As is increasingly observed in complex disease association studies, reports on AKT1
and schizophrenia reveal inconsistent replication of a particular allele or allelic combination, yet replication at a broader level, i.e. the SNP or haplotype and ultimately the gene itself (43
). Potential reasons for such diverse findings are sample size, heterogenous phenotypes, marker sets, allele frequencies or LD structure, genetic admixture, multiple testing issues, and indeed true heterogeneity in variants that influence disease.
In this study we sought to replicate association of AKT1
with schizophrenia in the ISHDSF, prompted by the susceptibility haplotype in this sample within the dysbindin (DTNBP1
) gene (21
). Although no single AKT1
marker showed association with disease, the data were consistent with global association of common alleles for selected SNP haplotypes including rs1130214, and a striking under
-transmission of these haplotypes with the rare allele (T) at rs1130214. This contrasts with Emamian et al
) and Schwab et al
) who found this allele in over-transmitted haplotypes, but is similar to Ikeda et al
) in the most significant association being a ‘protective’ (albeit different) haplotype. Our study is not unique – an increasing number of publications are replicating a previously reported disease-marker association with the risk allele(s) reversed. One explanation may be the ‘Flip-Flop Phenomenon’ of Lin et al
), whereby such associations may indeed be confirmations, with the ‘flip-flop’ due to interactive effects, haplotypic heterogeneity or linkage disequilibrium with a causal variant at another locus.
There is still debate over the ‘level of replication’ required to demonstrate true replication of association. Sullivan (46
) has explored the capacity of association studies to produce false positive findings and impact of various definitions of replication. He defines 4 levels of replication, A-D, being highly precise (same SNP, statistical test/phenotype, direction of association) and D very loose (any SNP, test, direction). Our study corresponds to definition B (same SNPs, test, different direction), predicted to give 0–10% false replications. We therefore consider it unlikely that our results represent false positive findings.
We note in our study the slight difference in results between TRANSMIT and PDTPHASE (e.g. and ). This is due to differences in how several key factors are treated in their algorithms (family structure, frequency of missing parental genotypes, number of informative families and assumed genetic model) creating differences in power. Nicodemus et al
) have evaluated various algorithms commonly used to implement TD-based tests, employing the joint null hypothesis of no linkage and no association, hence their results are applicable to our study. Although overall, PDTPHASE may be considered most rigorous for extended pedigrees, we included TRANSMIT as it was required for the FDR approach to correct for multiple testing, recovers missing parental data efficiently from unaffected sibling data, and with no linkage to 14q32, whereby using multiple families within pedigrees was permitted, TRANSMIT can use the maximal amount of data present in the sample. Both algorithms have strengths for use with the ISHDSF, both broadly concur, thus results of both are presented accordingly.
Many AKT1 haplotypes including rs1130214 appeared to be associated with schizophrenia in the ISHDSF to varying degrees. However, results from our exploratory haplotype analysis may have emerged through multiple testing and reflect false positives. It is unlikely that rs1130214 is an appreciable risk variant itself as single marker association does not reach significance. That it has functional relevance is suggested by bioinformatic analysis, which indicates a transcriptional regulatory role. Our data is consistent overall with prior Caucasian studies in highlighting AKT1 5’ in schizophrenia susceptibility.
The C allele of rs3730358 (in the T-C
-G haplotype) was associated with schizophrenia (6
) and in the core risk haplotype, rs1130214-rs3730358 (T-C
), with decreased AKT1 protein levels in lymphocyte cell lines (6
). However, lymphocytes from normal Caucasians homozygous for the C
-G portion of the haplotype T-C
-G have higher levels of AKT1 protein, and lowered reponse to apoptotic stimuli, than those harbouring the T-A haplotype (48
). As we found no clear relationship between haplotype and AKT1
transcription in the SBS, further studies are required to elucidate any differential effect in schizophrenic patients (associated with AKT1
or not) versus controls. We did observe reduced expression of total AKT1
RNA from schizophrenic and bipolar brain compared to controls, statistically significant for the schizophrenic cases (p
=0.029), and consistent with the reduced AKT1 protein in schizophrenic brain (6
). We found no significant differences between cases and controls in total RNA expression of AKT2
, reflecting again the protein level (6
), and pinpointing AKT1
especially for a role in schizophrenia.
Familial factors are thought not only to influence liability to psychiatric illness but also specific clinical features (49
). Along with shared environment, a major phenotype modifier is genetic, specific susceptibility alleles/haplotypes conferring liability to particular clinical subtypes (37
). The significant under-transmission in our study of the Emamian et al
risk haplotype across all OPCRIT symptom dimensions may indicate an effect on risk for psychosis unspecific to symptom factors. This may be consistent with the strongest associations being at broader diagnostic categories, which include individuals with mood disorders. Using this approach at other schizophrenia risk loci, associations were identified with 1–2 symptom factors (37
). Prior AKT1
/schizophrenia associations in Caucasian samples (6
) are with core schizophrenia or schizoaffective disorder (our D2 category). However, our results suggest the AKT1
effect to be general in nature, and underscore the possibility that AKT1
plays a role in multiple psychiatric syndromes. Although clear differences exist between schizophrenia and bipolar disorder (52
), neuropharmacological studies implicate dopamine dysregulation in both, and molecular studies suggest the conditions may share predisposing genes (53
In summary, our data is consistent with global association for AKTI
haplotypes including SNP rs1130214 with schizophrenia in the ISHDSF. We note an equally striking under
-transmission of haplotypes comprising the same markers but with the rare allele at rs1130214, in contrast to prior reports. We highlight the 5’UTR of AKT1
in disease association, and emphasise the role of genes, not specific alleles or haplotypes, as ‘units or replication’ in association studies (43
). We conclude that whilst AKT1
may play a minor role in susceptibility to schizophrenia in the ISHDSF, it appears to impact risk for a class of psychiatric syndrome with symptoms of schizophrenia and mood dysregulation.