Exosomes are the most studied EV subtype, with the term ‘exosome’ often used interchangeably with ‘extracellular vesicle’ despite the lack of consensus on exosome-specific biomarkers and unavailability of methods to selectively isolate exosomes from other subtypes of EVs [38]. Exosomes are specifically defined as EVs of endosomal origin, formed through the inward budding of the membrane of late endosomes/multivesicular bodies (MVBs) to form intraluminal vesicles (ILVs), followed by the fusion of these MVBs with the plasma membrane which then releases exosomes from the cell [42] (Figure 2). The endosomal sorting complexes required for the transport (ESCRT)-dependent pathway is the main well-characterised mechanism of exosome biogenesis. It consists of five complexes which work to form ILVs in MVBs [43]. Briefly, ESCRT-0, consisting of Hrs (hepatocyte growth factor-regulated tyrosine kinase substrate) and STAM1/2 (signal transducing adaptor molecule 1/2) binds phosphatidylinositol 3-phosphate at the endosomal membrane and ubiquitin on ubiquitinated proteins, enabling ubiquitin-dependent sorting into endosomes [44,45]. ESCRT-0 recruits ESCRT-I (composed of tumour susceptibility gene 101 (Tsg101), human multivesicular body 12 (hMvb12), and the vacuolar protein sorting (Vps) proteins, Vps28 and Vps37) [46]. ESCRT-I interacts with both ESCRT-0 and ESCRT-II. ESCRT-II consists of Vps22, Vps36 and two Vps25 subunits [47]. ESCRT-II has a high affinity for phosphatidylinositol 3-phosphate, enabling localisation to endosomes. ESCRT-III is composed of the charged multivesicular body proteins (CHMPs), CHMP2, CHMP3, CHMP4 and CHMP6 [48]. ESCRT-III monomers exist in an autoinhibited state in the cytoplasm and come together upon activation and recruitment by ESCRT-II [49]. Cargo is then deubiquitinated prior to loading into ILVs and ESCRT-III dissociates from the endosome membrane by the action of ATPase Vps4 and its co-factor, vacuolar protein sorting-associated protein VTA1 homolog (VTA1) which enables the recycling of the ESCRT machinery [43]. Through this process, the endosomal membrane undergoes inward budding before scission occurs and ILVs are released into the MVB [50]. RAB proteins then mediate the transport of the MVBs to the plasma membrane along microtubules where the MVBs then dock and SNARE (soluble N-ethylmaleimide-sensitive fusion protein attachment protein receptor) complexes mediate the fusion of MVBs with the plasma membrane, releasing the ILVs as exosomes [51,52].

Additionally, exosomes can be formed in an ESCRT-independent manner. Sphingomyelinase-enriched microdomains containing ceramides within endosome membranes are thought to promote a negative curvature in the membrane, resulting in inward budding [53]. Tetraspanins also participate in cargo selection [54]. Furthermore, RAB31 has been recently shown to control an ESCRT-independent pathway where it works through flotillin proteins to selectively package cargo into ILVs and inhibits RAB7, preventing lysosomal degradation and enhancing the release of ILVs as exosomes [55].

Microvesicles or ectosomes are formed at the plasma membrane by outward budding of the membrane followed by fission through the use of contractile machinery [56] (Figure 2). These are more heterogenous in size than exosomes. Although originally thought to be between 100 nm and 1000 nm, numerous studies have shown the existence of much smaller and larger vesicles sharing a similar biogenesis with microvesicles, for instance large oncosomes, which can reach around 10 μm in size. During microvesicle biogenesis, cargo is selected and trafficked to the membrane to be shed as microvesicles. Ras-related GTPase ADP-ribosylation factor 6 (ARF6)-positive recycling endosomes play an important role in directing cargo to the budding microvesicle. ARF6 has been found to promote the selective packaging of integrin β1, MHC-I (major histocompatibility complex class I) proteins and ARF6 itself into microvesicles [57]. The packaging of miRNA is less well-understood, however, recent evidence indicates that exportin-5 and hnRNPA2B1 play an important role. A number of other proteins participate in cargo delivery to the nascent microvesicle, including RAB22A, SNARE machinery, CD-9 and RNA-binding proteins [56]. ARF6 further promotes scission of the budding microvesicle. ARF6 promotes phospholipase D activation, recruiting ERK (extracellular signal-regulated kinase) to the membrane where it can phosphorylate myosin light chain kinase (MLCK) [58]. This then phosphorylates myosin light chain (MLC), promoting contraction around the budding microvesicle, enabling their release from the plasma membrane [57]. Moreover, RhoA works through the ROCK (rho associated coiled-coil containing protein kinase) signalling pathway to promote contractility, however, much of this pathway remains to be elucidated [59].

As with many biological components, the production of a pure EV fraction is currently unachievable. However, separation and enrichment of EVs from biological fluids can be carried out, enabling the use of EV-enriched fractions in research, diagnostics and disease monitoring [38]. The separation of EVs from conditioned cell culture media for in vitro studies and from biofluids such as plasma, urine and cerebrospinal fluid (CSF) for diagnostic purposes can be achieved in numerous ways. Among the most common methods are differential ultracentrifugation (DUC), size exclusion chromatography, ultrafiltration, precipitation and immunoaffinity-based techniques [60]. The EV fractions produced by these techniques differ in yield and purity. Importantly, they also produce EV fractions with different proportions of EV subtypes and with different levels of contaminating co-segregating components, such as apolipoproteins A1/2 and B in serum-derived EV fractions and bovine serum albumin (BSA) and secreted proteins in cell culture medium-derived EV fractions [38]. Although no method currently exists to purify different EV subtypes such as exosomes or microvesicles, this does not preclude the use of EVs in diagnostics as the EV fractions as a whole, once separated using one or a combination of these techniques, carry diagnostic, predictive and prognostic relevance in many cancers [61]. Analysis of EVs frequently includes nanoparticle tracking to determine size and quantity of EVs, transmission electron microscopy for observation of lipid bilayer vesicles and Western blotting for detection of membrane-associated and cytosolic EV markers, such as tetraspanins or integrins.

A vast wealth of knowledge exists on the specific molecular components of EVs derived from BC cells and TME cells which are discussed in the next section. Briefly, EV cargo consists of proteins, nucleic acids, lipids and other bioactive molecules [38]. Protein cargo can include various molecules involved in signal transduction, for instance receptors and ligands. Some proteins involved in the biogenesis of the different EV subtypes are also enriched in those EVs, for instance ARF6 is found in microvesicles and ESCRT components are frequently detected in exosomes (Table 1). Additionally, immunoregulatory molecules such as immune checkpoint inhibitors and immunosuppressive cytokines can be found in EVs. Nucleic acid cargo varies between EV subtypes. Large DNA fragments have reportedly been detected in apoptotic bodies whereas smaller EV subtypes contain a more limited selection of nucleic acids, mainly consisting of miRNAs, other non-coding RNAs and mRNA [62]. At around a general length of 22 nucleotides, miRNAs are powerful post-transcriptional regulators of gene expression. Typically, miRNAs interact with the 3′ untranslated region (UTR) of mRNA to suppress translation of target genes, resulting in downregulation of proteins [63]. Often, miRNAs within EVs target genes with tumour-suppressive functions, promoting cancer progression. Furthermore, the mRNAs contained within EVs can be translated in target cells [64]. Less research has been carried out into the lipid cargo of EVs, however sphingomyelin, glycosphingolipids and phosphatidylserine are enriched in EVs [65]. Prostaglandins have also been detected in EVs and shown to contribute to signalling in recipient cells [66].

Changes in EV cargo during oncogenesis enable BC cells and cells of the TME to convey powerful chemical messages to surrounding cells, enhancing tumour progression and they could be used to monitor, detect and classify BC. The isolation of EVs from biofluids including, plasma, serum, urine and ascites, and the analysis of their cargo is emerging as a potential diagnostic and prognostic tool, allowing for the early detection and post-treatment surveillance of BC patients [67]. Due to their stability in biological fluids and their ability to protect and maintain the integrity of their content, preventing degradation and enabling its further study, tumour-associated EVs are currently deeply investigated through several omic techniques in order to identify novel and specific biomarkers that reflect the biological landscape of BC, the state of the tumour, disease progression and the response to cancer treatments. Specific examples of the implementation and clinical implication of these techniques are given later on in the review, highlighting how liquid biopsy, based on the non-invasive sampling and analysis of easily accessible non-solid biological tissue is a promising approach for the detection of good diagnostic, prognostic and therapeutic BC biomarkers [67].