Characterization and selective isolation of human extracellular vesicle subpopulations from simple and complex biosamples for novel clinical liquid biopsy strategies

All living cells selectively package a wide variety of biomolecules in nanometric entities termed extracellular vesicles (EVs), which actively mediate vital biological events, especially biomolecule recycling and horizontal intercellular communication. To do so, EVs can diffuse through interstitial space, reaching even circulating biofluids to promote long-range paracrine signalling in recipient cells, while sometimes directly functioning as effectors. Despite fundamental knowledge gaps within the field, EVs are frequently dysregulated during disease, often directly sustaining disease hallmarks. Therefore, they are extremely valuable circulating biomarker sources that can provide timed snapshots of individual (patho)physiological status, while carrying cell or tissue-specific cargo, indicative of their origin. Consequently, attention towards EVs has exponentially increased during the last decade, fostered by translational research efforts that have focused on two major EV applications, either as drug delivery vehicles or as analytes in liquid biopsies, which consist in bodily fluid collection for biomarker detection. In contrast to standard tissue biopsies for cancer management, liquid biopsies are heralded as the holy grail of personalized medicine because they are minimally invasive, low-risk, high-throughput and repeated sampling is easy and cheap. These features enable full understanding of tumour heterogeneity, while tumour progression can be tracked longitudinally, as well as resulting alterations in healthy tissues. Liquid biopsies aid clinicians in early disease detection and screening, patient stratification, treatment response monitoring and resistance mechanism identification. Still, the small size and low amount of cargo in biological nanoparticles limit current state-of-the-art technologies employed for EV profiling. These fail to grasp the full complexity of EV heterogeneity, ultimately straggling the establishment of model EV samples and identification of clinically relevant EV subpopulations. Hence, the first central aim of this thesis, further described in the first chapter of the results section, concerns the identification of EVs among particle noise and characterization of relevant subpopulations, using state-of-the-art, dedicated instruments, optimized for high-resolution single nanoparticle detection. Label-free and fluorescence measurements allowed either indiscriminate or selective detection of EV subpopulations and common contaminants. Analysing the expression of classical surface markers facilitated EV sample characterization. Using thoroughly optimized staining procedures, fluorescently-labelled EVs could be reproducibly generated, and later employed as tracer EV models for accurate measurements in downstream spike-recovery experiments. Liquid biopsy studies have often selected blood plasma for EV biomarker discovery and detection, as it potentially transports EVs from any cell type, together with other biomarkers. Since blood collection is a well-established routine clinical practice, it is often considered the ideal biofluid for liquid biopsy tests. The majority of these studies purify bulk EVs with co-isolated contaminants from plasma, failing to investigate the specific EV subsets that actually contain relevant biomarkers, which might end up undetected, diluted among many other macromolecules. Thus, in the second chapter of results, the present thesis aimed to further dissect EV heterogeneity in human plasma samples, and to determine whether distinct EV subpopulations can indeed confer differential clinical utility. To this end, several affinity-based EV isolation approaches were optimized and explored for their efficiency and specificity in simple and complex matrices. However, the complexity of biological matrices, especially plasma, drastically hamper affinity interactions. Hence, additional critical gaps in the field regarding sample pre-analytical processing were addressed. Results demonstrated that target EV subpopulations could be efficiently enriched from plasma, but also, that both complex matrix composition and target EV surface phenotypes can dramatically influence the performance of affinity isolation methods. Furthermore, distinct plasma EV subpopulations were indeed captured when targeting different surface moieties. Platelet-derived EVs, but not other EV subsets, evidenced 8 mRNA expression signatures that could be linked to early-stage lung cancer, proving that distinct EV subpopulations are indeed more relevant than others for liquid biopsy-based biomarker detection. In conclusion, the original experimental work reported in thesis evidenced that EV labelling protocols should be optimized for each single nanoparticle profiling platform. Moreover, fluorescent dyes, affinity reagents and methodologies employed must be carefully selected and tested, as all contribute to accurate EV measurements in simple matrices, but can also promote confounding particle detection and biased analysis. Also, elimination of dyes in excess is extremely important for precise small EV detection, and protocols used to conduct this step must not affect the original subpopulation composition in EV samples. On the other hand, the immunoaffinity-based EV isolation protocol devised in this PhD project purified distinct EV subpopulations from plasma. This protocol is compatible with a wide range of downstream procedures and analytical platforms. It is a simple, quick, scalable and automatable pre- analytical workflow, fitting the requirements of routine clinical assays, as to encourage the inclusion of EVs in novel liquid biopsy tests. Importantly, clinically relevant mRNA expression profiles associated to early-stage lung cancer, were obtained upon isolation of platelet-derived EV subpopulations from plasma, using this immunoaffinity protocol. Furthermore, it was evident that pre- analytical variables, sample matrices and EV surface phenotypes are critical parameters to account for when affinity-based methodologies are employed to target specific EV subsets. As result of the research output throughout this PhD project, ongoing multi-centre collaborations have been established, with the goal of integrating EV multianalyte and multi-omics data. Large-scale validation, prospective and explorative studies will be fundamental for robust biomarker detection and inclusion of EVs in novel liquid biopsy-based assays.