UCLA

Cancer is one of the leading causes of premature death worldwide. Currently, when diagnosed at advanced stages, the five-year survival rates of cancers remain low even with tremendous amounts of medical resources dedicated to the treatments. To enhance the survival rates and reduce the cancer-caused socioeconomic burdens, population-based cancer screening is desired. Such screening requires the detection techniques to be accurate, patient friendly, and easy to operate at low costs. Over the past decade, cell-released small extracellular vesicles (sEVs, sized ⌀ 30−150 nm) have been attracting increasingly amounts of attention as a source of biomarkers for minimally invasive cancer detections. While existing stably in the bodily fluids, sEVs reflect their cells of origins, including cancer cells, through the encapsulated cargos such as DNA/RNA and proteins. Therefore, it is possible to detect cancers by investigating the sEVs captured in the bodily fluids. This thesis aims to explore the potential clinical applicability of analyzing individual sEVs by surface-enhanced Raman spectroscopy (SERS) for minimally invasive cancer detections. To objectively examine the SERS spectral features collected from sEVs, customized machine learning algorithm is utilized. Further, the combination between SERS and machine learning is named as “SERS Identification of Molecules” or “SIM”.The thesis work begins with understanding the heterogeneity of sEVs, namely sEV subpopulations, from the single-vesicle level by SIM. Currently, sEVs, even from the same cell of origin, are recognized to be heterogeneous which can be further sub-fractionated. Size discrepancy is one of the most popularly used criteria for separating the vesicular subpopulations. In Chapter 2, SIM was employed to spectrally detect and analyze individual sEVs from cell lines isolated based on their size discrepancies, forming subpopulations. The results suggested that sEVs in different size groups carried different chemical compositions, which could be reflected by the distinguishable SERS spectral features. Chapter 3 & 4 include the explorations of the clinical applicability of SIM in the sEV-based minimally invasive cancer detections. In the first case, SIM was applied to analyze sEVs derived from the saliva, blood, and tissue samples between gastric cancer (GC) patients and non-GC participants (n = 15 each). The algorithm prediction accuracies were reportedly 90, 85, and 72%. “Leave-a-pair-of-samples out” validation was further performed to test the clinical potential. The area under the curve of each receiver operating characteristic curve was 0.96, 0.91, and 0.65 in tissue, blood, and saliva, respectively. In addition, by comparing the SERS fingerprints of individual vesicles, a possible way of tracing the biogenesis pathways of patient-specific sEVs from tissue to blood to saliva was provided. The second case involved SIM in analyzing sEVs derived from bronchoalveolar fluid (BAL) for non-small cell lung cancer (NSCLC) detection. BAL samples were collected from NSCLC patients and non-cancer participants (n = 10 each). Analyzing the SERS spectra collected from the BAL-derived sEVs revealed the SERS spectral distinguishability between the patient and the control group in general. In addition, the sEV spectra collected from patients with early and late-stage NSCLC were also distinct. Blind test to examine the clinical usage of SIM was perform (n = 6 each for the machine learning model developing and n = 4 for the testing). Under such setup, the model correctly predicted diagnostic results from all the eight individuals in the testing set. Collectively the results obtained from the two cancer case studies indicate the clinical potential of the minimally invasive cancer detections via SIM analyzing sEVs derived from bodily fluids.