Common genetic variants and gene expression associated with white matter microstructure in the human brain

Highlights

• Five genome-wide significant SNPs were detected in association with FA.

• HTR7 expression is associated with FA and with intronic genome-wide significant hit.

• GNA13 and CCDC91 expression may mediate effects of genome-wide significant SNPs.

• Additive genetic effects on FA are largely homogeneous across white matter tracts.

Abstract

Identifying genes that contribute to white matter microstructure should provide insights into the neurobiological processes that regulate white matter development, plasticity and pathology. We detected five significant SNPs using genome-wide association analysis on a global measure of fractional anisotropy in 776 individuals from large extended pedigrees. Genetic correlations and genome-wide association results indicated that the genetic signal was largely homogeneous across white matter regions. Using RNA transcripts derived from lymphocytes in the same individuals, we identified two genes (GNA13 and CCDC91) that are likely to be cis-regulated by top SNPs, and whose expression levels were also genetically correlated with fractional anisotropy. A transcript of HTR7 was phenotypically associated with FA, and was associated with an intronic genome-wide significant SNP. These results encourage further research in the mechanisms by which GNA13, HTR7 and CCDC91 influence brain structure, and emphasize a role for g-protein signaling in the development and maintenance of white matter microstructure in health and disease.

Introduction

Studies in twins and extended pedigrees have established that white matter microstructure, as measured in vivo by diffusion tensor imaging (DTI), is heritable (Chiang et al., 2011b, Jahanshad et al., 2013a, Kochunov et al., 2010). However, the genetic variants contributing to this heritability are unknown and little is understood about the mechanisms that govern the development, maintenance, plasticity and pathology of white matter microstructure. White matter plays an important role in several neurological diseases (Stebbins and Murphy, 2009) and psychiatric disorders (Kubicki et al., 2007, Mahon et al., 2010), which are phenotypes that also have substantial but poorly characterized genetic components. There is increasing evidence that compromised white matter microstructure is part of the inherited risk for these disorders, as indicated by reduced FA in unaffected relatives (Gold et al., 2012, Hoptman et al., 2008, Sprooten et al., 2011a, Sprooten et al., 2013a), and polygenic risk score analysis (Whalley et al., 2013). Therefore, identifying genes that influence white-matter microstructure could provide a biological anchor for disentangling basic molecular mechanisms that predispose to these debilitating disorders, potentially leading to novel treatment agents and prevention strategies.

DTI is a magnetic resonance imaging technique that is based on the orientation and magnitude of the motion of water molecules, and its restriction by surrounding tissue. Because of the parallel alignment of white matter fibers that restrict motion primarily in directions perpendicular to the fibers, DTI is ideally suited to measure properties of white matter microstructure (Beaulieu, 2002). Fractional anisotropy (FA) is an index of the extent to which this motion is directionally constrained and, as validated in animal (Li et al., 2011) and post-mortem research (Schmierer et al., 2007), it reflects a combination of myelin thickness, fiber coherence and axon integrity. Studies using a priori selected candidate genes and SNPs have associated FA with genetic variation in NRG1 (McIntosh et al., 2008, Sprooten et al., 2009, Winterer et al., 2008), ErbB4 (Konrad et al., 2009, Zuliani et al., 2011), DISC1 (Sprooten et al., 2011b), NTRK1 (Braskie et al., 2012), BDNF (Chiang et al., 2011a) and APOE (Jahanshad et al., 2012), amongst others. However, FA is a complex, polygenic phenotype and for most complex phenotypes data-driven GWA have not implicated a priori candidate variants in their top results (Flint and Munafo, 2013, Stein et al., 2012), hence many more novel SNP-associations contributing to variation in FA could be discovered using GWA.

Numerous common variants correlated with complex disease risks have been reported using GWA (Hindorff et al., 2009, Hirschhorn and Daly, 2005, Ripke et al., 2013), but the effect size of individual common variants on complex phenotypes tend to be small (Flint and Munafo, 2013, Hindorff et al., 2009). Significant genome-wide association reflects the presence of a relevant functional variant in the surrounding genomic region and thus is indicative of causal gene localization but not of the identification of an underlying biological mechanism, which is the ultimate goal of complex disease genetics. It is difficult to infer a specific gene's involvement in trait variance solely based upon a statistically significant association, since the polymorphisms tagged in GWA rarely influence gene function directly and the effect of a tagging SNP reflects, in addition to its own effect, the effects of all SNPs within the surrounding linkage disequilibrium (LD) block, which may span many genes any of which could be driving the observed association. Examining complementary biological information, such as RNA expression, can refine inferences made from GWA and identify potential genes through which the associated SNPs are likely to exert their effect.

In the present paper, we aim to characterize the common variation contributing to the genetics of white matter microstructure. Firstly, to identify common variants affecting white matter microstructure we performed GWA of a global FA measure in a sample of 776 Mexican-American members of extended pedigrees. Secondly, to propose genes that may be responsible for the effects of these common variants, we correlated lymphocyte-derived RNA transcripts of nearby genes both with genetic variation in the genome-wide significant SNPs and with white matter microstructure.

We chose a global index of white matter microstructure, namely average FA across the white matter skeleton derived from tract-based spatial statistics (TBSS), as our primary phenotype. This phenotype was previously shown to be heritable in multiple cohorts (Jahanshad et al., 2013a), and relevant to genetic risk for bipolar disorder (Sprooten et al., 2011a, Sprooten et al., 2013a) and major depressive disorder (Whalley et al., 2013). To examine the neuroanatomical specificity of genetic effects on FA, we performed voxel-wise analyses for our top SNPs, and used the family structure and additional GWA results of regional FA to investigate the degree to which additive genetic effects across the genome are anatomically homogeneous throughout the brain.