Along with their comprehensive review in the target article, Crespi & Badcock (C&B) present an intriguing, if rather speculative, new hypothesis which attempts to reconcile imprinted gene dysfunction with abnormal brain development and vulnerability to neuropsychiatric disorder. The central tenet of the article is that psychosis and autism represent phenotypes at opposite ends of a neurodevelopmental spectrum, and that the disruption of paternally expressed imprinted genes may predispose to one disorder, whereas the disruption of maternally expressed imprinted genes may predispose to the “opposite” disorder. C&B's detailed discussion represents an extension of ideas relating to the role that imprinted genes play in social behaviour, which has been alluded to by others previously (Brown & Consedine Reference Brown and Consedine2004; Haig Reference Haig, LeCroy and Moller2000a; Isles & Holland Reference Isles and Holland2005; Isles et al. Reference Isles, Davies and Wilkinson2006). Imprinted genes (i.e., genes that are epigenetically marked such that they are monoallelically expressed in a parent-of-origin dependent manner) are currently an extremely hot topic in behavioural genetics, given their important role in key neurodevelopmental processes (Wilkinson et al. Reference Wilkinson, Davies and Isles2007). Thus, this review is timely and provides novel testable ideas regarding the role of these fascinating genes in conferring vulnerability to mental disorder.

To assess the utility of the proposals contained within the target article, it is necessary to answer two pertinent questions: First, can we really say that autism and psychosis are diametrical disorders of the social brain? Second, what is the evidence for (oppositely) imprinted gene dysfunction in these disorders?

On the first point, the authors themselves point out that much of their evidence comes from complex, multigenic disorders where psychosis or autistic traits are endophenotypes. The type of psychosis, neuroanatomy, and cognitive function are likely to be extremely heterogeneous both within, and between, each so-called psychotic-spectrum condition, making the grouping of these conditions together as a single entity a potentially hazardous practice. For example, the genetic abnormalities that cause gross developmental abnormalities such as reduced brain size or low birth weight in schizophrenia are likely to be significantly different from those producing a similar outcome in other psychotic-spectrum conditions. Indeed, in some psychotic-spectrum conditions (such as psychotic bipolar disorder), reduced brain size may not be apparent (McDonald et al. Reference McDonald, Marshall, Sham, Bullmore, Schulze, Chapple, Bramon, Filbey, Quraishi, Walshe and Murray2006), making generalisations about psychotic-spectrum disorders nigh on impossible. It is also relevant to consider the specificity and extent of the “opposite” phenotypes of autism spectrum and psychotic-spectrum disorders. The authors have highlighted a number of endophenotypes that differ between the two types of disorder (bearing in mind the caveat outlined above); however, the two types of disorder share many similarities. For example, dysfunction of attentional and inhibitory processes (both fundamental to normal social functioning) is common to many neuropsychiatric disorders ranging from autism to schizophrenia to bipolar disorder.

In answering the second question posed above, it is necessary to appreciate our current lack of knowledge about imprinted gene function in the brain. The evidence that imprinted gene function may be disrupted in autism and psychosis is largely circumstantial, mainly coming from parent-of-origin dependent linkage/association effects or from studies of subjects with cytogenetic abnormalities in imprinted gene-rich regions. In the former case, candidate genes have often not been identified, or when they have been identified, the molecular basis of the parent-of-origin association effect has not been clarified. Work which includes genetic, epigenetic, functional, and association approaches (such as that by Francks et al. Reference Francks, Maegawa, Lauren, Abrahams, Velayos-Baeza, Medland, Colella, Groszer, McAuley, Caffrey, Timmusk, Pruunsild, Koppel, Lind, Matsumoto-Itaba, Nicod, Xiong, Joober, Enard, Krinsky, Nanba, Richardson, Riley, Martin, Strittmatter, Moller, Rujescu, St Clair, Muglia, Roos, Fisher, Wade-Martins, Rouleau, Stein, Karayiorgou, Geschwind, Ragoussis, Kendler, Airaksinen, Oshimura, Delisi and Monaco2007) is likely to improve our knowledge of imprinted gene dysfunction in neuropsychiatric disorder, and will permit us to propose more refined hypotheses. In the latter case, the cytogenetic abnormalities often encompass multiple oppositely imprinted genes, together with non-imprinted bystanders, making drawing conclusions as to the functions of individual imprinted genes complicated.

A number of independent studies have now shown that disruptions of the imprinted gene-rich cluster on human chromosome 15q11-q13 result in mental disorder, including both autism and psychosis. Evidence for this comes from a number of general genomic association studies (Martin & Ledbetter Reference Martin and Ledbetter2007), and also from the explicit imprinted gene syndromes, Angelman syndrome (AS) and Prader-Willi syndrome (PWS), which display aspects of autistic-like behaviour and psychosis, respectively. Indeed, it is the reciprocity of the neuropsychiatric problems observed in AS and PWS that forms the cornerstone of C&B's theory, as these disorders are caused by the absence of maternal gene expression, and paternal gene expression at 15q11-q13, respectively. One important consideration, however, is that the original suggestion of such reciprocity (Boer et al. Reference Boer, Holland, Whittington, Butler, Webb and Clarke2002) also identified that it is an overdosage of maternal gene expression found specifically in maternal uniparental disomy and imprinting centre defect sub-types of PWS that predisposes to psychosis. This finding does not contradict C&B's theory, but it does point to the fact that it is not the lack of paternal gene expression from this cluster per se which results in psychosis. However, these waters are further muddied by the fact that one of the most consistently observed cytogenetic abnormalities in autistic subjects is duplication of the maternally inherited chromosomal region 15q11-q13 (Martin & Ledbetter Reference Martin and Ledbetter2007); this implies that overdosage of maternally expressed imprinted genes in this region may predispose to autistic tendencies. This observation is inconsistent with the simple idea that over-expression of maternal genes in this region leads to psychosis, but nevertheless may help to explain why a significant proportion of PWS subjects demonstrate symptoms reminiscent of autism spectrum disorders (Dimitropoulos & Schultz Reference Dimitropoulos and Schultz2007). Thus, the findings from perhaps the best characterised imprinted disorders only partially support C&B's prediction, in that while the genes in the imprinted cluster on 15q11-q13 are almost certainly involved in social functioning, the functional dissociations between paternally and maternally expressed genes in this region are still unclear.

On a technical note, C&B state that imprinted genes can be dysregulated in more ways than non-imprinted genes; this is incorrect. However, the monoallelic expression of imprinted genes does render them more vulnerable to complete inactivation. Moreover, given that imprinted genes are important players in fundamental neurodevelopmental processes (Wilkinson et al. Reference Wilkinson, Davies and Isles2007), it is likely that the consequences of their dysregulation may be profound. Finally, the complexities of genomic imprinting have necessitated the use of a standardised terminology. For example, it is now common practice to describe imprinted genes on the basis of their expressed allele (i.e., paternally or maternally expressed genes) rather than their silenced allele. As such, ambiguous terms like “paternal-gene bias” need to be clarified to prevent confusion.

In summary, we believe that the ideas presented in the review article are highly original and should be considered when interpreting genetic data from neuropsychiatric conditions. However, the caveats mentioned herein should be borne in mind upon such interpretation.