Membrane lipid order predicts potency decline in therapeutic extracellular vesicles following handling, storage, and reconstitution stress

Introduction: Potency assays for therapeutic extracellular vesicles (EVs) are widely recommended, yet common quality-control readouts (e.g., protein concentration and EV marker levels) are not always stability-indicating. We asked whether a membrane lipid order metric can sensitively detect potency deterioration of EVs after membrane damage and logistics-relevant handling, storage, and reconstitution stresses. Methods: Human fibroblast-derived EVs were isolated by ultracentrifugation. The study comprised two parts. First, as an artificial membrane-disruption model, EVs were exposed to graded concentrations of Triton X-100 to induce controlled membrane perturbation. Second, to mimic real-world post-manufacturing stresses, EVs were subjected to vortex mixing, -80 °C freeze-thaw cycling, or liquid nitrogen-assisted lyophilization followed by reconstitution in PBS. Membrane lipid order was quantified using a polarity-sensitive dye and expressed as corrected generalized polarization (cGP) on a standard plate reader. Conventional QC candidates (particle size by DLS, protein by BCA, and CD63 by ELISA) were evaluated in parallel. Potency was assessed as wound closure in an immortalized human keratinocyte scratch assay. Bivariate and multivariable regression analyses were performed to identify QC metric(s) that best predicted potency. Results: In the Triton model, EV potency declined even under ultralow detergent conditions that produced minimal changes in particle size, accompanied by increased extravesicular protein and a cGP shift consistent with reduced lipid order. This size-potency uncoupling and membrane-leakage signature motivated us to test whether the same lipid order readout could capture potency deterioration under practical stresses. In the real-world stress models, vortex mixing, freeze-thaw, and lyophilization-reconstitution again measurably decreased lipid order. Across stressed conditions within each EV lot, cGP tracked potency loss and showed stronger predictive performance than particle size, protein, or CD63. In contrast, baseline potency differences among independently manufactured lots were not well captured by cGP, supporting lipid order as a within-lot stability indicator rather than a between-lot potency ranking tool. Conclusion: A plate reader-based membrane lipid order metric provides a rapid, practical approach to detect potency decline of therapeutic EVs after membrane disruption and logistics-relevant stresses, and may support acceptance criteria for distribution control and bedside go/no-go decisions.