Abstract

In recent years, very large association studies have begun to reveal the contributions of specific common genetic variants in conferring risk for mental illnesses such as autism, schizophrenia and bipolar disorder. However, the size of effect of each individual risk variant has turned out to be surprisingly small (other than for some large structural genetic abnormalities that exist in only very few patients). In the context of very high heritability estimates, which suggest that genetic variation explains a substantial proportion of the variance in risk for psychiatric disease, the lack of many single genetic variants of at least small-to-moderate effect sizes is surprising. This problem has been termed “missing heritability”, and there is some controversy as to its explanation.

An alternative strategy for assessing the influence of genetic variation on risk for psychiatric disease is to consider as the phenotype not a categorical diagnosis, but instead a continuous trait that may be closer to some genetically-influenced process than the diagnosis itself. This strategy of examining “endophenotypes” has yielded some successes in non-psychiatric illness. For example, some of the genes implicated in cardiac arrhythmias were initially identified through an electrocardiogram endophenotype known as the “prolonged QT interval”. Some patients with cardiac arrhythmias were found to have a prolonged QT interval, as were their unaffected first-degree relatives. Ultimately, genetic linkage studies using the prolonged QT interval as an endophenotype successfully identified several risk variants.

The endophenotype approach has not yet been adopted widely in psychiatry. Nonetheless, several candidate endophenotypes have been proposed, including brain imaging measurements. Importantly, several brain imaging measurements conform to a number of the criteria for endophenotypes: they can be heritable; they can differentiate reliably psychiatric patients and non-patients; these differences can occur independent of medication or illness phase; and reliable effects can be identifiable in unaffected first-degree relatives. However, the expense and technical challenge inherent in gathering sufficiently large samples has generally precluded the use of brain imaging methods in large-scale genome-wide association studies to date. Instead, neuroimaging genetics studies have more commonly used the reverse approach, aiming to understand how risk variants for psychiatric disorders might alter brain function in non-psychiatric subjects. While this strategy has yielded some important insights into the mechanisms by which genetic variants affect neural function in healthy individuals, the greatest promise for neuroimaging genetics lies in the identification of new variants that would not have been identified using traditional psychiatric diagnoses as phenotypes.