Bone marrow immune remodeling in depression: TNF/NF-κB mediated leukocyte redistribution and construction of an interpretable predictive model.

Environmental stress contributes to the development of depression through neuro-immune interactions, yet the underlying molecular mechanisms and associated clinical diagnostic biomarkers remain unclear. We established a psychosocial stress mouse model and systematically investigated the immune dysregulation induced by stress through integrated analysis of blood cell profiles, leukocyte transcriptomics, protein-protein interaction networks, single-cell RNA sequencing, and targeted pharmacological intervention. Additionally, we constructed and validated a depression predictive model using multiparametric peripheral blood data and machine learning, and assessed feature importance using the SHapley Additive exPlanations (SHAP) analysis. The results show that stressed mice exhibited increased neutrophils and decreased lymphocytes, which were positively correlated with depressive-like behaviors. Differential expression analysis identified 127 differentially genes enriched in pathways related to immune response, inflammation regulation, and hematopoietic dysfunction. Hub-genes (Rel, Cd274, Icam1, Tnf, Maff, Nfkbia) were upregulated and drove leukocyte redistribution via inflammatory cascades and adhesion mechanisms, whereas TNF/NF-κB pathway inhibition (Etanercept/Infliximab) reversed bone marrow immune imbalance and ameliorated depressive-like behavior. The random forest (RF) model trained with RF-mean features and random oversampling resampling based on blood cell parameters achieved 91.60 % accuracy and an AUC of 0.65. SHAP analysis identified RDW, MOPCT, PLTSI, SII, and NPR as biomarkers. In short, this study elucidates the regulatory role of bone marrow immune remodeling in stress-induced depression and proposes a feasible predictive strategy based on multiparametric peripheral blood profiling.