Dr. Dan Geschwind from UCLA spoke at the Illumina Neuroscience Discovery Symposium and also on the final day of the Society for Neuroscience meeting on efforts that his lab has made toward understanding neurodevelopmental disorders, particularly autism spectrum disorder (ASD) using integrative genomics.
Sequencing and linkage studies can give information on genes that can shed light on mechanisms of causality, and there is value in examining single genes for modeling and looking at circuits. However, genetic studies also highlight the extreme heterogeneity of conditions such as ASD, where no matter if rare or common, no major gene variations identified to date account for more than 1% of cases. Genome-wide studies are important to understand not only the gene function in complex disease, but also effects from epigenetics, epistasis, and environmental signals. Then on top of that, placing this information within the genetic and molecular contexts of cell-cell interactions, circuit dynamics, nervous system physiology, and behavior makes the picture even more complex than previously imagined.
Dr. Geschwind presented work primarily intended to answer the following questions:
- Despite various factors contributing to heterogeneity, is/are there shared molecular pathologies in ASD?
- Do ASD risk genes converge on similar risk pathways?
He described an initial series of experiments1 looking at pathway annotation from a small cohort of cortical samples from ASD and normal brains using RNA microarrays. From this, they found genes related to synaptic pathway expression were decreased, while genes associated with inflammatory pathways showed increased expression. Examining the transcriptome reveals convergent molecular pathways, or “hubs” of expression in inflammatory/immune genes and synaptic genes. Similar to an air traffic control map, perturbations that cause changes at O’Hare or JFK have greater effects on their respective flight paths than nodes with fewer connections. Differential expression of inflammatory and synaptic genes in ASD cortex seemed to have shared molecular pathology.
While these hypothesis-free gene expression data were certainly interesting, it was not clear if having ASD causes this pattern, or these changes contribute directly to the ASD phenotype. Imputation from larger GWAS studies showed patterns of enrichment in some genes associated with synaptic dysfunction or microglial inflammation, but Dr. Geschwind wanted to pursue other experiments to understand whether these modules cause disease, or are a downstream result of having disease.
Researchers in the Geschwind lab and others then performed a larger gene expression2 study of 113 matched ASD cases and controls using RNA-Seq, which allowed them to not only look at the expression levels of the same genes, but also to assess the effects of splicing. They were interested in seeing if the pathways involved in a monogenic form of ASD (due to rare, de novo variants) were similar to pathways seen in idiopathic ASD. To their surprise, they found that almost all changes in gene expression changes occurred in the same direction, for the idiopathic and monogenic ASD groups, but that they were of a greater magnitude in the monogenic cases. They also found that splicing level changes are even more strongly shared in both ASD cohorts, including targets of FOX1, NOVA1, and SRRM4. The overlap of gene expression patterns between the ASD forms was a “striking convergence that they did not expect”, and can begin to define a shared molecular pathology.
The next goal was to expand further on the information revealed by the splicing and transcriptional networks to map out the circuitry of causal networks from early developmental stages and compare gene interactions and expression levels identified in the post-mortem adult brain samples, to ultimately, evaluate or confirm early developmental involvement of the ASD risk genes. They analyzed expression data from the BrainSpan project (a project undertaken to look at normal molecular networks from early stages of human development) to determine how and when the risk genes are expressed. They found that prenatal neuro-developmental processes do have convergence, involving both transcriptional and translational co-regulation orchestrated by various transcription factors and the recently-discovered fragile X mental retardation protein (FMRP)3. In conclusion, although further developmental modeling is needed, this approach successfully identified some common molecular ASD pathways and risk genes which converge during human neocortical development.
The systems approach of this study shows the value of integrating data from different sample sources and experimental designs. By co-evaluating hypothesis-free expression data from post-mortem samples, permuting values from GWAS to see if there is an enrichment of risk genes, understanding splicing affects, comparing findings from post-mortem samples to data from normal neurodevelopment, following up with functional mechanistic studies (perhaps mouse, as suggested), and replicating studies to confirm associations, it is the hope that with the help of genomics we may reach the point of prediction, and ultimately improve outcomes for people with autism and other complex diseases.
- Voineagu, I., et al. (2011) Transcriptomic analysis of autistic brain reveals convergent molecular pathology. Nature 474; 380-84.
- Parikshak, N., et al. (2013) Integrative functional genomic analyses implicate specific molecular pathways and circuits in autism. Cell 155(5); 1008-1021.
- Darnell, J.C. & E. Klann. (2013) The translation of translational control by FMRP: therapeutic targets for FXS. Nature Neurosci 16; 153-36.