researchers then used RNA sequencing to look for abnormalities in alternative splicing between three autism samples and three controls. This analysis turned up differential splicing in a number of neuronal genes, including A2BP1 targets, and adds splicing abnormalities to the possible mechanisms contributing to autism.
Standard gene-by-gene expression profiling was first applied to schizophrenia brain samples over 10 years ago (Mirnics et al., 2000), but the kind of network analysis used in the new study may help distill the complexity of expression patterns of thousands of genes into something more tractable. Though findings in schizophrenia postmortem tissue are complicated by confounds like medication history, the strategy of using GWAS signals to parse cause from consequence may resolve this. Ultimately, a systems-level understanding of what distinguishes a brain with schizophrenia from other brains could hasten the discovery of new, more widely effective treatments.—Michele Solis.
Voineagu I, Wang X, Johnston P, Lowe JK, Tian Y, Horvath S, Mill J, Cantor RM, Blencowe BJ, Geschwind DH. Transcriptomic analysis of autistic brain reveals convergent molecular pathology. Nature. 2011 May 25. Abstract