Functional polygenic risk score of glucocorticoid-dependent regulatory element activity and its relation to clinical...

Chronic stress and adverse life events are major contributors to depression and anxiety disorders. Genetic variants that influence glucocorticoid (GC) hormone responses may moderate the impact of stress on psychiatric outcomes. Here, we use a functional polygenic risk score (fPRS) based on GC-responsive gene regulatory elements (GREs) following dexamethasone (DEX) exposure (DEX-fPRS) and examined its association with clinical and biological stress-related phenotypes.

We analyzed data from 2286 participants in the Dutch NESDA cohort, with baseline, 2-year, and 6-year follow-up visits, including 1115 with baseline depression/anxiety diagnoses. Univariable and multivariable regression models, adjusted for age, sex, batch, and ancestry principal components, were used to test associations between DEX-fPRS and lifetime diagnosis, cumulative symptom severity, and clinical features among cases (e.g. symptom severity, recurrence). Subgroup comparisons (e.g. single vs. recurrent episodes) and biological outcomes (salivary cortisol, immune markers) were also assessed. A vehicle-treated fPRS control (VEH-fPRS) was additionally tested, capturing genetic variants associated with baseline, non-GR-stimulated enhancer activity. Gene-environment interactions were tested and replication in the MARS cohort (N = 1160).

In the full sample, DEX-fPRS showed no significant associations with lifetime diagnosis, symptom severity, or biological measures after Bonferroni correction (all p_bonf>0.05), although a nominally significant negative association was observed for morning cortisol (AUCg) in the univariable model (B=-0.05, p_nominal=0.03, p_bonf=0.19). Among cases, evening cortisol showed nominally significant negative associations in both models (univariable: B=-0.05, p_nominal=0.02, p_bonf=0.09; multivariable: B=-0.04, p_nominal=0.02, p_bonf=0.14). A significant positive association was found for recurrent vs. single episodes in both models (univariable: β=1.24, p_nominal=0.01, p_bonf=0.03; multivariable B=1.21, p_nominal=0.01, p_bonf=0.04). VEH-fPRS showed similar nominal associations for recurrent episodes (univariable: B=1.13, p_nominal=0.03, p_bonf=0.11; multivariable B=1.21, p_nominal=0.02, p_bonf=0.08) and a negative association for single episodes versus controls in the multivariable model (B=-1.52, p_nominal=0.02, p_bonf=0.07). No gene-environment interactions were detected, and none of the NESDA findings were replicated in the MARS cohort.

While the DEX-fPRS was not associated with lifetime diagnosis or symptom severity, the results suggest a potential role in HPA axis sensitivity and episode recurrence. Notably, the VEH-fPRS, initially a control, may itself capture baseline transcriptional regulation. However, the lack of replication highlights the need for cautious interpretation and further validation.