The primary analyses did not demonstrate associations between schizophrenia and any of the 32 studied DAOA, DAO, PPP3CC, and DTNBP1 SNPs. This finding is in line with other studies performed in larger samples
The secondary analyses instead confirmed effects of the DAO, PPP3CC, and DTNBP1 genes on schizophrenia susceptibility; these effects were mediated not only by the gender and paranoid/non-paranoid dichotomy but also by epistatic/interaction mechanisms.
The discrepancies between the primary and the secondary analyses may be reconciled assuming that the DAO, PPP3CC, and DTNBP1 genes modulate the susceptibility for schizophrenia only under definite circumstances.
Gender-related differences in schizophrenia are far from a novel finding: an extensive literature indicates that males and females differ from each other in some psychopathological domains, treatment efficacy, and the presence of structural and brain abnormalities
[24–26]. The influence of gender on schizophrenia has been related to oestrogens
[25, 26], neurodevelopment
[27, 28], and lateralization
However, it is possible that in schizophrenia, as in many other complex disorders
, several genes or genetic variations may affect susceptibility to the disorder in relation to the gender of the patient both for female or male specific protective or risk effects
Present data concerning an influence of gender not only on the DAO, DTNBP1, and PPP3CC genes but also on the epistatic effect between the DAO and PPP3CC genes undoubtedly fit well within this scenario and have some specific antecedents.
In particular, in a Korean population, the DAO gene evidenced gender-specific differences in relation to the allele distributions and haplotypes of three SNPs, rs2070586, rs2070587, and rs3918347, which acted as risk or protective factors in relation to the sex of the patients
. The Korean study
 and the current study differ in the haplotype combination involved; this discrepancy could be explained, as already reported
, by ethnic background differences in the DAO gene, with variations between Asians and Caucasians in both allele frequencies and linkage patterns (http://hapmap.ncbi.nlm.nih.gov/index.html.en). Furthermore, animal studies have suggested that the DAO activity differs according to sex, most likely due to the effects of sex hormones
[35, 36]. Given the effect of oestradiol in reducing DAO activity in the liver tissues of female guinea pigs
, oestrogens could have a protective effect on schizophrenia by preventing increases of DAO activity and subsequent NMDA hypofunction. In particular, the presence of a hormonally influenced transcriptional regulation of NMDA genes, mainly the DAO gene, could explain our results in females.
Since the significant interaction of gender and DAO and PPP3CC genes, an epistatic effect was performed and a 66% lower risk of developing schizophrenia was observed among females with the PPP3CC CAG triplotype and the DAO GT diplotype carriers as compared with non-CAG or -GT haplotype carriers of the same sex, whereas the CAG triplotype and the GT diplotype alone showed a 47% and a 42% lower risk for schizophrenia, respectively. Therefore, it could be hypothesised not only that the major protective effect of DAO could be mediated by the non-interaction with oestradiol
[35, 36], but also that the potentiation played by PPP3CC may be sustained by some interaction with the sex hormone pathway.
Other relevant findings come from the analysis of DTNBP1 gene polymorphisms and haplotypes: the observation of an influence of DTNBP1 on schizophrenia susceptibility in males could indicate the presence of sex-specific effect and/or of interactions with sex hormones.
With respect to the diagnostic subtype, the 42% and 47% excess of the estimated/phased DAO GT diplotype and PPP3CC CAG triplotype female carriers respectively found in controls compared to patients did not increase when exclusively female paranoid patients were considered (48% and 46% respectively). Therefore, the protective effect of the both DAO GT diplotype and PPP3CC CAG triplotype carriers is reasonably not paranoid-specific but comparable to schizophrenia as a whole. Similar results were obtained for PPP3CC CAG triplotype carriers and paranoid patients versus schizophrenia as whole (see Additional file
1: Table S3B). Although this putative involvement of DAO and PPP3CC in this specific subgroup was observed also for other genes related to the same glutamatergic/NMDAR regulation
[37–40] associated with paranoid schizophrenia, a larger sample is needed to confirm this finding.
The present study requires some further comments. Although our SNPs are substantially informative due to their distribution covering the whole genes, they do not all correspond to those used in other studies. Thus, in the case of the analysis of the same SNP, our association can be attributed to stratification, even though we cannot exclude other confounding clinical phenotypes that manifest differently in male and female schizophrenia patients, such as treatment response, comorbid mood symptoms, and cognitive impairment, rather than a direct link between these genes and female/male schizophrenia. In the case of positive associations with different SNPs in the same gene, we cannot exclude the replica value because haplotype-specific functional variations have not been analysed. A recent GWAS study on bipolar patients showed that there is likely common genetic variation associated with the disorder near exons (±10 kb) that could be identified in larger studies and also provided a framework for assessing the potential for replication when combining results from multiple studies
Moreover the major limitation of this paper is the relatively small sample size mainly after stratification for gender and diagnostic subtypes; however, because of we have based our study on the phenotype dissection strategy, we have computed the power for higher OR. This represents a suggestion for future replications in larger samples.
Finally, there is a formal possibility of type I error due to the limited number of samples available for analyses after different stratifications. However mainly in the case of stratifications by sex, significant results persist even after correction for multiple testing in the subgroups, and this makes it reasonable to assume that there could be only a low possibility of type I error in the results.