Neuroanatomical correlates of transdiagnostic clinical phenotypes in major depressive and bipolar disorders

Major depressive disorder and bipolar disorder are characterized by substantial clinical heterogeneity and overlapping symptom profiles. However, it remains unclear whether transdiagnostic symptom phenotypes show partially dissociable associations with cortical and subcortical morphology and whether these brain-symptom associations show cross-sectional age-related variation across adolescence and emerging adulthood. In a clinical cohort of patients with major depressive disorder or bipolar disorder ( N = 1,023), item-level K-means clustering was applied to 174 clinical measures to derive non-overlapping transdiagnostic symptom phenotypes. High-resolution 3T structural MRI data were processed using FreeSurfer to extract cortical thickness, cortical surface area, and subcortical volumes. Sparse canonical correlation analysis was used to characterize multivariate associations between each symptom phenotype and each structural modality. Model significance and feature stability were evaluated using permutation testing and bootstrapping. Continuous age moderation was further tested in 960 participants aged 10 ~ 29 years. K-means identified two internally stable transdiagnostic phenotypes: Negative Affect-Anhedonic phenotype, characterized primarily by self-injurious tendencies and depressive/anhedonic symptoms, and Manic-Activation phenotype, characterized by manic and hypomanic activation. Both phenotypes showed significant multivariate associations with cortical thickness and cortical surface area, but not with subcortical volumes in the full sample. Cortical thickness associations involved focal higher-order association regions, whereas cortical surface area associations were spatially distributed but network-organized. Phenotype dominance partially aligned with categorical diagnosis but showed substantial cross-diagnostic overlap and within-diagnosis heterogeneity. In revised age-interaction analyses, no age-by-symptom interaction survived FDR correction across the six tested models. The smallest nominal effect was observed for the Negative Affect-Anhedonic-subcortical volume model, but this effect did not survive correction and was therefore interpreted as exploratory. Transdiagnostic symptom phenotypes capture clinically meaningful heterogeneity beyond categorical diagnosis and map onto partially dissociable, modality-specific structural brain patterns. Cortical thickness and surface area showed divergent spatial organizations, whereas age-moderation analyses did not identify any interaction that survived correction across the six tested models. Together, these findings support a multimetric dimensional framework for understanding shared and distinct brain-symptom relationships across major mood disorders, while indicating that age-related moderation effects require cautious interpretation. This study was registered at the Chinese Clinical Trial Registry with the registration number ChiCTR2500106094. Registered on July 17, 2025.