This is an excellent article that contains a meta-analysis of the
putative association between G72/G30 with schizophrenia (SZ) and
bipolar disorder
(BP), together with a review of the relevant data from linkage and biology
that place the association findings in context. I think that the
authors are correct to conclude that, while the association data
certainly favor the view that variation at this locus confers risk for
both BP and SZ, a number of nagging concerns remain. The lack of
consistency between the associated alleles in different studies could
reflect true heterogeneity, but there are a number of artifactual causes that need to be excluded.
Second, some of the best associated SNPs actually map some distance
from the gene and in a different LD block to other associated SNPs. If
both these sets of findings are correct, this also suggests that
whatever is going on at the level of the gene is more complex than a
simple association with a single functional variant. Third, no one has
yet succeeded in demonstrating expression of the G72 protein. This
might be because it is expressed only fleetingly or at very low levels.
Alternatively it might be a member of an increasingly recognized class
of genes that encode RNA only. However, it should be recalled that G72
gains its biological plausibility as a candidate gene for SZ and BP
from its putative impact on glutamatergic neurotransmission via
interaction (of the
protein!) with DAO.
In many ways, this is a model of the sort of skeptical and thoughtful
review of candidate genes that the field needs. It makes it clear that
we need larger and more comprehensive studies of this locus, and points
out some of the difficulties in interpretation of the data we already have.
I would caution that G72 has not been shown to be an actual gene, and in the four years since Chumakov and colleagues' report, the biochemistry has not been reproduced.
Reply to comment by Dr Scolnick
We agree that caution is required regarding the assumption that the genetic association at this locus is causally related to the DAOA "gene," and this is the reason that in the paper we have referred to the "DAOA/ G30 locus." Establishing robust genetic association in a restricted region of the genome is clearly the first step on a path to characterizing the biological and phenotypic relationships associated with the variation. It is entirely possible that pathologically relevant variation occurs at the DAOA/G30 locus that does not involve a protein product of the DAOA DNA sequence.
The DAOA/G30 locus is a paradigm of association in psychiatric genetics, where positive reports are followed by both confirmation of association and failures to associate, with the observers of the glass being half-full commenting that it is unlikely that replication would occur spuriously multiple times, and those seeing the glass as half-empty (or three-quarters empty) emphasizing allelic inconsistencies, lack of identified causative SNPs, and in the case of DAOA/G30, lack of conclusive evidence of a gene expressed in brain. Clearly, we are just scratching the surface of understanding the reasons for any association signal in this region of the genome. It is important to remember that the DAOA/G30 locus was cloned from a region that has shown linkage in a number of studies, giving prior probability to association analyses, and that association has been reported in samples from a number of corners of the world. Expression may be restricted to discrete times in development and may not be present in abundance in middle-aged brains. It is also possible, as noted by Mike Owen, that the association signal reflects variation that impacts on a gene or genetic network not yet fully characterized.
This study by Craddock and colleagues makes the case for variation in the gene being related to nondiagnostic aspects of psychopathology, consistent with the reasonable expectation that genes for mental illness will not respect DSM-IV boundaries. Nevertheless, the confirmation of a role for this locus in the pathophysiology of psychiatric disorders will not be based on statistics but on evidence that genetic variation impacts on the biology of brain functions related to the psychopathology in question. We recently reported evidence that the SNP in the 2002 report by Chumakov et al., 2002 that showed association in both of their clinical samples—M10—is associated with cognitive function in a large family sample and with physiologic activation of the medial temporal lobe measured with fMRI even in normal subjects (Goldberg et al., 2006). These associations were not found for SNPs that were negative in the Chumakov et al. study, and the pattern of association with risk alleles was in the direction of abnormalities associated with schizophrenia and with pharmacological NMDA antagonism. In our view, there is dense smoke in the DAOA locus, though fire has yet to be conclusively observed.
I recommend the Primary Papers