Intelligent Probabilistic System For Digital Tracing Cellular Origin Of Individual Clinical Extracellular Vesicles

Extracellular vehicles(EVs)are small vesicles secreted by various cell types to mediate cell-to-cell communication through the transferring of macromolecules.EVs carry multiple cargo molecules that reflect the origins of their donor cells,thus be considered as reliable and noninvasive biomarkers for early cancer diagnosis.However,the diverse cellular origin of EVs masks detection signals generated by both tumor or non-tumor-derived cells.Thereby,the capability to recognize cellular origin of EVs is the prerequisite for their diagnostic applications.In the present study,we develop an intelligent probabilistic system for tracing cellular origin of individual EVs that enabled by single-molecule multicolor imaging.Through quantitative analysis of the expression profile of two typical protein markers,CD9 and CD63,on single EVs,accurate and rapid probabilistic recognition of individual tumor and non-tumor cells derived EVs in clinical samples is achieved,the correlation between cellular origin and surface protein phenotyping of single EVs is also exemplified.The proposed system holds great potential for advancing EVs as reliable clinical indicators and exploring their biological functions.This paper is divided into three parts,as follows:Chapter 1 OverviewFirst,this chapter introduces EVs and tumors,including their production pathways,components,and their applications in tumor diagnosis and treatment.Then,it focuses on the separation and characterization of EVs and the application of total internal fluorescence microscopy to the detection of EVs.Finally,the specific content of this paper and the research significance of the paper are elaborated.Chapter 2 Construction of a single EV capture interface and fluorescence imaging of a single EVsUsing the glass interface modification method,the EV-specific antibody CD81 /Ep CAM is fixed on the interface to capture EVs.The morphology,concentration,and surface protein of EVs were characterized by TEM,NTA and WB.Afterward,the EVs were labeled with anti-CD9/Alexa Fluor 647 and anti-CD63/Alexa Fluor 488 fluorescent antibodies,and the fluorescence of different channels was observed under the TIRF microscope to achieve the purpose of detecting the surface protein of a single EV.Experiments have found that the fluorescence intensity of EVs secreted by different cells is very different,indicating that the CD9 and CD63 proteins expressed by EVs from different cell sources are very different.The construction platform realizes the fluorescence imaging of single EVs,and the results show that,compared with CD81 to capture EVs,the effect of using Ep CAM to capture EVs is better.Chapter 3 Using machine learning algorithms to trace the source of tumor EVs in plasma samplesBased on the colocalization pictures of two fluorescent antibodies,this chapter extracts the fluorescence intensity,statistically analyzes the fluorescence intensity of EVs from different sources.Labeling and classification.Using machine learning algorithms(KNN and SVM)to extract features of EVs from different sources.Using this as a basis to trace the source of exosomes.The principle is that the amount of surface protein expression of EVs secreted by different cells is inconsistent,resulting in different fluorescence intensities.Based on this,we have successfully used single-molecule fluorescence intensity distribution to trace the source of EVs.This method has been successfully used in the detection of plasma from clinical samples and can distinguish whether EVs are derived from cancer cell plasma or normal plasma.