Single-cell extracellular vesicle-program scoring maps immunometabolic rewiring and immune crosstalk of mesenchymal stromal cells in intervertebral disc degeneration, prioritizing AP2S1 and CSTB

Background Intervertebral disc degeneration (IDD) is increasingly viewed as an immune-perturbed and metabolically stressed niche rather than a purely mechanical or aging-related disorder. Mesenchymal stromal cells (MSCs) and their extracellular vesicles (EVs) may shape local immune signaling, yet MSC EV–associated transcriptional programs and their immune-network context remain poorly defined at single-cell resolution. Methods We analyzed single-cell RNA sequencing (scRNA-seq) data from GSE230809 (8 IDD and 3 control nucleus pulposus samples) with standard preprocessing and batch correction. We derived a transcriptome-inferred MSC EV–associated program score by integrating complementary gene-set scoring strategies, stratified MSCs into EV-score–high versus –low states, and prioritized candidate regulators using a consensus LASSO/SVM-RFE workflow. CellChat was used to infer ligand–receptor communication, while pathway-level immunometabolic remodeling was assessed with GSVA/GSEA and immune-signature correlation analyses. Selected genes were examined by RT–qPCR in an in vitro inflammatory NP-cell model, and docking was performed to explore tractable compound–target hypotheses. Results We resolved seven major cell populations and observed a marked shift toward higher EV-program scores in MSCs from IDD tissues. EV-score–high MSCs showed increased incoming signaling and interaction potential in CellChat networks, consistent with an “immune cue–responsive” state. Machine-learning prioritization converged on five hub genes (AP2S1, CSTB, GSTP1, RPL28, and TSG101). The hub-gene program aligned with immune–metabolic axes including IL6/JAK–STAT3 and interferon-related signaling, together with oxidative phosphorylation and ROS stress signatures, and was systematically associated with immune mediators (notably chemokines and checkpoint-related genes). AP2S1 and CSTB displayed dynamic expression along MSC pseudotime trajectories and were supported by RT–qPCR under inflammatory stimulation. A hub-gene model distinguished IDD from control samples within the discovery cohort (AUC = 0.836). Docking suggested a plausible interaction between CSTB and ripasudil (−5.98 kcal/mol). Conclusions This study maps an MSC EV–associated program at single-cell resolution and places it within an immune–communication and immunometabolic framework in IDD. AP2S1 and CSTB are candidate nodes linking EV-program states to immune signaling and metabolic stress, providing candidate biomarkers and tractable intervention hypotheses.