The neuregulin-1 (NRG1) gene is widely known as one of the most well-replicated genetic risk factors for schizophrenia. Converging evidence shows that it is associated with schizophrenia at the gene expression and mouse model levels which are consistent with its molecular functions in neural development. However, in several recent genome-wide association studies (GWAS), there appeared nary a blip of association at the 8p12 locus where NRG1 resides. What gives?
While there are many possibilities for this phenomenon (some discussed here), the recent paper, “Support for NRG1 as a Susceptibility Factor for Schizophrenia in a Northern Swedish Isolated Population” by Maaike Alaerts and colleagues, suggest that the typical GWAS study may not adequately probe genetic variation at a fine enough scale – or, if you will, use a netting with sufficiently small holes. By holes, I mean both the physical distance between genetic markers and the frequency with which they occur in populations. While GWAS studies may use upwards of 500,000 markers – that’s a pretty fine scale net for a 3,000,000,000bp genome (about 6,000bp apart) – Alaerts and colleagues set forth with slightly finer-scale netting. They focus on a 157kb region that is about 60kb upstream from the start of the NRG1 gene and construct a net consisting of 37 variants between the markers rs4268087 and rs17601950 (average spacing about 5kb). They used the tagger program to select markers that account for all haplotypes whose frequency is higher than 1.5%. Thus – even though there are still more than 500 possible snps in the region Alaerts and colleagues are exploring, they are using a slightly finer netting than a typical GWAS.
The results of their analysis (using GENEPOP) of 486 patients and 514 ethnically matched control participants from northern Sweden did reveal significant associations in an area slightly downstream (about 50kb closer to the start point of the NRG1 gene) than the location of the “previously often replicated variants”, suggesting that the region does confer some risk for schizophrenia, but, that diagnostic markers for such risk will be different for different populations. More telling however are the very weak effects of the haplotypes that show significant association. Those haplotypes with the most significance show meager differences in how often they are observed in patients vs. controls. For example, one haplotype was observed in 5% of patients vs. 3% of controls. Others examples were, 11 vs. 9, 25 vs. 22 and 40% vs. 35% – revealing the very modest (krill sized) effects that single genetic variants can have in conferring risk toward mental illness.
However, there are potentially lots of krill in the genomic sea!