Circulating immune cells phenotypes discriminate bipolar disorder from major depressive disorder during a depressive...

MDD and BD are characterized by differences in peripheral immune markers, in frequencies of T helper, T cytotoxic, Natural Killer (NK) and T regulatory cells that may contribute to the development of mood disorders. However, few studies have used immune cells populations to discriminate MDD from BD with inconsistent findings. We applied an elastic-net machine learning pipeline predictive model using 5-fold nested cross-validation based on a wide panel of immune cell populations to discriminate between MDD and BD. The sample included 124 MDD and 89 BD hospitalised for depressive episode, and 40 healthy controls (HC). The predictive model achieved an accuracy of 78%, with 76% sensitivity, 80% specificity. The model performed significantly better than chance (p < 0.001). BD diagnosis was associated with higher frequencies of cells populations producing proinflammatory cytokines (i.e. CD4⁺IFN-γ, CD4⁺IL-17, CD4⁺TNF-α, CD8⁺TNF-α, NK cells/TNF-α), and regulatory cells (FOXP3), whereas MDD diagnosis was associated with elevated frequencies of early activating memory cells both T helper and cytotoxic (i.e. CD4⁺CM/CD69⁺, CD4⁺EM/CD69⁺, CD8⁺CM/CD69⁺), and cytotoxic effector cells producing apoptotic proteins (CD8⁺EM/Perf⁺GrzB⁺, NK cells/Perf⁺GrzB⁺). Excluding patients in lithium only early activated and cytotoxic cells differentiated the two diagnoses. Limitations include possible residual instrument-related effects and cryopreservation effects, lack of generalization to other phases of the illness, lack of external validation. Our data confirm that MDD and BD show different immune signatures and suggest a more pronounced alteration of immune cell populations in MDD especially in cytotoxic CD8+ cells. Future studies should address these differences to develop more effective therapies for mood disorders.