Mapping reproducible brain-based depression stratification across clinical, cognitive, and neurotransmitter dimensions.

Identifying reproducible brain-based stratifications across independent cohorts is critical for parsing heterogeneity in Major Depressive Disorder (MDD), yet robust profiles spanning micro- and macroscales remain poorly defined. We applied stability-based clustering to cortical thickness (CT) data from 1531 MDD individuals in UK Biobank (UKB), with replication in 144 inpatients from IRCCS San Raffaele Hospital (HSR). Two stable clusters emerged (normalized stability = 0.125) that generalized to a hold-out UKB sample with 96.5% accuracy (majority-cluster baseline = 59.5%). The clusters differed in widespread CT, largely reflected in global CT, with one showing cortical thinning, childhood trauma, and diabetes comorbidity. A two-cluster structure was independently replicated in HSR (normalized stability = 0.29), with a broadly similar cortical topography, although systematic CT differences between cohorts limited direct cross-cohort comparability. The clinical correlates identified in UKB did not replicate after correction for multiple comparisons. In UKB, mapping the CT differences between strata onto Neurosynth meta-analytic activation patterns revealed a ventral-dorsal gradient linked with emotion regulation, interoceptive, and motivational processes. Spatial correlations with 19 neurotransmitter receptors and transporters obtained from positron emission tomography identified the dopamine transporter as the dominant contributor in UKB, and the histamine H3 receptor in HSR. These findings provide a reproducible framework linking CT-based clusters in MDD to multiscale biological organization, although their clinical relevance warrants further investigation.