16 November 2012. Many of the diverse genes emerging for schizophrenia are bound together by a common role in brain development, according to a network analysis published online November 11 in Nature Neuroscience. Led by Dennis Vitkup of Columbia University in New York, the study deciphers biological themes underlying a variety of genes linked to schizophrenia through studies of common and rare variations. Within a background network estimating the phenotypic relationships among all human genes, 47 schizophrenia-related genes formed a cluster related to axon guidance and cell migration that is highly expressed in the brain during prenatal development.
Recent genetic studies of schizophrenia point to a large number of loci—the newest GWAS report identifies more than 60 (see SRF related news story), and a recent study estimates over 850 genes involved (see SRF related news story). Researchers have speculated that these diverse genes converge on related biological pathways, which comprise many different molecules working together to do something for a cell. But testing this idea requires a comprehensive description of all pathways, which biologists are still grappling to understand in single-celled organisms, let alone the human brain (Alberts, 2012). The computational approach designed by Vitkup and colleagues attempts to fill this gap with a network that describes the chance that any two human genes contribute to the same phenotype. In it, the strength of the connection between any two genes is based on multiple sources of information, including gene ontology classifications, functional annotations, evolutionary similarities, and physical interactions with other molecules. Genes that share a similar profile are strongly connected, whereas those with dissimilar profiles would be connected weakly or not at all, thus producing something like an airline flight map showing the varying traffic patterns between cities. Vitkup’s network differs from others built from more limited datasets, which, for example, take into account only molecular interactions.
Vitkup has already used this approach in an effort to describe the roles of genes contained within CNVs found in autism (Gilman et al., 2012). In the new study, he teamed up with Joseph Gogos and Maria Karayiorgou, also of Columbia, to analyze a wider collection of genes associated with schizophrenia through de novo CNVs, de novo protein-altering point mutations detected by exome sequencing, and genomewide association studies (GWAS). Considering all of these genes, rather than merely a subset deemed as more important to schizophrenia, provided an unbiased look at any shared roles among them.
Clusters of connections
First author Sarah Gilman and colleagues analyzed a total of 1,044 genes with links to schizophrenia: 173 came from regions within 250 kb of single nucleotide polymorphisms flagged by GWAS; 712 from genes contained within de novo CNVs; and 159 hit by de novo protein-altering point mutations found by exome sequencing, including data from Gogos and Karayiorgou’s most recent study (see SRF related news story). After mapping these genes to the background network, the researchers then asked whether they formed groups of well-connected genes, termed clusters. To do this, they first scored each possible cluster based on the connection strengths between the genes within the cluster, and plucked out the highest-scoring ones.
This process identified one cluster composed of 47 putative schizophrenia genes, which were enriched for genes with roles in brain development and intracellular signaling. As a control, the researchers used random genes with the same average connection strengths as the schizophrenia-related genes, and found these formed more fragmented clusters within the background network. A second, marginally significant schizophrenia gene cluster was also identified (p = 0.071), which contained genes related to chromosomal remodeling. Clustering worked best when combining information from all sources of genetic data: when the input consisted of only the 159 genes hit by point mutations, a smaller, marginally significant cluster emerged. This suggests that clues coming from studies of common and rare variations in schizophrenia mutually reinforce each other.
Because even healthy people carry rare variants, the researchers also tested whether the genes impacted by de novo point mutations or CNVs in controls or in unaffected siblings of people with autism (which gives a measure of background de novo mutation; see SRF related news story) formed clusters. They did not, which argues that the clusters emerging from the schizophrenia-derived input have something to do with the disorder.
Same pathways, different disorders
Further characterizing their two clusters using the Human Brain Transcriptome database (see SRF related news story), the researchers found that these cluster genes had high levels of expression in the brain, particularly during prenatal development. Also, the functional categories given to their cluster genes overlapped with those ascribed to differentially expressed genes in neurons grown from pluripotent stem cells derived from people with schizophrenia (see SRF related news story), which offers further validation for the cluster genes.
The researchers then grappled with the emerging evidence for genetic overlaps between schizophrenia and other disorders. When considering gene sets already implicated in autism or intellectual disability, they found these were significantly connected with the schizophrenia cluster genes; in contrast, genes hit by de novo mutations in healthy controls, or by synonymous mutations (which do not alter protein structure) in schizophrenia were not. This suggests that genetic glitches involved in schizophrenia, autism, and intellectual disability involve similar biological pathways.
But hitting the same pathway does not necessarily result in the same consequences, and the researchers suggest that the clinical phenotype may depend on the types of mutation involved. In their previous study of autism, the researchers found that cluster genes derived from CNVs were likely to increase the growth of dendrites, the specialized structures on the receiving end of synapses (Gilman et al., 2011). Using a similar analysis, the researchers reported that, of the schizophrenia cluster genes derived from CNV data, those with known phenotypes tended to decrease dendritic growth when hit by a CNV in someone with schizophrenia. Although the story is likely to be more complicated, the analysis reminds us that there are different ways to break a biological pathway, and the details in how they are perturbed may explain different outcomes.—Michele Solis.
Gilman SR, Chang J, Xu B, Bawa TS, Gogos, JA, Karayiorgou M, Vitkup D. Diverse types of genetic variation converge on functional gene networks involved in schizophrenia. Nat Neurosci. 2012 Nov 11. Abstract