Identifying tissues implicated in Anorexia Nervosa using Transcriptomic Imputation

Abstract

Anorexia nervosa (AN) is a complex and serious eating disorder, occurring in ~1% of individuals. Despite having the highest mortality rate of any psychiatric disorder, little is known about the aetiology of AN, and few effective treatments exist.Global efforts to collect large sample sizes of individuals with AN have been highly successful, and a recent study consequently identified the first genome-wide significant locus involved in AN. This result, coupled with other recent studies and epidemiological evidence, suggest that previous characterizations of AN as a purely psychiatric disorder are over-simplified. Rather, both neurological and metabolic pathways may also be involved.In order to elucidate more of the system-specific aetiology of AN, we applied transcriptomic imputation methods to 3,495 cases and 10,982 controls, collected by the Eating Disorders Working Group of the Psychiatric Genomics Consortium (PGC-ED). Transcriptomic Imputation (TI) methods approaches use machine-learning methods to impute tissue-specific gene expression from large genotype data using curated eQTL reference panels. These offer an exciting opportunity to compare gene associations across neurological and metabolic tissues. Here, we applied CommonMind Consortium (CMC) and GTEx-derived gene expression prediction models for 13 brain tissues and 12 tissues with potential metabolic involvement (adipose, adrenal gland, 2 colon, 3 esophagus, liver, pancreas, small intestine, spleen, stomach).We identified 35 significant gene-tissue associations within the large chromosome 12 region described in the recent PGC-ED GWAS. We applied forward stepwise conditional analyses and FINEMAP to associations within this locus to identify putatively causal signals. We identified four independently associated genes; RPS26, C12orf49, SUOX, and RDH16. We also identified two further genome-wide significant gene-tissue associations, both in brain tissues; REEP5, in the dorso-lateral pre-frontal cortex (DLPFC; p=8.52×10−07), and CUL3, in the caudate basal ganglia (p=1.8×10−06). These genes are significantly enriched for associations with anthropometric phenotypes in the UK BioBank, as well as multiple psychiatric, addiction, and appetite/satiety pathways. Our results support a model of AN risk influenced by both metabolic and psychiatric factors.

Molecular neuroanatomy of anorexia nervosa

“…Predicting gene expression from genetic information avoids these pitfalls by using reference expression data from healthy subjects. This transcriptomic imputation approach was applied to loci identified in the Duncan et al GWAS 65 . Within the rs4622308 locus, they identified 35 associations where differential expression was predicted for a specific gene and tissue.…”

Molecular neuroanatomy of anorexia nervosa

“…Predicting gene expression from genetic information avoids these pitfalls by using reference expression data from healthy subjects. This transcriptomic imputation approach was recently applied to the genetic data used in the Duncan et al GWAS (Huckins et al 2018) . Within the rs4622308 locus, they identified 35 associations where differential expression was predicted for a specific gene and tissue.…”

Chronic stress and adipose tissue in the anorexic state: endocrine and epigenetic mechanisms

“…Gene-tissue associations from a transcriptomic imputation study attributed 13.9% of the phenotype variance in AN to sWAT [83]. Moreover, decreased expression of receptor accessory protein 5 in the dorsolateral prefrontal cortex, which promotes the expression of olfactory receptors and is positively correlated with body weight, is associated with a higher risk of AN [83], probably due to the dysfunction of odour discrimination and sequential desire of food.…”