Abstract
BACKGROUND: Bipolar disorder (BD) is a heterogeneous psychiatric condition characterized by distinct episodes: manic (BipM), depressive (BipD), mixed (mBD), and remission (rBD). Current evidence indicates alterations in brain functional connectivity in BD, yet a comprehensive understanding across all episodes remains incomplete. METHODS: Here, to investigate how different BD episodes alter brain functional organization, we calculated the sensory-association axis using diffusion map embedding on the functional connectome matrix and compared this axis between the four BD groups and neurotypical controls. Then, we employed regression dynamic causal modeling to investigate the directional information flow along the reorganized sensory-association axis across different BD episodes. Furthermore, we applied Nested Spectral Partitioning to decode functional integration and segregation along the same axis. Finally, we compared the reorganization patterns with normative maps of clinical symptomatology, cellular composition, and receptor distribution to elucidate symptom-related and molecular-level associations. RESULTS: Compared to healthy controls, we observed sensory region expansion and association region compression in BipM, BipD, and rBD. The mBD showed expanded visual and prefrontal regions but compressed motor and precuneus regions. Analyzing neural information flow revealed reduced connectivity in association regions for BipM and BipD, indicating association dominance in functional reorganization. Conversely, mBD exhibited heightened bidirectional signal flow between sensory and association regions, emphasizing increased integrative processing. Network analyses further revealed increased integration and decreased segregation across unipolar episodes, with the highest integration in mBD. Clinical correlations highlighted that emotional fluctuations primarily related to association region reorganization, suggesting potential biomarkers for mood episode detection. Moreover, these functional reorganizations spatially correlated with serotonin transporter, gamma-aminobutyric acid type A receptor, alpha-4-beta-4 nicotinic acetylcholine receptor, and specific cortical neuron layers (layer 4 and layer 5 excitatory neurons). CONCLUSIONS: Our findings propose functional reorganization as both a biomarker and a simplified neural phenotype framework for systematically quantifying BD-related neural abnormalities.