WCPG 2011—Scanning Sequences in Schizophrenia
22 September 2011. With rare and common variants emerging from studies of psychiatric disease, understanding the function of so many variants will be a challenge. A Monday morning plenary talk at the 2011 World Congress on Psychiatric Genetics, in Washington, D.C., by Nicholas Katsanis of Duke University showed a way for “functionalizing” many different alleles.
Katsanis's example was Bardet-Biedel syndrome and other cilia-disrupting disorders, which, like psychiatric conditions, are marked by heterogeneous phenotypes and many contributing variants. Using zebrafish as a model system to probe human allele function, Katsanis and his colleagues have revealed a complex, but tractable, picture in which variants interact with each other, at times giving rise to more severe phenotypes or even new phenotypes than those resulting from an isolated variant.
Though the ability to model the anatomical anomalies of these disorders in zebrafish would seem to rule out the approach for psychiatric disease, it may help understand the function of genes hit by CNVs associated with psychiatric disorders, and which are often accompanied by physical abnormalities. As an example, Katsanis described new experiments in which he was able to model the mirror phenotypes of macrocephaly and microcephaly found in 16p11.2 deletions and duplications, respectively. These deletions are found in autism, and duplications have been reported schizophrenia.
Turning to the transcriptome
Profiling the transcripts found in the brain is one way to identify risk genes for schizophrenia, and Xiangning Chen of Virginia Commonwealth University, Richmond, described his efforts to directly examine them with RNA-sequencing. Starting with tissue samples from more than 80 postmortem brains, including schizophrenia, bipolar disorder, and control groups, he identified hundreds of differentially expressed genes between these groups. Groups of genes whose expression co-varied constituted “hubs” which were enriched for genes involved in synaptic function, hippocampal formation and cell polarity. The hub structure was similar between schizophrenia and bipolar disorder, which supports the idea that they stem from some of the same risk factors. However, the schizophrenia network was also enriched for genes involved in Wnt signaling and GABAergic function. Testing the differentially expressed genes for association with schizophrenia revealed some nominally significant hits, including RIMS1, which encodes a protein localized to synaptic vesicles.
Chao Chen from the University of Chicago took the same approach using data from microarrays to measure mRNA levels. Differentially expressed genes with correlated transcript levels were organized into hub-like groups, which he called “modules.” Two modules distinguished schizophrenia samples from controls: one that was enriched for genes involved in neural differentiation, in astrocytes, and GABRG1, a subunit of the GABA receptor, and another enriched for genes belonging to the metallothionein family, coding for proteins that bind heavy metals. Similarly, Zhongming Zhao of Vanderbilt University in Nashville, Tennesssee, used RNA-seq to identify 218 differentially expressed genes between schizophrenia and control brain samples. A pathway analysis found these were enriched for genes encoding cell adhesion molecules, which is consistent with neurodevelopmental origins for schizophrenia.
Sifting through sequences
With next-generation sequencers churning out data, researchers are trying to make sense of it all, as described in an afternoon session. Todd Lencz of Zucker Hillside Hospital in New York presented preliminary findings from a project to sequence the exomes of two families multiply affected by schizophrenia. Each family featured two affected brothers and unaffected parents, though in both families there was no DNA from the father. Over 5000 variants were identified within each family, so the researchers whittled them down with a homozygous model of disease in mind. This resulted in 7 candidate variants in the first family and one in the second family. Those predicted to be damaging were all on the X chromosome, and lay within genes expressed in the brain. This raises the possibility that X chromosome variants may contribute to risk of schizophrenia, which could help explain the male bias in the disorder (1.4 males to 1 female). Lencz conceded that their filtering procedure in effect favored these single copy X chromosome variants.
CNVs popped up again in this session, with Menachem Fromer of the Broad Institute of MIT and Harvard presenting a method based on principal components analysis to detect CNVs from exome sequences. Using ~1000 exome sequences from a Swedish population, half with schizophrenia and half controls, Fromer came up with more than 3000 rare CNVs. Some were nominally associated with schizophrenia, and these included CNVs involving CHRNA7, which encodes a subunit of the acetylcholine receptor, a histone gene in the major histocompatibility complex (MHC) region, and genes within 16p12.2. Overall, the CNV burden was significantly higher in schizophrenia than in controls. Fromer reported that the method compared favorably with SNP array-based CNV calling in the same samples, and so the method could maximize findings from sequencing data.
Researchers are also picking and choosing particular regions of interest for resequencing, as described by Pamela Sklar of Mount Sinai School of Medicine. About 1800 samples from the same Swedish population used by Fromer were studied, half of them with schizophrenia and half controls. Regions involving about 500 putative schizophrenia risk genes—1.53 Mb per sample—were chosen based on findings from CNV, GWAS, and exome studies. Sklar reported finding more than 9000 rare variants in all, some 4000 of which were amino-acid changing (“non-synonymous”). Though the schizophrenia cases did not carry an excess burden of rare variants compared to controls, there was some evidence for nominal association with variants within a calcium channel gene, CACNA1S, and within NEK4 and ITIH3, which have both turned up as hits in bipolar GWAS (e.g., Scott et al., 2009).—Michele Solis.