A Novel Polygenic Risk Score Indexing Somatostatin-Expressing Inhibitory Neurons Predicts Somatostatin-Expressing Cell Proportions and Severity of Symptoms in Late-Life Depression

Background Postmortem studies suggest that somatostatin-expressing (SST+) interneurons are selectively vulnerable in aging and depression. We developed a novel polygenic risk score (PRS) as an in vivo index of interindividual variability of SST-related function and evaluated its association with late-life depression (LLD)-related phenotypes. Methods We identified genes co-expressed with SST and their corresponding cis -expression quantitative trait loci in dorsolateral prefrontal cortical (dlPFC) tissue. We aggregated these variants into multiple PRSs (SST-PRSs), testing each as a predictor of SST+ cell proportion estimated from bulk postmortem tissue RNA sequencing data. The SST-PRS most predictive of SST+ cell proportion was computed in a cohort of older adults (393 with LLD; 416 never-depressed) and tested as a predictor of depressive symptoms (Montgomery–Åsberg Depression Rating Scale [MADRS]) and cognitive function (Repeatable Battery for the Assessment of Neuropsychological Status [RBANS]; Delis-Kaplan Executive Function System). To assess generalizability, we sought to replicate LLD-specific associations in a subset of older-adult UK Biobank participants (≥60 years old) with a history of recurrent depression. Results The SST-PRS was associated with reduced dlPFC SST+ neuron proportion (R = 0.5, p = 1.783 × 10⁻⁴) but not with LLD diagnosis (p = .39). Among individuals with LLD, the SST-PRS was associated with higher MADRS scores regardless of sex (t ₃₆₉ = 2.267, p = .024) and lower RBANS language scores in males only (t ₁₂₀ = −3.077, p = .003; p _interaction = .002). In the UK Biobank, the SST-PRS was modestly associated with lower Symbol Digit Substitution scores (p = .044) and higher depressive symptom scores (p = .07) in males. Conclusions The SST-PRS may serve as a novel biomarker of SST+ neuron pathology and of depressive and cognitive symptom burden in LLD.