Genetic hypotheses for schizophrenia

In their ‘common disease – rare alleles’ hypothesis McClellan et al (2007) come close to formulating an untestable theory. Although they refer to currently fashionable ‘candidate genes’ – e.g. dysbindin, neuregulin and DISC1 – it appears that they do not regard these as established. I agree that there is no consistency in the findings across even the largest genome scans conducted to date (sample sizes exceeding 300 and totally over 1000 sib-pairs; Crow, 2007) but disagree profoundly about the alternative.

We know that in schizophrenia: (a) incidences are more uniform across populations as one moves to the core syndrome, e.g. nuclear symptoms (Jablensky et al, 1992); (b) structural brain changes (e.g. ventricular enlargement) are consistent across populations (Chua et al, 2003) and uniform across patients relative to controls (Vita et al, 2000); (c) age at onset has a specific distribution throughout the reproductive phase of life; (d) there are gender differences (earlier onset and worse outcome in males); (e) the core syndrome comprises symptoms that are language related (i.e. specific to Homo sapiens). None of these findings would be expected if schizophrenia were a result of random mutations in a large number of genes such as McClellan et al postulate, nor would one expect variation in the form of illness within families as is generally observed.

While McClellan et al's hypothesis promises a search for elusive rare alleles that will never reach a conclusion, Craddock et al (2007) perseverate in their claim that ‘Several genes have been implicated repeatedly as conferring risk for schizophrenia or bipolar disorder’. Comparison of the largest and most systematic linkage studies, including those of Craddock et al themselves, shows that these claims cannot be sustained (Crow, 2007).

Alternative hypotheses to the ‘rare alleles of major effect’ and the ‘polygenes of small effect’ deserve consideration. One such hypothesis (Crow, 2007) is that the variation arises in relation to characteristics that are specifically human, i.e. recent in evolution, and that it is ‘epigenetic’ in form, i.e. involves a modification of the sequence (methylation of DNA) or the associated chromosomal structure (methylation, phosphorylation or acetylation of histones) rather than a change in the DNA sequence itself. We do know that the risk for first-degree relatives is approximately 10%, whereas that for second-degree relatives is very little increased compared with the population as a whole. Although this is often held to be consistent with polygenic influence, it is also compatible with an ‘imprint’ that is applied and reapplied in meiosis (i.e. with a short time course between generations). The solution proposed is that the variation arises in relation to the change (speciation event) that defined the species, and that this is associated with the cerebral torque – the bias from right frontal to left occipital across the anteroposterior axis that is characteristic of the human brain. In contrast to McClellan et al's rare alleles and Craddock et al's polygenes of small effect, this hypothesis is specific and refutable – a gene has been identified that duplicated at 6 million years from the X to the Y chromosome to give rise to the ProtocadherinX/Y gene pair. This pair has been subject to accelerated evolution since the duplication event (Williams et al, 2006) and is in an unusual situation with respect to epigenetic modulation. This variation can be assessed and the hypothesis thereby tested.