Insights into the genetics of perinatal depression via polygenic scores

Perinatal depression (PND) is a form of depressive illness that affects women in the period before and after birth. Heritability of PND (40-55%) is estimated to be higher than that of non-PND depression (28-44%). However, to date, there are no sufficiently powered genome wide association studies (GWAS) of PND, resulting in its underlying genetic structure remaining largely understudied. In this work, we leverage the UCLA-ATLAS Biobank, which links genotypes to electronic health records (EHR), to study the genetics of PND. We define PND cases as women with no prior history of depression who develop depression or receive antidepressants during pregnancy or up to 1 year after birth. We generate polygenic scores (PGS) from existing GWAS of psychiatric traits along with hormonal, pregnancy-related, and inflammation related traits. Scores were generated by SBayesR, a Bayesian method that adjusts GWAS summary statistics for linkage disequilibrium. Using logistic regression, we perform three separate analyses. We compare PGS between PND cases to mothers with non-PND pregnancies, PND cases to depression unrelated to birth, and stratifying ante- (APD) and post-partum (PPD) depression in PND cases. Within the 36778 UCLA Health patients in ATLAS, we identify 417 PND cases, along with 986 non-PND pregnancies (PND controls) and 7141 non-PND depression cases. Of the PND controls, 848 were retained after filtering for history of major depression or antidepressant use. The majority (79.7%) of PND cases were diagnosed after giving birth. PND cases compared to controls (without history of depression) showed a nominally increased PGS for bipolar disorder (OR = 1.14; p = 0.026). PND cases compared to non-PND depression patients have lower PGS estimates for major depression (MDD) (OR = 0.87; p = 0.009) and ADHD (OR = 0.88; p = 0.009), but higher estimates for age at menarche (OR = 1.14; p = 0.012). After stratifying PND cases by time of onset (i.e.: APD and PPD) and comparing them separately to controls, the association of the MDD PGS was limited to APD (OR = 1.25; p = 0.05) and absent in PPD (OR = 1.07; p=0.29). This effect is even larger than the one observed between PND cases and controls. (OR = 1.11; p = 0.096). In the absence of cohorts sufficiently powered to perform GWAS for PND, polygenic scores of related phenotypes facilitate a cursory investigation into the genetic basis of PND. We observe genetic differences between PND and non-PND depression, suggesting a distinct depression phenotype between the two. While existing studies of depression around childbirth place much emphasis on the postpartum period, this work aims to broaden this scope to consider depression perinatally. Interestingly, we also find differences in estimates of genetic risk for MDD between patients with onset of PND before or after pregnancy. This may be indicative of varied disease etiology. We are currently comparing risk predictions from a clinical predictor of PND to assess the relationship between clinical covariates and genetic risk estimates. We are also analyzing a larger release of ∼60000 ATLAS patients, and aim to replicate our results in an external validation cohort.