Large extracellular vesicles subsets and contents discrimination: the potential of morpho mechanical approaches at single EV level

Extracellular vesicles are heterogenous lipid bound membranous structures released by different cells, showing a great potential to be used as biomarkers. They have been explored by the researchers for their role in the context of environmental toxicity. When exposed to pollutants like polycyclic aromatic hydrocarbons, they undergo surface modification as well as the change of contents in their cargo. Subpopulations of large EVs (lEVs) have shown to contain both damaged and intact mitochondria which is inexorably linked to oxidative stress conditions. In this study, we aimed to perform morpho-mechanical characterization of the lEVs derived from benzo[a]pyrene (B[a]P) treated endothelial cells. The combination of fluorescence microscopy and atomic force microscopy (AFM) allowed to accurately perform the measurements on the lEVs that contained mitochondria or not. Through morphological study, we obtained the size profile of different EV sub-populations, size of EVs containing mitochondria coming from treated condition (1.8 +/- 0.89 micrometers) and from the control condition (1.63 +/- 0.76 micrometers). Through quantitative measurement of Young\'s modulus, we found that EVs containing mitochondria from treated condition showed higher stiffness than the control condition ones with the average Young\'s modulus 3.09 and 1.25 MPa respectively. We also observed the heterogeneity within the single vesicles, having majority of the Young\'s modulus values ranging from 0.1 up to 30 MPa for the treated condition and 0.1 to 5 MPa for the control condition. Finally, linear discriminant analysis (LDA) was applied as a statistical tool to discriminate EV subpopulations based on maximum diameter, height, and the Young\'s modulus value distributions. The statistical analysis using LDA confirmed our ability to detect mitochondria within EVs based on morphological and Young\'s modulus data and demonstrated the potential for discrimination between conditions. Following this, we applied the Random Forest algorithm to the morpho-mechanical data to classify EV subpopulations successfully. This approach enabled us to distinguish EVs containing mitochondria and determine whether these mitochondria-containing EVs originated from control or treated conditions, achieving accuracies of 84% and 76%, respectively. These findings highlight the combined power of morpho-mechanical analysis and machine learning for identifying and discriminating EV subpopulations.