Genetic imaging. There are two major fields in epilepsy research – functional imaging and genetics. Both fields live parallel lives and hardly ever interact. When they do, the interaction is usually short-lived and full of disappointments, as nothing has really ever worked. However, a grant application due today has led me to a recent publication in the Journal of Neuroscience, which combines imaging and GWAS. And believe it or not, the ion channels are back.
Functional connectivity. Connectivity refers to the strength of both anatomical and functional connections between two regions in the human brain. Connectivity can be both functional and anatomical. Functional connectivity refers to the correlation of synchronized activity between two brain regions as measured by the so-called BOLD response or correlation of EEG activity. Anatomical connectivity, which was used in the study by Chiang and colleagues, is usually assessed by Diffusion Tensor Imaging (DTI). DTI basically looks at the diffusion of water molecules in the CNS and measures the direction of diffusion. In white matter tracts, diffusion is easiest along the existing fibres. Therefore, DTI can be used to determine the direction of nerve fibres and for the visualisation of neural tracts (tractography). DTI gives you a measure called fractional anisotropy (FA), i.e. the asymmetry of the diffusion process at a given point in the brain. This can be interpreted as a measure of fibre density.
A twin study on FA. The authors investigated a large cohort of twins and non-twin siblings with MRI and performed DTI. Based on a statistical twin model, they identified 18 regions, in which FA as the measure fibre density was under strong genetic control. These regions were then candidate regions for a genome-wide association study (GWAS), using FA as the endophenotype, for which association with any of 500K SNPs were tested. The results were then corrected for multiple testing, as usual with results for genome-wide studies. Most of the 18 candidate regions were in the superior corona radiata.
Genes for white matter integrity. The authors identified 24 SNPs that were associated with white matter integrity, i.e. FA as a quantitative endophenotype. While many of the associated genes are not immediately intuitive, two classes of genes stand out. First, the authors find an association with SYN3 and SYT17. The encoded proteins Synapsin 3 and Synaptotagmin 17 are involved in synaptic development and function, highlighting the connection between synaptogenesis and the fibre strength. Even more interesting is the association with SCN3A and CACN1AC, genes coding for ion channels. While ion channels are implicated in a broad range of functions, an association with brain structure is interesting. The authors performed gene network analysis and found an enrichment for genes implicated in cell-cell adhesion.
Dusting off the endophenotype concept. Probably for most people in the field of epilepsy genetics, the term endophenotype sounds a little stale. Initially borrowed from the field of psychiatric genetics, it was frequently used in the early 2000’s to describe potential genetic disease markers in epilepsy. Few of the suggested markers had a lasting impact. And amongst those endophenotypes that were repeatedly used including EEG markers (centrotemporal spikes, photosensitivity), the genetic background was not necessarily easier to understand than in epilepsies. However, with the advent of sophisticated functional imaging studies and the possibility to apply these studies in families or larger groups of patients, the endophenotype concept might undergo a revival. Studies like the work of Chiang and colleagues may help identify some genetic aspects of measurable brain function or microstructure.
A few comments. There were some aspects relating to the paper by Chiang and colleagues that might be critically discussed. First of all, they selected brain regions thought to underlie genetic influence based on twin modelling. As mentioned in earlier posts, heritability through twin studies may not necessarily guarantee strong candidate genes. Therefore, I wonder whether other regions might have stronger assocations and what the best method could be to match connectome and genome. In fact, a method for a voxel-wide GWAS has been suggested. Secondly, I feel that the authors proceed from gene finding to gene network a little too quickly. The GWAS field is burdened with false positive findings. This situation is not necessarily improved by relaxing the selection criteria and by performing gene network analysis. The twin dataset is unique, therefore validation might be impossible. However, the lack of a possibility to validate does not add to the statistical significane. Third, it would have been nice if the authors had provided a measure of effect, providing the readers with an idea of how strong the effects actually were. This might help us to estimate whether imaging phenotypes like this qualify as useful endophenotypes that are tightly linked to causative genes.