2 Apr 2015
April 3, 2015. The genetic clues found in the latest genomewide association study (GWAS) of schizophrenia mostly lie in non-coding regions of the genome, which are in charge of regulating gene expression in various ways. This suggests that a large part of the pathophysiology of schizophrenia is embedded in the process by which a gene is turned into a protein—a process that not only controls how much of a protein is made, but also which version of a protein that is made through alternative splicing or other RNA alterations.
To get a handle on how this unfolds in schizophrenia, an afternoon symposium on Sunday, March 29, at the 2015 International Congress on Schizophrenia Research in Colorado Springs presented the efforts of the Lieber Institute Pharma RNA-seq Consortium to measure changes in gene expression in the brain. Introduced by Daniel Weinberger of the Lieber Institute of Brain Development in Baltimore, Maryland, the consortium consists of five companies (Eli Lilly, Roche, Astellas, Pfizer, and Lundbeck) that have joined forces with the Lieber Institute to sequence the transcriptomes of 1,600 postmortem brains. Compared to older microarray methods that could detect only known transcripts, sequencing RNA (referred to as RNA-seq) can detect all transcripts, including novel ones, and so will give a fuller picture of gene expression in the brain in health and in disease. Weinberger said the first stage of the project will focus on brain tissue from dorsolateral prefrontal cortex (DLPFC) of 800 subjects and compare expression between schizophrenia and controls. The second stage will extend to the hippocampus, and the third will consider expression differences found in the caudate.
The resulting datasets will be released to the research community, and the first two stages are expected to finish by the end of 2015, said Joo Heon Shin of the Lieber Institute. Within the RNA-seq data from DLPFC, Shin was able to detect cases of RNA editing when the RNA sequence deviates slightly from the corresponding DNA sequence. This could result in different versions of a protein being made, and he found some indication of a difference in the number of RNA editing sites between schizophrenia and control brains from people aged 40 or older.
If RNA sequencing seemed straightforward to any attendees in the room at this point, they were set straight by Phillip Ebert of Eli Lilly in Indianapolis, Indiana. Ebert bluntly described the quality control challenges in obtaining these kinds of data, which ranged from contamination of RNA sequence data by non-human sources to indicators of an unexpected ethnicity, which flags a sample mix-up. Ebert said that, because studies of postmortem brain tissue tend to rely on a small number of brains, it was imperative to enforce quality control to prevent one poor sample from poisoning an entire dataset.
The transcriptome data can help make sense of the 108 loci found in the Psychiatric Genomic Consortium's acclaimed schizophrenia GWAS (see SRF related news report). Andrew Jaffe of the Lieber Institute described his hunt for expression quantitative trait loci (eQTLs)—the single nucleotide polymorphisms (SNPs) associated with gene expression changes in the DLPFC from healthy controls. Testing eight million SNPs, he found over 9,000 genes that had an SNP significantly associated with their expression levels. That is, for a significant SNP, one version of the allele was associated with, say, lower levels of a gene's expression while the other version tracked with higher expression. These SNPs overlapped with risk SNPs identified by the PGC's GWAS, and so link them to a function; they also suggest that non-coding SNPs regulate a vast network of genes that is disrupted in schizophrenia.
Jens Wendland of Pfizer in Cambridge, Massachusetts, outlined the company's plan for finding targets amid the 108 loci (Schubert et al., 2014). To help with this endeavor, he noted a need for more RNA-seq datasets, more fMRI data focused on intermediate phenotypes, and more rare variants, which are easier to interpret functionally. He also worried that, because the PGC's GWAS results flag loci that make a person vulnerable to schizophrenia, drugs based on these data may not help much in decreasing symptom severity in a person who is already ill, but instead may be more relevant for people at illness onset.
Finally, Daniel Weinberger took the podium again as the discussant to reveal a tantalizing example of how RNA-seq data can help find genes hiding in GWAS signals. Noting a genomewide-significant hit in the 10q24 region containing 10 genes, he showed how consulting the brain transcript data could clarify things. In this case, the risk-associated SNPs in the region were linked to expression changes of one gene, AS3MT. The full transcript for AS3MT did not change with SNP genotype, however, but a splice variant did. This splice variant was overly abundant in schizophrenia compared to controls and other psychiatric disorders, giving a clear direction for how to normalize this gene's expression.—Michele Solis.