Dissecting the many genetic faces of schizophrenia

Machine learning approaches: from theory to application in schizophrenia

In recent years, machine learning approaches have been successfully applied for analysis of neuroimaging data, to help in the context of disease diagnosis. We provide, in this paper, an overview of recent support vector machine-based methods developed and applied in psychiatric neuroimaging for the investigation of schizophrenia. In particular, we focus on the algorithms implemented by our group, which have been applied to classify subjects affected by schizophrenia and healthy controls, comparing them in terms of accuracy results with other recently published studies. First we give a description of the basic terminology used in pattern recognition and machine learning. Then we separately summarize and explain each study, highlighting the main features that characterize each method. Finally, as an outcome of the comparison of the results obtained applying the described different techniques, conclusions are drawn in order to understand how much automatic classification approaches can be considered a useful tool in understanding the biological underpinnings of schizophrenia. We then conclude by discussing the main implications achievable by the application of these methods into clinical practice.

Canonical correlation analysis for RNA-seq co-expression networks

Digital transcriptome analysis by next-generation sequencing discovers substantial mRNA variants. Variation in gene expression underlies many biological processes and holds a key to unravelling mechanism of common diseases. However, the current methods for construction of co-expression networks using overall gene expression are originally designed for microarray expression data, and they overlook a large number of variations in gene expressions. To use information on exon, genomic positional level and allele-specific expressions, we develop novel component-based methods, single and bivariate canonical correlation analysis, for construction of co-expression networks with RNA-seq data. To evaluate the performance of our methods for co-expression network inference with RNA-seq data, they are applied to lung squamous cell cancer expression data from TCGA database and our bipolar disorder and schizophrenia RNA-seq study. The preliminary results demonstrate that the co-expression networks constructed by canonical correlation analysis and RNA-seq data provide rich genetic and molecular information to gain insight into biological processes and disease mechanism. Our new methods substantially outperform the current statistical methods for co-expression network construction with microarray expression data or RNA-seq data based on overall gene expression levels.

TCF4 (e2-2; ITF2): a schizophrenia-associated gene with pleiotropic effects on human disease

Common SNPs in the transcription factor 4 (TCF4; ITF2, E2-2, SEF-2) gene, which encodes a basic Helix-Loop-Helix (bHLH) transcription factor, are associated with schizophrenia, conferring a small increase in risk. Other common SNPs in the gene are associated with the common eye disorder Fuch's corneal dystrophy, while rare, mostly de novo inactivating mutations cause Pitt-Hopkins syndrome. In this review, we present a systematic bioinformatics and literature review of the genomics, biological function and interactome of TCF4 in the context of schizophrenia. The TCF4 gene is present in all vertebrates, and although protein length varies, there is high conservation of primary sequence, including the DNA binding domain. Humans have a unique leucine-rich nuclear export signal. There are two main isoforms (A and B), as well as complex splicing generating many possible N-terminal amino acid sequences. TCF4 is highly expressed in the brain, where plays a role in neurodevelopment, interacting with class II bHLH transcription factors Math1, HASH1, and neuroD2. The Ca(2+) sensor protein calmodulin interacts with the DNA binding domain of TCF4, inhibiting transcriptional activation. It is also the target of microRNAs, including mir137, which is implicated in schizophrenia. The schizophrenia-associated SNPs are in linkage disequilibrium with common variants within putative DNA regulatory elements, suggesting that regulation of expression may underlie association with schizophrenia. Combined gene co-expression analyses and curated protein-protein interaction data provide a network involving TCF4 and other putative schizophrenia susceptibility genes. These findings suggest new opportunities for understanding the molecular basis of schizophrenia and other mental disorders.

ANK3, CACNA1C and ZNF804A gene variants in bipolar disorders and psychosis subphenotype

OBJECTIVES. The ANK3, CACNA1C and ZNF804A genes have been implicated in both bipolar disorders (BPD) and schizophrenia (SCZ). It has been suggested that BPD with psychosis may be a clinical manifestation of genes overlapping between BPD and SCZ. We therefore tested the association of these genes with BPD in a large family-based sample, and then dissected the phenotype into psychosis present or absent subgroups. METHODS. We genotyped four high interest single nucleotide polymorphisms from ANK3 (rs10994336, rs9804190), CACNA1C (rs1006737), and ZNF804A (rs1344706). Family based association testing (FBAT) was performed on 312 families, and within psychotic (N = 158) and non-psychotic BPD (N = 119) subgroups. RESULTS. In the whole sample, we found a nominal association in ZNF804A (rs1344706, P = 0.046), and a trend in CACNA1C (rs1006737, P = 0.077). In the psychotic BPD subgroup, as hypothesized, stronger signals were observed in ZNF804A (P = 0.019) and CACNA1C (P = 0.017). We found no association in the ANK3 markers, but the rs10994336 variant was nominally associated with non-psychotic BPD (P = 0.046). Exploratory analysis revealed the rs1344706 variant was also implicated in suicide-attempt behaviour (P = 0.038). CONCLUSIONS. These tentative results are consistent with the hypothesis that the subphenotype of BPD with psychosis may represent a clinical manifestation of shared genetic liability between BPD and SCZ.

Advances in multidisciplinary and cross-species approaches to examine the neurobiology of psychiatric disorders

Current approaches to dissect the molecular neurobiology of complex neuropsychiatric disorders such as schizophrenia and major depression have been rightly criticized for failing to provide benefits to patients. Improving the translational potential of our efforts will require the development and refinement of better disease models that consider a wide variety of contributing factors, such as genetic variation, gene-by-environment interactions, endophenotype or intermediate phenotype assessment, cross species analysis, sex differences, and developmental stages. During a targeted expert meeting of the European College of Neuropsychopharmacology (ECNP) in Istanbul, we addressed the opportunities and pitfalls of current translational animal models of psychiatric disorders and agreed on a series of core guidelines and recommendations that we believe will help guiding further research in this area.

Microdeletions of 3q29 confer high risk for schizophrenia

Schizophrenia (SZ) is a severe psychiatric illness that affects approximately 1% of the population and has a strong genetic underpinning. Recently, genome-wide analysis of copy-number variation (CNV) has implicated rare and de novo events as important in SZ. Here, we report a genome-wide analysis of 245 SZ cases and 490 controls, all of Ashkenazi Jewish descent. Because many studies have found an excess burden of large, rare deletions in cases, we limited our analysis to deletions over 500 kb in size. We observed seven large, rare deletions in cases, with 57% of these being de novo. We focused on one 836 kb de novo deletion at chromosome 3q29 that falls within a 1.3-1.6 Mb deletion previously identified in children with intellectual disability (ID) and autism, because increasing evidence suggests an overlap of specific rare copy-number variants (CNVs) between autism and SZ. By combining our data with prior CNV studies of SZ and analysis of the data of the Genetic Association Information Network (GAIN), we identified six 3q29 deletions among 7545 schizophrenic subjects and one among 39,748 controls, resulting in a statistically significant association with SZ (p = 0.02) and an odds ratio estimate of 17 (95% confidence interval: 1.36-1198.4). Moreover, this 3q29 deletion region contains two linkage peaks from prior SZ family studies, and the minimal deletion interval implicates 20 annotated genes, including PAK2 and DLG1, both paralogous to X-linked ID genes and now strong candidates for SZ susceptibility.

Mechanisms underlying synapse-specific clustering of GABA(A) receptors

A principle that arises from a body of previous work is that each presynaptic terminal recognises its postsynaptic partner and that each postsynaptic site recognises the origin of the synaptic bouton innervating it. In response, the presynaptic terminal sequesters the proteins whose interactions result in the dynamic transmitter release pattern and chemical modulation appropriate for that connection. In parallel, the postsynaptic site sequesters, inserts or captures the receptors and postsynaptic density proteins appropriate for that type of synapse. The focus of this review is the selective clustering of GABA(A) receptors (GABA(A)R) at synapses made by each class of inhibitory interneurone. This provides a system in which the mechanisms underlying transynaptic recognition can be explored. There are many synaptic proteins, often with several isoforms created by post-translational modifications. Complex cascades of interactions between these proteins, on either side of the synaptic cleft, are essential for normal function, normal transmitter release and postsynaptic responsiveness. Interactions between presynaptic and postsynaptic proteins that have binding domains in the synaptic cleft are proposed here to result in a local cleft structure that captures and stabilises only the appropriate subtype of GABA(A)Rs, allowing others to drift away from that synapse, either to be captured by another synapse, or internalised.

Copy number and sequence variants implicate APBA2 as an autism candidate gene

We recently reported an autistic proband and affected sibling with maternally inherited microduplications within the 15q13.1 and 15q13.3 regions that contain a total of 4 genes. The amyloid precursor protein-binding protein A2 (APBA2) gene is located within the 15q13.1 duplication and encodes a neuronal adaptor protein essential to synaptic transmission that interacts directly with NRXN1 at the presynaptic membrane. We interpreted this as evidence for a putative role of APBA2 in autism as larger maternal duplications of 15q11-q13 are the most common known cause of autism. We therefore resequenced 512 subjects with autism spectrum disorder (ASD) and 463 controls, and identified 7 novel nonsynonymous coding variants in ASD subjects compared with 4 in controls. Five of the seven variants in the ASD group were predicted to affect protein function, alter residues conserved across 18 species, or both. All of the variants for which parental DNA was available were inherited. We also found two different nonsynonymous variants in two siblings with autism: (1) a paternally inherited heterozygous 6 bp deletion and (2) a maternally inherited heterozygous missense mutation, the latter also found in a single control. These results indicate compound heterozygous mutations of APBA2 in this autism sibship. The co-occurrence of two nonsynonymous mutations in both affected siblings in a single family, each transmitted from a different unaffected parent, suggest a role for APBA2 mutations in rare individuals with ASD.

A large replication study and meta-analysis in European samples provides further support for association of AHI1 markers with schizophrenia

The Abelson helper integration site 1 (AHI1) gene locus on chromosome 6q23 is among a group of candidate loci for schizophrenia susceptibility that were initially identified by linkage followed by linkage disequilibrium mapping, and subsequent replication of the association in an independent sample. Here, we present results of a replication study of AHI1 locus markers, previously implicated in schizophrenia, in a large European sample (in total 3907 affected and 7429 controls). Furthermore, we perform a meta-analysis of the implicated markers in 4496 affected and 18,920 controls. Both the replication study of new samples and the meta-analysis show evidence for significant overrepresentation of all tested alleles in patients compared with controls (meta-analysis; P = 8.2 x 10(-5)-1.7 x 10(-3), common OR = 1.09-1.11). The region contains two genes, AHI1 and C6orf217, and both genes-as well as the neighbouring phosphodiesterase 7B (PDE7B)-may be considered candidates for involvement in the genetic aetiology of schizophrenia.