Extracellular vesicles (EVs) are an exciting mode of intercellular communication that function in nearly every aspect of health and disease. Many characteristics of these particles make them particularly attractive to inform development of diagnostics and therapeutics. In health, EVs are active players in maintenance and cell signaling, functions which could be exploited for use as therapeutics or drug delivery vehicles, while in disease, EVs are often co-opted to contribute to disease progression, making them a potential target for therapeutics and a source of diagnostic markers. In parallel to our understanding of EVs’ functional diversity, there is also a growing understanding of EVs’ compositional heterogeneity both between EVs from different sources (i.e. interpopulation heterogeneity) and within EVs from a single cell source (i.e. intrapopulation heterogeneity). This heterogeneity consists of differences in chemical compositions, such as protein and nucleic acid expression, and physical characteristics, such as size and density, which can both exist across a spectrum or in distinct subpopulations. This compositional heterogeneity appears to be directly tied to EVs’ role in intercellular communication, allowing for diverse functional signaling dependent on these features. Yet, the contributions of this heterogeneity to both native function and its impact on our use of these particles in the clinic remains understudied. To this end, the objectives of this work were to identify how EV heterogeneity may contribute to disease progression and how we can exploit heterogeneity to develop more effective EV-based diagnostics and therapeutics.

To address the contribution of EV interpopulation heterogeneity to disease progression, the functions of EVs from healthy and diseased cells were compared in the context of multiple diseases. The specific targeting of EVs from breast cells known to metastasize to the brain was compared with EVs from breast cancer without organotropic metastasis and EVs from a non-cancer cell line using a novel assay combining neurons, astrocytes, and microglia. This work examined the difference in inflammatory response dependent on cell disease state and probed how inclusion of microglia altered uptake. In a similar vein, EVs from osteosarcoma cells with high metastatic potential were compared to EVs from a less metastatic cell line to understand how their interaction with macrophages may contribute to formation of the metastatic niche. This work identified potential differences within these EVs from highly metastatic cells including differential protein expression that may have contributed to the driving of macrophages to a more tumor-permissive phenotype. Finally, EVs were examined for contribution to feline calicivirus infection. Specifically, we utilized iodixanol gradient to separate EVs from co-isolated virions and reported on the contribution of each to infection. Together this work examined how interpopulation heterogeneity of EVs and their co-isolates between normal and diseased cells is co-opted during disease progression.To address the impact of EV intrapopulation heterogeneity in development of diagnostics and therapeutics, we examined EV subpopulations in the context of each. First, we reported on EV subpopulations from ovarian cancer, identifying specific subpopulations of EVs that offer lower and higher sensitivity to cancer-associated biomarkers. Second, we developed methods for quantifying heterogenous EV engineering, showing that certain methods produce different ratios of native to engineered EVs in the final product. Finally, in a systematic review, we probed the field of EV-based clinical trials focusing on the degree of incorporation of intrapopulation heterogeneity in trial design, highlighting that this is an underutilized resource. Together this work highlights how ignoring intrapopulation EV heterogeneity can weaken our use of EVs while examining how direct use of these subpopulations may allow for much more effective use of EVs in the clinic.

For more effective early-stage cancer diagnostics, there is a need to develop sensitive and specific, non- or minimally invasive, and cost-effective methods for identifying circulating tumor-associated biomolecules, including extracellular vesicles (EVs). As a rapid, label-free, non-destructive analytical measurement requiring little to no sample preparation, Raman spectroscopy shows great promise for liquid biopsy cancer detection and diagnosis. While many studies have demonstrated the promise of Raman spectroscopy to provide value for clinical diagnostics, the sensitivity and specificity of such platforms typically drops when applied to larger patient cohorts. Additionally, Raman can suffer from low signal due to the number of inelastically scattered photons (~1 in a million) produced after a sample is interrogated with a laser. Surface enhanced Raman scattering (SERS) is a powerful extension of this technique, providing orders of magnitude increase in chemical sensitivity compared to spontaneous Raman scattering. Yet it remains a challenge to synthesize robust, uniform SERS substrates quickly and easily. Raman technology has not been successfully moved into the clinic and is hindered by the need to develop more miniaturized and automated systems that are integrated with inexpensive and useful SERS materials. Thus, the objective of this dissertation work is three-fold: 1) to carry out experiments on large clinical datasets to validate Raman and SERS diagnostics; 2) to examine the value of spectroscopic analysis of EVs; and 3) to develop novel SERS materials that are robust, biocompatible, and inexpensive.To address these objectives, we carried out Raman analysis of plasma, serum, and saliva from hundreds of cancer patients and benign controls (from patients undergoing similar procedures or screenings without cancer), including patients diagnosed with head and neck, ovarian, and endometrial cancers. Several notable findings were reported arising from this analysis, ranging from optimization of Raman data collection and data analysis, discovery and application of new plasmonic materials, and applied clinical testing of EVs. First, we showed that a simple data augmentation routine of fusing plasma and saliva spectra provided significantly higher clinical value than either biofluid alone, pushing forward the potential of clinical translation of Raman spectroscopy for liquid biopsy cancer diagnostics. Next, we reported the utilization of a simple plasmonic scaffold composed of a microscale biosilicate substrate embedded with silver nanoparticles for SERS analysis of ovarian and endometrial cancer EVs. We observed a major loss of sensitivity for ovarian and endometrial cancer following enzymatic cleavage of EVs’ extraluminal domain, suggesting its critical significance for diagnostic platforms. Using SERS, we also confirmed that three common EV isolation methods (differential ultracentrifugation, density gradient ultracentrifugation, and size exclusion chromatography) yield variable lipoprotein content. However, in combining SERS analysis with machine learning assisted classification, we showed that the disease state is the main driver of distinction between EV samples, and largely unaffected by choice of isolation. Finally, we demonstrated the synthesis and characterization of a new homogeneous gold nanofoam (AuNF) substrate produced by a rapid, one-pot, four-ingredient synthetic approach. These novel AuNFs were rapidly nucleated with macroscale porosity and then chemically roughened to produce nanoscale features that confer homogeneous and high signal enhancement (~10^9) across large areas, a comparable performance to lithographically produced substrates, with high utility for application in low-resource settings The work presented below comprehensively shows the promise of Raman as a clinical diagnostic tool and takes measured steps toward validating the technology in the context of cancer disease states. The technique has high disease discrimination in whole biofluids and isolated EV populations, and the addition of novel nanomaterials increases the sensitivity and specificity to reach clinically necessary levels. This work is foundational in promoting the continued emphasis on translating Raman to be a clinically relevant diagnostic tool.