Abstract
To get beyond the "low-hanging fruits" so far identified by genome-wide association (GWA) studies, new methods must be developed in order to discover the numerous remaining genes that estimates of heritability indicate should be contributing to complex human phenotypes, such as obesity. Here we describe a novel integrative method for complex disease gene identification utilizing both genome-wide transcript profiling of adipose tissue samples and consequent analysis of genome-wide association data generated in large SNP scans. We infer causality of genes with obesity by employing a unique set of monozygotic twin pairs discordant for BMI (n = 13 pairs, age 24-28 years, 15.4 kg mean weight difference) and contrast the transcript profiles with those from a larger sample of non-related adult individuals (N = 77). Using this approach, we were able to identify 27 genes with possibly causal roles in determining the degree of human adiposity. Testing for association of SNP variants in these 27 genes in the population samples of the large ENGAGE consortium (N = 21,000) revealed a significant deviation of P-values from the expected (P = 4x10⁻⁴). A total of 13 genes contained SNPs nominally associated with BMI. The top finding was blood coagulation factor F13A1 identified as a novel obesity gene also replicated in a second GWA set of approximately 2,000 individuals. This study presents a new approach to utilizing gene expression studies for informing choice of candidate genes for complex human phenotypes, such as obesity.
Introduction: Addressing the "Missing Heritability"
Obesity is a highly heritable trait, with estimates ranging from 45% to 85%. However, early Genome-Wide Association Studies (GWAS) struggled to explain more than 1% of this variation. Most obesity-related genes reside in a statistical "gray zone"—their individual effects are too small to pass the stringent significance thresholds required for whole-genome scans.
A further complication in transcriptomic studies of adipose tissue is the "chicken or egg" problem: are observed differences in gene expression the cause of obesity, or a reaction to the metabolic stress of an overweight state? This study, published in PLoS Genetics, addresses these challenges by integrating transcriptomics with large-scale genetic data to filter out "reactive" noise and identify genuine drivers of adiposity.
Methodology: An Integrative "Subtraction" Approach
The researchers employed a multi-step strategy to prioritize candidate genes before mining large-scale GWAS data.
- The Reactive Filter: Adipose tissue biopsies were analyzed from 13 pairs of monozygotic (identical) twins discordant for BMI. Since these twins are genetically identical, any expression differences must be a reaction to environmental factors or obesity status, not a genetic cause.
- The Correlation Scan: Expression was also analyzed in 77 unrelated individuals, where differences represent a mix of reactive and causal factors.
- Subtraction Logic: By subtracting the "reactive" genes identified in twins from the "correlated" genes found in unrelated individuals, the team isolated 27 genes putatively involved in causal processes.
- Targeted GWAS Mining: The researchers then "mined" existing GWAS data from the ENGAGE consortium (N=21,000) specifically for these 27 genes. This targeted approach reduced the multiple-testing burden, allowing them to detect signals that would otherwise be lost in the noise.
Key Results and Insights
The study successfully identified a specific set of genes that correlate with BMI but are not merely reactive to the obese state.
- F13A1 Validation: The top hit was SNP rs2274393 in the F13A1 (Coagulation Factor XIII A1) gene, which showed a significant association with BMI (P = 0.003). In a meta-analysis including a replication cohort of 2,000 individuals, the association reached a p-value of 0.0004.
- The Coagulation Link: While F13A1 is primarily known for its role in blood clotting, its association with BMI suggests a direct genetic link between obesity and cardiovascular risks like thrombosis.
- Gender Specificity: The study also noted sex-specific genetic architectures, such as variants in the URB gene associating with BMI specifically in women.
Discussion: Beyond the Low-Hanging Fruit
By narrowing the search from millions of SNPs to variants within 27 biologically prioritized genes, the authors demonstrated a powerful method for uncovering the "missing heritability" of complex traits. While the initial discovery sample was small, the biological value of the twin-control model provided the necessary leverage to find significant signals in the GWAS "gray zone."
Future research will focus on functional "wet lab" studies to determine the exact molecular mechanisms by which factors like F13A1 influence adipose tissue expansion. This integrative methodology offers a blueprint for studying other complex phenotypes where distinguishing between cause and effect is paramount.
Summary updated