The impact of functional architecture dynamics in bipolar disorder staging

Bipolar disorder (BD) follows a neuro-progressive trajectory, yet current clinical staging models lack robust biological markers. Based on recent evidence, we hypothesized that BD stages are associated with alterations in the functional architecture of the Triple Network Model. We analyzed resting-state fMRI data from 123 individuals spanning five clinically defined BD stages. Using a hierarchical approach, we investigated how functional connectivity features characterize these stages, transitioning from coarse- to fine-grained analyses in both clinical classification (binary early-versus-late grouping to five-stage discrimination) and connectivity metrics (mean and standard deviation of connectivity to topological dynamic features). First, low-level functional connectivity features, combined with gene expression data, discriminated early- from late-stage illness, reaching higher performance than unimodal approaches. Next, evaluating the full staging spectrum revealed a non-linear trajectory in whole-brain connection density, which was lower at each successive stage from 0 to 3 but unexpectedly higher at stage 4. Network topology analysis contextualized this finding, revealing a reorganization of functional hubs in the final stage. Specifically, three hubs within the Salience Network exhibited decreased centrality despite the overall increase in connection density. This shift corresponded to a loss of temporal synchronization between these hubs and default-mode hubs, an axis of communication mediated by the salience network that is known to be important in healthy subjects and disrupted in psychiatric disorders. Overall, our findings suggest that late-stage BD is characterized by a subtle but significant reorganization of brain architecture, accompanied by alterations in the dynamics of the salience network which loses its coordinating role.