Eukaryotes generate multiple kinds of EVs. The major classes include exosomes, which are generated from multi-vesicular endosomes (MVEs), shedding microvesicles, which originate from the plasma membrane, apoptotic blebs, which result from programmed cell death, and gesicles, which are produced by vesicular stomatitis virus and potentially incorporate specific cargo proteins [40]. Exosomes are generated by the inward budding of MVEs and early endosomes which is mediated by the actions of endosomal sorting complexes required for transport (ESCRTs), ceramides, and tetraspanins [41]. The inward budding results in the potentially selective incorporation of cytoplasmic components including RNA, DNA, and proteins into the lumen of the exosome, as well as membrane lipids and transmembrane proteins. If the resultant MVEs are targeted to the plasma membrane instead of the lysosome, the fusion of MVEs with the plasma membrane leads to the release of exosomes. The fusion event is mediated by the action of multiple Rab proteins, including RAB27A (RAB27A, member RAS oncogene family), RAB35 (RAB35, member RAS oncogene family), and RAB11A (RAB11A, member RAS oncogene family) [42] in addition to the action of the cytoskeleton and fusion machinery (soluble NSF attachment protein receptors [SNAREs], etc.) [41]. The proteins involved in the processing of MVEs into EVs, including the ESCRTs, Rabs, and SNAREs, are present across the eukaryotes, which argues for EV production in at least the last common eukaryotic ancestor. The presence of orthologs of ESCRTs in archaea suggests that the last universal common ancestor may also have produced EVs [43]. However, since these proteins are also essential to ubiquitination pathways, it is possible that the conservation of EV production is a by-product of the conservation of the targeting of proteins to the lysosome for degradation.

The budding of microvesicles from the plasma membrane requires the actin-myosin machinery and small GTPases such as ADP ribosylation factor 6 (ARF6), but does not appear to require ESCRTs, although there is evidence that ESCRT-I is associated with some microvesicles. Further advances in microscopy, including autofluorescence contrast microscopy, are likely to enable the visualization of EV production [44].

The content of EVs often differs significantly from the cellular compartments from which the EVs are generated, both in terms of membrane composition and surface markers, as well as the contents of the lumen (RNA, DNA, and proteins) (Fig. 2) [16,17]. While some of the differences in membrane composition may be due purely to free energy considerations within the membrane such as curvature and charge, the enrichment of specific RNA motifs (ACCAGCCU, CAGUGAGC, and UAAUCCCA) [45] and protein surface markers (epithelial cell adhesion molecule [EPCAM], Erb-b2 receptor tyrosine kinase 2 [ERBB2], protein tyrosine phosphatase, receptor type C [PTPRC], or CD63 molecule [CD63]) [46] argues for the selective packaging of EV contents. Proteins that may fulfill the role of RNA sorting include members of the ESCRT-II complex as well as the RNA-induced silencing complex (RISC) (including trinucleotide repeat containing 6A [TNRC6A] and argonaute 2 [AGO2, RISC catalytic component] [41,47]), but the complete picture of the RNA sorting is still unclear. For example, it is not yet clear whether the sorting of RNA into exosomes is for the primary purpose of export in EV or if it is a side effect of targeting RNA to the lysosome which is sometimes exploited to provide export in EV. Pulse-chase experiments with labelled RNA may help elucidate the kinetics of this transport, and thereby determine whether specific RNAs are targeted for secretion, or if the process is RNA agnostic.

There are multiple hypotheses for the ancestral function of EVs. The first set of hypotheses deal with the function of the content of EVs; the second with the need to enclose the content of EVs within vesicles, and the degree of importance of the vesicle membrane surface markers. EVs from gram-negative bacteria have been shown to contain Quorum sensing molecules [48], toxins (Shiga toxin [49], ClyA [50]), and immune system modulators [5,8]. Enteroparasites such as nematodes have also been shown to modulate innate immunity by transferring small RNAs (miR-100 and Y RNA) to mammalian cells [19,51].

In humans, EVs can modulate the immune system by presenting antigens to antigen presenting cells (APCs) through internalization or fusion, being released by APCs to activate natural killer (NK), CD8⁺ T, and CD4⁺ T cells, directly activating neutrophils, macrophages and NK cells, repressing immune responses through CD8⁺ T cells, NK cells, or myeloid derived suppressor cells, and finally by conveying opsonins and complement components to modulate phagocytosis and apoptotic cell phagocytosis [52]. These mechanisms are even exploited by some cancer cells by releasing EVs which induce apoptosis of activated T cells [52]. During human pregnancy, EVs with a distinct population of mRNA and miRNA are released from villous trophoblasts in the placenta; these EVs also contain FasL and TRAIL which promote apoptosis and subsequent immunosuppression of T cells [53,54]. These EVs are likely utilized by the placenta to help maintain allograft tolerance during pregnancy [24].

Using vesicles for packaging EVs can reduce the exposure of the contents to host proteases and antibodies [48], as well as coupling cargo with targeting receptors/agonists without requiring direct targeting molecule to cargo interactions. This mechanism of packaging hints at the vast potential of therapeutic cargo inside of EVs with specific surface markers for the treatment of many diseases, such as cancer, where a cargo needs to be delivered to a precise tissue or cell type which is otherwise difficult to access (Table 1) [28].

The methods of isolation currently used include ultracentrifugation, density gradient centrifugation, size exclusion through membranes, polymeric precipitation, immunoaffinity capture, and microfluidics [46,88,89]. Ultracentrifugation is an easily executed protocol, but also results in the enrichment of proteins with high sedimentation rates which are not bound to exosomes, including major vault protein, heparan sulfate proteoglycan 2 (HSPG2), fatty acid synthase (FASN), and the 26S proteasome [88]. Density gradient centrifugation reduces the contamination with high sedimentation rate protein complexes, but requires additional western blotting steps to verify that the density fraction isolated contains exosomes and has lower effciency than immunoaffinity capture.

Immunoaffinity capture using bead-conjugated antibodies to EPCAM or another exosome marker (such as ERBB2, PTPRC, or CD63) is simpler than density gradient centrifugation, and produces exosomes of high purity, but is limited to exosomal populations which express specific surface markers and may potentially miss novel exosome populations. Therefore, discovery experiments looking for novel populations of exosomes should use density gradient centrifugation, whereas established assays looking at known exosome populations can use appropriate immunoaffinity capture methods. Witwer et al. [89] and future recommendations of the International Society for Extracellular Vesicles provide valuable recommendations for investigators examining EVs in humans.

For circulating miRNA, Fesler et al. [90] suggests uniform volumes of plasma or exogenous sequences and points out that endogenous control genes may vary between subjects (as well as in the control population).

Because EVs located in body fluids can potentially originate from any source tissue for which there is a path to that fluid, accurate diagnostics of remote tissues also require identification of the tissue of origin of a particular EV or set of EVs. As the miRNA and mRNA contents of EVs came from cells in the originating tissue, miRNA and mRNA which are specific to those originating tissue but found within EVs in body fluid indicate that at least some of the EVs came from the original tissue. For example, the presence of high abundance placenta-specific transcripts such as LGALS14 in a sample of EVs would indicate that at least some of the EVs came from the placenta. Determining the fraction of EVs which are from a particular tissue of origin would allow for changes in miRNA and mRNA abundance over time to be assigned to a particular source tissue when multiple samples of different per-tissue abundances are obtained at a single timepoint. Marker-free (such as CellCODE [91]) and/or marker-dependent methods will likely be necessary to determine the proportion of exosomes which come from different source tissues in a mixed population.

In cases where the source tissue has not been assayed, an alternative is the (much more diffcult) process of sequencing individual EV separately. Presumably, the advances in single-cell and single nucleus sequencing will provide the tools for single EV sequencing as a side benefit. Previous work by Smith et al. [92] using laser tweezers Raman spectroscopy that identified the lipid content of EVs on a per-EV basis identified multiple exosome subtypes which do not correlate strongly with the originating cell type, and therefore may represent different exosome subclasses which may be present in a single bulk exosome preparation. However, Raman spectroscopy is very slow and does not easily enable identification of the contents of exosomes. In an alternative approach, Kibria et al. [93] used micro flow cytometry after differential centrifugation to analyze CD44 (CD44 molecule) surface marker intensity on individual exosomes. This overcomes the limitations of typical flow cytometers, which can only detect minimum particle sizes of 200–500 nm, which is much larger than typical EVs. We suspect that the combination of micro flow cytometry with droplet based barcoding and sequencing techniques can enable the sequencing of the RNA contents of individual EVs in the not-too-distant future.