Genetic risk for anorexia nervosa and association with general dimensions of psychopathology in childhood

Anorexia nervosa (AN) is an eating disorder characterized by restriction of food intake which results in a dangerously low body weight. AN affects approximately 1% of the population, and incidence peaks during adolescence. The most recent genome wide association study (GWAS) of AN identified 8 risk loci associated with the condition, as well as significant genetic correlations with anthropometric and metabolic traits. Gaps remain in our understanding of how genetic susceptibility to AN leads to the development of the disorder. Genetic susceptibility to AN (as measured by a polygenic risk score, PRS) is associated with lower body mass index (BMI) and with lower weight gain trajectories in the general population. Whether genetic susceptibility to AN associates with dimensions of childhood psychopathology, including how the genetic overlap with anthropometric and metabolic traits affects these associations, is unknown. We aim to address these questions in a population-based cohort of children. We obtained GWAS summary statistics for AN (n=72,517) and BMI (n=681,275) from the Psychiatric Genomics Consortium and the GIANT data repository, respectively. To disentangle the genetic correlation between AN and BMI we ran a "GWAS-by-subtraction", resulting in a set of "conditional AN" summary statistics. We developed PRSs for children in the Canadian Healthy Infant Longitudinal Development (CHILD) study, a birth cohort of pregnant women recruited between 2008 and 2012. Genotyping data were available for 2,833 children. At age 5, parents completed the Child Behavior Checklist (CBCL) assessing internalizing (e.g. anxious, withdrawal, somatic, emotionally reactive), externalizing (e.g. attention problems, aggressive behavior), and total problems (a sum of internalizing, externalizing, sleep issues and other problems). We analyzed the associations between PRSs and CBCL scores using linear regression, adjusting for sex and the first 6 genetic principal components. Our GWAS-by-subtraction results showed that after adjusting for the genetic correlation with BMI, the 8 risk loci for AN had reduced effect sizes, and 3 were no longer genome-wide significant. Conditional AN GWAS results remained highly correlated with psychiatric traits (rg=0.34 for major depressive disorder and rg=0.41 for obsessive compulsive disorder) but showed decreased correlations with anthropometric and metabolic traits (e.g. rg=-0.11 for waist-hip-ratio vs. unconditional rg= -0.22). We restricted initial analyses to 1453 children of European ancestry who had both genotype and CBCL data. Externalizing problem scores were higher for boys than girls on average (6.5 and 5.2 respectively), while internalizing problem scores were similar (5.7 and 5.4). Higher AN PRS was associated with lower externalizing problems (beta= -0.37, p=0.033), specifically attention problems (beta= -0.12, p=0.005). Higher conditional AN PRS was associated only with lower attention problems (beta= -0.09, p=0.044) but not with the composite problem scales. We found weak associations between genetic susceptibility to AN and dimensions of childhood pathology in a population-based sample of 5-year-old Canadian children. Case-control studies of AN report internalizing symptoms during adolescence as risk factors for the development of eating disorders. Therefore, we will next test the replicability of our findings in the CHILD sample using CBCL data collected at the 8-year visit, and in the UK-based ALSPAC cohort.