¹Osaka Metropolitan University Graduate School of Medicine, Neuropsychiatry, Osaka, Japan, ²Osaka Metropolitan University Graduate school of Medicine, Neuropsychiatry, Osaka, Japan, ³Weill Cornell Medical College, Psychiatry, New York, United States

Background: Mood disorders still lack objective biomarkers and mechanism-guided therapeutics, partly because direct access to living brain tissue is limited. Extracellular vesicles (EVs) are lipid bilayer-enclosed nanoparticles released by cells and detectable in blood. Brain-enriched EVs may provide a minimally invasive window into brain biology, yet a scalable EV transcriptomic liquid-biopsy framework for bipolar disorder (BD) and major depressive disorder (MDD) remains underdeveloped.

Aims & Objectives: We developed a brain-enriched EV liquid-biopsy method using immunoprecipitation with neural cell adhesion molecule 1 (NCAM1), a neuron-enriched membrane marker, and tested whether plasma NCAM1-positive EV transcriptomes capture brain-relevant signals and distinguish BD, MDD, and healthy controls. We then integrated BD-associated signatures with Connectivity Map (CMap) analysis to nominate candidate therapeutics predicted to reverse BD-related transcriptional programs.

Method: Total EVs were isolated from 500 µL plasma by ultracentrifugation (110,000 g, 70 min, 4°C; 70Ti rotor) and validated by electron microscopy, nanoparticle tracking analysis, and EV marker assays. Brain-enriched (NCAM1-positive) EVs were obtained by immunoprecipitation, verified by western blotting, and processed for RNA-seq (Smart-seq2; Nextera XT; NextSeq 500). Differential expression and pathway analyses used Enrichr/Reactome, and orthogonal partial least squares discriminant analysis was used to evaluate group separation. For CMap, the 145-gene BD-versus-control differential-expression signature was used to query the CLUE/L1000 resource (>1.5M gene-expression profiles from ~5,000 small-molecule compounds and ~3,000 genetic perturbagens). Candidate perturbations were selected based on significant connectivity after multiple-testing correction (FDR-adjusted p < 0.05), focusing on those predicted to reverse BD-associated signatures. Human plasma samples included 39 MDD, 13 BD, and 15 healthy controls.

Results: EVs and NCAM1-positive EVs were validated as above. Enrichr showed that genes expressed in NCAM1-positive EVs were significantly enriched for fetal brain cortex, midbrain, prefrontal cortex, and neuronal epithelium, supporting successful brain enrichment. Between MDD and BD, 159 genes differed after multiple-testing correction (top genes: STOM, RGS18, UBLCP1, MBD2, TRBC1), and orthogonal partial least-squares discriminant analysis separated the groups. Comparing BD with controls, 145 genes were differentially expressed (FDR-adjusted), and pathway analyses suggested activation of immune/inflammatory signaling (NF-κB, IL-1/JAK-STAT, interferon response), proliferative programs (PI3K/MYC), and metabolic pathways (glucose metabolism, SREBP pathway), consistent with a coordinated immune–metabolic shift. CMap nominated potential therapeutics including dinaciclib (CDK inhibitor), TGX-221 (PI3Kβ inhibitor), fatostatin (SREBP inhibitor), KW-2449 (FLT3/Aurora kinase inhibitor), acalabrutinib (BTK inhibitor), and the muscarinic M3 antagonists otilonium and diphemanil; all nominated hits met FDR-adjusted p < 0.05.

Discussion & Conclusions: Plasma NCAM1-positive EV transcriptomes capture brain-enriched signals and provide a scalable, minimally invasive readout for biomarker discovery and biological stratification in mood disorders. Crucially, integrating this liquid-biopsy signature with CMap establishes an end-to-end, mechanism-aware route from blood to actionable drug hypotheses—prioritizing compounds predicted to counteract immune-inflammatory, proliferative, and metabolic programs activated in BD. These candidates warrant independent validation in larger cohorts and experimental models and may enable a pragmatic path toward precision repurposing and longitudinal monitoring of treatment-related molecular changes in BD.