Genetics and Brain Morphology

Abstract

A wealth of empirical evidence is accumulating on the genetic mediation of brain structure phenotypes. This comes from twin studies that assess heritability and genetic covariance between traits, candidate gene associations, and genome-wide association studies (GWAS) that can identify specific genetic variants. Here we review the major findings from each of these approaches and consider how they inform on the genetic architecture of brain structure. The findings from twin studies show there is a strong genetic influence (heritability) on brain structure, and overlap of genetic effects (pleiotropy) between structures, and between structure and cognition. However, there is also evidence for genetic specificity, with distinct genetic effects across some brain regions. Candidate gene associations show little convergence; most have been under powered to detect effect sizes of the magnitude now expected. GWAS have identified 19 genetic variants for brain structure, though no replicated associations account for more than 1 % of the variance. Together these studies are revealing new insights into the genetic architecture of brain morphology. As the scope of inquiry broadens, including measures that capture the complexity of the brain, along with larger samples and new analyses, such as genome-wide common trait analysis (GCTA) and polygenic scores, which combine variant effects for a phenotype, as well as whole-genome sequencing, more genetic variants for brain structure will be identified. Increasingly, large-scale multi-site studies will facilitate this next wave of studies, and promise to enhance our understanding of the etiology of variation in brain morphology, as well as brain disorders.

Glossary

Bonferroni correction

simplest method of correcting for multiple testing ($\alpha /{N}_{tests}$). In GWAS, the number of independent tests has been estimated to 10⁶, leading to a genome wide significance threshold of 5 × 10⁻⁸.

Causal variant

a variant that has a direct or indirect functional effect on disease risk. Because of the LD structure of the genome, identifying the causal variant among highly correlated signals is not straightforward. A lower p-value, expression data or a biological understanding of the causal mechanism are only suggestive evidence.

Endophenotypes

endophenotypes are heritable phenotypes; genetically correlated with the disease/complex trait (John and Lewis 1966; Gottesman and Gould 2003).

Epigenetic

molecular process that causes gene expression to change in time through environmental changes in the DNA methylation sites or RNA sequence.

eQTL

(expression Quantitative Trait Loci): Loci that regulate the expression of a gene by regulating the number of micro RNA copies.

(Genetic) Variant

used to refer to observed/tagged SNPs but also to SNPs or any structural variants imputed from the reference panel.

Haplotype

set of SNPs, within one chromosome, that show non random association (linkage disequilibrium). Within one haplotype, SNPs can be highly correlated, which reduces the number of independent testing but creates collinearity issues when simultaneously testing the effect of several SNPs (e.g. gene based tests).

Imputation

the statistical method consisting in inferring a missing value. In genetics, it uses the LD structure of a fully sequenced reference panel to predict the unobserved SNPs (up to 3 millions) based on the set of tagged SNPs (usually 500,000 from a SNP chip).

Linkage Disequilibrium (LD)

non-random association of alleles across a population genome. The LD structure of a genome, defines blocks of strongly correlated alleles called haplotypes.

Intergenic region

DNA region located between genes. Intergenic regions are non-coding and tend to regulate nearby gene expression.

Intronic region

DNA region located inside a gene.

Minor Allele Frequency (MAF)

the frequency of the least frequent nucleotide version at one SNP. MAF can be different across ethnic groups.

Meta-analysis

statistical method that allows merging results from different centres without having to share the (raw) individual data. Only summary statistics of the associations (size effect, number of individuals, p-values) are shared and combined. Studies where the raw data is shared are usually called mega-analyses.

Mode of inheritance

The manner in which a particular genetic trait or disorder is passed from one generation to the next.

(DNA) Methylation

epigenetic process that sees a methyl group binding to the DNA molecule with consequences on gene expression. In the human adult, methylation only happens on sites where a cytosine is directly followed by a guanine in the DNA sequence (CpG site). However, highly conserved non-CpG methylation has been shown to accumulate in neurons from fetal to early adult age, to compose the main form of methylation in the neuronal genome (Lister et al. 2013).

Missing Heritability

discrepancy between heritability estimated from twin or family studies and heritability calculated from genome-wide significant SNPs identified through GWAS.

Pleiotropy

One gene influences multiple traits

Polygenic Score

Also known as a genomic profile score. Individual scores are calculated from participant’s genotype data by summing the number of effect (risk) alleles weighted by the variant effect size as determined in an independent Discovery sample. The SNP list is limited to variants with a p-value less than a defined threshold (or several thresholds may be considered, e.g. <.00001, .001, .001, .01, .1, etc. (Wray et al. 2014)).

SNP

Single Nucleotide Polymorphism, a variant in the genome where the nucleotide can differ between individuals.

Statistical Power

Probability of detecting a significant association, in presence of true association i.e. 1- probability of false negative. In GWAS, the power is specific to a SNP and depends on the level of statistical significance (α), the SNP effect size and MAF, the sample size.