THE INS-AND-OUTS OF EXOSOME BIOGENESIS, SECRETION, AND INTERNALIZATION

Abstract

Exosomes are specialized cargo delivery vesicles secreted from cells by fusion of multivesicular bodies (MVBs) with the plasma membrane (PM). While the function of exosomes during physiological and pathological events has been extensively reported, there remains a lack of understanding of the mechanisms that regulate exosome biogenesis, secretion, and internalization. Recent technological and methodological advances now provide details about MVB/exosome structure as well as the pathways of exosome biogenesis, secretion, and uptake. This review aims to outline our current understanding of these processes and to highlight questions that remain to be answered based on recent discoveries in the field.

EXOSOME: THE CARGO DELIVERY VEHICLE

Autocrine and paracrine intercellular communication is key to regulate the functionality of a tissue or organ and ultimately the organism. The intracellular trafficking pathways such as the ER-Golgi secretory pathway and secretory autophagy (see glossary) govern short-range intercellular signaling by facilitating the secretion of intracellular proteins, such as cytokines and growth factors that diffuse into the external environment [1]. To provide signal specificity and targeted delivery, cells package specific proteins, lipids, RNA, and DNA in 40-180 nm membrane-bound vesicles called exosomes [2]. Exosomes are a subclass of small Extracellular Vesicles (sEVs) that also comprise microvesicles (50-1000 nm) and small ectosomes (30-150 nm), which are generated by the outward budding of the PM. In contrast, exosomes originate intracellularly via inward membrane budding and vesiculation of large, 800-2000 nm vesicles named multivesicular bodies (MVBs) [2,3]. The fusion of these MVBs with the PM leads to the release of the intraluminal vesicles (ILVs) as exosomes [4]. The cell type and intracellular origin of MVBs influence the type of cargo present inside exosomes and on their surface. Diversity in exofacial (see glossary) proteins is required to establish the specificity of intercellular communication. In this context, numerous physiological [5,6], pathological [7-11], and therapeutic conditions [12,13] have been linked to changes in the composition and quantity of secreted exosomes. The focus of this review is to highlight recently described mechanisms involved in MVB biogenesis, exosome secretion, and uptake of exosomes by recipient cells. Table 1 provides an exhaustive list of components involved in the regulation of MVB biogenesis and exosome secretion, and Table 2 describes the tools used to study the mechanisms regulating these processes.

TABLE 1.

LIST OF COMPONENTS INVOLVED IN THE REGULATION OF ILV BIOGENESIS AND CARGO SORTING

Component	Description	Role	Ref.
Caveolae	A 50-80 nm cup-shaped structure that is formed during PM endocytosis facilitated by caveolin 1.	Mechanoadaptaion, signal transduction, cholesterol homeostasis, and cargo sorting in ILVs.	[14,39]
Clathrin	A self-assembling protein that polymerizes into a polyhedral lattice and facilitates membrane curvatures.	Facilitates molecular crowding required for the initiation of ILV biogenesis.	[18,22]
Ceramide	A negative curvature generating lipid formed by nSMase-mediated sphingomyelin hydrolysis.	Cargo sorting, ILV formation, and MVB maturation.	[21,41,88]
Lipid rafts	Dynamic lipid-ordered microdomains enriched in cholesterol and sphingolipids that influence membrane fluidity and membrane protein trafficking.	ILV Biogenesis, cargo sorting, and ILV retrofusion. Exosome internalization	[12,63,104]
Dynamin	A multimeric GTPase that regulates membrane fission.	MDV biogenesis, and endocytosis.	[5]
Neutral Sphingomyelinase	A hydrolase involved in sphingomyelin catalysis to produce ceramide.	ILV biogenesis, cargo sorting, and ER-endosome MCS formation.	[40,43,45,59,71,88]
RalA/RalB	Small GTPases crucial for exocytosis and tumorigenesis. Localized on CD63-positive endosomes/MVBs and PM. Facilitates the recruitment of PLD1 on endosomes, thereby regulating the synthesis of the positive curvature generating phospholipid phosphatidic acid from phosphocholine.	ILV biogenesis, cargo sorting and exosome secretion.	[9]
Rab 31	A small GTPase involved in exocytic trafficking.	ESCRT-independent cargo sorting, MVB docking on PM.	[6,26]
CHMP 4B/7	Components of the ESCRT III complex involved in membrane budding and fission.	Regulate viral egress, ILV biogenesis and tetraspanin sorting in ILVs.	[28,47,117]
ALIX	An ESCRT scaffolding protein that recruits the ESCRT III family proteins CHMP3/4/1 and VPS4B. ALIX localize to late endosomes in an LBPA-dependent manner and recognizes LYPX(n)L (late domain-like motif) in syntenin forming a complex that promotes ILV budding.	With ESCRT III, facilitates selective retention of ubiquitinated cargo, mostly tetraspanin. Also facilitates ILV biogenesis. Controls exosome secretion by regulating lysosome functions.	[21,28,110,113,117]
TSG101	A component of ESCRT I associated with tumor proliferation and metastasis.	Induces molecular crowding to initiate ILV biogenesis. Recognizes PSAP motif on Gal3 for ILV cargo sorting.	[86,110]
GPR143	An atypical GPCR involved in melanosome biogenesis, organization and transport.	Recruits the ESCRT-0 subunit Hrs and promotes its association with EGFR and integrins to regulate their sorting in ILVs/exosomes.	[10]
CERT	Also known as STARD11, dictates the ratio of ceramide to sphingomyelin by transferring ceramide from the ER to the Golgi.	Facilitates ER-endosome MCS by interacting with VAP-A. Promotes ILV biogenesis and RNA sorting in ILVs.	[40,41]
VAP	An ER-resident protein involved in membrane contact formation by facilitating protein complex assembly.	Anchors FFAT-containing proteins like CERT to the ER during ILV biogenesis.	[40,41]
IDR-condensates	Membrane-less, microscale compartments in eukaryotes formed by the condensation of proteins containing intrinsically disordered regions (IDR).	Spatially regulates the sorting of specific miRNAs in ILVs.	[118]
Exocyst	Octameric protein complex that tethers secretory vesicles to the PM facilitating SNARE-mediated fusion.	Present in LAMP2A-enriched exosomes. Regulates exosome secretion by interacting with Rab11 and facilitating MVB-PM fusion.	[6,77]
V-ATPase	An ATP-driven proton pump controlling intra- and extra-cellular pH.	Regulates ILV cargo sorting with the membrane fusion machinery and exosome secretion.	[119,120]
LAMP2A,	A heavily glycosylated type-1 lysosomal membrane protein that regulates autophagy.	Regulates selective cargo sorting in ILVs.	[6]
KFERQ-like sequence	Signal sequence for endosomal micro autophagy.	Identifies exosomal cargo for selective sorting by LAMP2A.	[6]
LAPTM4B	An oncogenic protein present on late endosomes and lysosomes that binds ceramide and promotes its endosomal escape.	Facilitates exogenous ceramide induced MVB transport to PM thereby increasing exosome secretion.	[86]
LC3	A ubiquitin-like modifier involved in membrane growth during autophagosome maturation.	Regulates MVB-acidification. Controls the loading of RNA and RNA-binding proteins in ILVs	[51]
IST	An ESCRT III component involved in cytokinesis and nuclear envelope reformation	Regulates the cargo sorting of GPRC5B	[30]
EHD family proteins	Proteins involved in endosome sorting	Regulate ILV biogenesis and the cargo sorting of GPRC5B. Present in small EVs secreted from apoptotic T-cells	[30,120]
CD2AP	Facilitates endocytosis by acting as an adapter between membrane proteins and the actin cytoskeleton	Controls the sorting of GPRC5B	[30]
Actin remodeling machinery	A group of proteins that control actin polymerization and disassembly, along with the structural organization of branched actin. Includes components such as, Cortactin, Filamin and Arp2/3.	Regulate the stability of cortical actin required for MVB docking at the PM and exosome secretion. Promotes the formation and maturation of MVBs involved in apoptotic exosome-like vesicle biogenesis.	[52,73,74,119]

TABLE 2.

LIST OF TOOLS AND TECHNIQUES USED TO STUDY EXOSOME BIOLOGY

Isolation and purification
Method	Principle	Merit	Pitfall	Ref.
Differential ultra-centrifugation	Stepwise increase in the centrifugal force leads to the separation of components based on their density.	No additional reagents required other than bench top, and an ultracentrifuge	Very low “purity” of isolated exosomal fraction	[80,113]
Density-gradient ultra-centrifugation	High centrifugal forces lead to the separation of particles in a gradient of some viscous solution based on their relative density, with the denser ones settling to the bottom of tube.	Efficient removal of non-exosomal fractions. Adjustable resolution of particle separation.	Cannot distinguish between exosomes with similar density but with distinct origin/molecular composition
Top loading	Loading exosome suspension on the top of density gradients (1.02-1.226 g/ml).	Easy gradient establishment	Inefficient resolution	[40,45]
Bottom loading	Loading exosome suspension at the bottom of density gradient (1.06-1.32 g/ml).	Better resolution	Requires very high speed (200,000 X g)	[65]
High resolution	Exosome suspension loaded at the bottom of i density gradient (12-36 %).	Best resolution	Fails to resolve molecular heterogeneity	[113]
ExtraPEG	Precipitation of smaller particles at low centrifugation speeds by increasing the particle size using PEG-based molecular crowding.	Fast and efficient with robust increase in exosome yield	Still requires other exosome purification procedures for downstream applications	[45,121]
Size exclusion chromatography	Size-based separation of particles by filtration through a gel.	Minimal alteration of exosomal corona	Very low yield	[65]
Antibody-conjugated agarose beads	Separation of particles based on the binding of antigens present on the exosomal surface to an antibody immobilized on agarose/magnetic beads.	Provides molecular specificity for exosome isolation	Depends on the identification of surface antigens, and on the availability of antibodies. Difficult to separate antibodies from exosomes.	[45,113]
Sorting and Analysis
Nanosight	Based on the characteristic movement of particles in solution according to Brownian motion, the camera documents the scatter light upon laser illumination of particles in a defined volume. Particle movement is subsequently calculated using nanoparticle tracking analysis (NTA).	Can assess the changes in exofacial cargo by measuring hydrodynamic radii. Can accurately assess particle size.	Cannot distinguish between membranous and non-membranous particles of same size. Not ideal for concentration measurements.	[45,113,121,122]
Zetaview	Tracks the in-flow micro-electrophoretic migration of particles based on surface charge (zeta potential) of the particle. Tracks are quantified by NTA.	Superior for reliable concentration measurements of serially diluted particles. Can detect the impact of freeze-thaw.	Fails to accurately measure the particle size.	[122]
Interferometric plasmonic imaging (SP-IRIS)	Based on interferometric scattering light from the adsorption of single exosomes on thin gold films in real-time.	Circumvents the problem of scattering sensitivity limit past sixth power of object diameter, noted for NTA. Enables imaging of molecular motions in single exosomes. Allows to distinguish multiple types of exosomes by proteomic profile.	Limited by the distribution density of proteins on the exosome surface.	[123]
STORM imaging	Stochastic optical reconstruction microscopy of exosomes in 3D using super resolution microscopy.	Provides information on the distribution of multiple proteins on specific microdomains within single exosomes.	Requires specialized equipment and trained personnel	[124]
Biotinylation-based mass spectrometry coupled with proteinase K (PK) degradation of surface proteins	By comparing the peptides found in samples of exosomes between PK treated and control samples and then cross-referencing this with biotinylated surface proteins one can determine the presence and deduce the topology of exofacial proteins	PK treatment provides information on the proteome of exosomal cargo and their location as well as their topological orientation	PK treatment may damage membrane proteins to allow endogenous proteins to degrade them and obscure their presence	[125]
Secretion
TIRFM	Total Internal Reflection Fluorescence Microscopy enables high resolution imaging of thin (<200 nm) specimens close to high refractive index imaging surfaces.	Enables visualization of exosome secretion in real time.	Requires fluorescent tagging of exosomal cargo	[70]
Hyperspectral imaging of exosome microarray	Microscopic hyperspectral imaging of exosomes based on rapid binding of membrane proteins on a corresponding array of antibodies printed on photonic crystal biosensor	Requires very low sample volume and can efficiently generate molecular fingerprint of exosomes secreted by different cell types	No details about size. Fails to assess exosome internal cargo. Requires prior information of exosome proteome.	[126]
FLIM imaging of exosome membrane viscosity	Uses fluorescence lifetime imaging of BODIPY-tagged benzene-based molecular rotors sensitive to membrane viscosity that correlates with exosome size	Can measure exosome size differences during pathogenicity	Might yield erroneous information in the case of changes in exosome membrane lipid composition	[127]
DRCA single particle fluorescence	Selective dual rolling circle amplification-based encapsulation of single cell-derived exosomes for flow cytometric or microscopic analysis	Can distinguish the exosomes from two different origin from a single sample	Requires the information of cell type specific exosome markers	[128]
Internalization
pHlurion (M153R)-CD63-Fluorescent tag	Use of brighter pH-sensitive fluorophore, pHlurion enables the appearance of green signal upon fusion of CD63-positive MVB with the PM, followed by subsequent loss of pHlurion signal upon exosome internalization in acidic endosomes. mCherry signal is not pH-sensitive and allows for the tracking of CD63 MVBs and exosomes over time.	Brighter fluorophore, real-time imaging of exosome secretion and internalization	Restricts visualization of exosome trafficking only to the MVBs that have acquired acidic pH.	[4,71]
Aptamer-based approach	DNA-based fluorescent tagging of exosomes for subsequent analysis of internalization pathways in cells by flow cytometry or microscopy	Can easily distinguish between exosome-PM fusion or exosome endocytic pathway of exosomal cargo release	Expensive and tedious	[108]
Proximity barcoding assay	Profiling of surface proteins of individual exosomes using antibody-DNA conjugates and next gen sequencing	Useful for identifying exosomal heterogeneity during pathogenesis	Resource intensive approach	[129]
GFP-nanobody based donor-acceptor pair	CLEM based identification of intracellular exosomal cargo release site by GFP-tagging membrane or soluble cargo and expressing GFP nanobody in acceptor cells	Tracking exosome internalization and cargo release	Requires genetic manipulation of cell lines	[107]
MVB Ultrastructure
CLEM	Correlative Light and Electron Microscopy obtains and correlates high-resolution transmission electron microcopy with images visualized using fluorescent-tagged proteins.	Enables the colocalization of protein (fluorescence) with membranes (electron microscopy) structures within MVB and at PM during exosome secretion	Can introduce fixation induced structural artefacts	[49,71]
Expansion microscopy	Biochemical isotopic expansion (4x or 10X) of cells to visualize the distribution, composition, and size of ILVs within the MVBs.	Provides intracellular details of exosome biogenesis avoiding inter cellular and intraorganellar heterogeneity	Requires cells to be fixed thus limits real time imaging	[45,49]

MVB BIOGENESIS

Canonical pathway

The canonical exosome biogenesis pathway is initiated at the PM, where the distinctive vesicle-within-a-vesicle architecture of MVBs is generated by two-step membrane budding process. In the first step, the initial membrane curvature required for PM deformation and endocytosis (see glossary and Figure 1) initiation is facilitated by multimeric proteins such as caveolae [14] (Table 1) and clathrin (Table 1), along with the asymmetric distribution of lipids such as ceramide and cholesterol [15]. This PM bud then undergoes dynamin [16] (Table 1) and/or ESCRT [17] (see glossary) dependent fission to produce a new vesicle. The second step involves the formation of ILVs, where the limiting membrane of the endosome invaginates to create numerous nanoscale sized ILVs within a large vesicle now referred to as an MVB [2,18,19]. This step can either be ESCRT-dependent [17-20] or ESCRT-independent. The ESCRT-independent pathway has been reported to be facilitated through the generation of (i) ceramide via neutral sphingomyelinase 2 (nSMase2) activation (Table 1) [21-24], (ii) phosphatidic acid via phospholipase D2 activation [25], or (iii) lipid-raft formation in a Rab31-flotillin dependent manner [26]. A recent study shows that ILVs are generated one at a time rather than simultaneously with a wave of periodic, concerted, and clathrin-dependent ESCRT recruitment on endosomes [20]. The recruitment of clathrin and ESCRT on developing MVBs however, is independent of endosomal maturation [20].

Figure 1. Heterogeneity of MVB Biogenesis.

(A) Canonical multivesicular body (MVB) biogenesis is initiated at the plasma membrane (PM) through endocytosis and the generation of an early endosome structure, which matures to a late endosome and then a CD63-positive MVB before eventually fusing with the PM and releasing intraluminal vesicles (ILVs) as CD63-enriched exosomes. The zoomed inset 1 highlights the budding and endosomal sorting complexes required for transport- (ESCRT) (blue spiral) mediated scission of the limiting membrane of MVBs. This leads to the formation of ILVs, which exist as distinct subpopulations (represented by different colors) of either single or tethered ILV clusters within the MVB. At the bottom of the inset, the retrofusion (large green ILV) of dynamic ILVs with the MVB limiting membrane is depicted. In zoomed inset 2, the loading of miRNAs and RNA-induced silencing complex (RISC) through endoplasmic reticulum (ER)-MVB membrane contact sites (MCS) formed during maturation of late endosome to MVB is highlighted. The thick purple and pink lines on ER and endosome membranes represent membrane microdomains enriched with ceramide (purple) and cholesterol (pink).

The non-canonical MVB biogenesis pathways are illustrated as:

(B) the biogenesis of Rab11-enriched exosomes, depleted of canonical exosomes markers, originating from recycling endosomes during glutamine deprivation;

(C) the fusion of a LC3-II positive autophagosome with a canonical MVB, leading to the formation of an amphisome that culminates into a hybrid structure containing both canonical ILVs and LC3-II-positive ILVs enclosed by the inner-membrane of autophagosomes that are then secreted as LC3-II-positive exosomes;

(D) proteasome-mediated biogenesis of apoptotic-extracellular vesicles (AEVs) within autolysosomes that are secreted in a caspase 3-dependent manner from early apoptotic cells;

(E) budding of mitochondria into mitochondria-derived vesicles (MDVs) and their fusion with canonical MVBs, followed by secretion of exosomes containing mitochondrial proteins and mitochondrial DNA (mtDNA);

(F) the outward budding of the inner nuclear envelope in activated neutrophils leads to the biogenesis of nuclear envelope-derived MVBs (NE-MVB) followed by secretion of NE-derived leukotriene B₄ (LTB₄)-containing exosomes.

The deformation of the MVB limiting membrane required for ILV biogenesis is thermodynamically unfavorable. To overcome this energy barrier, lipid segregation and protein molecular crowding are necessary. As a result, the membrane deforms continuously rather than abruptly [18]. This initial stage of ILV biogenesis is therefore proposed to be a passive process facilitated by the free diffusion of ESCRT 0-III proteins and transmembrane cargo. The increase in limiting membrane curvature after the initial deformation facilitates the assembly of the ESCRT III helical spiral [18] (Figure 1, inset 1). The tension generated by the helical spiral provides relaxation forces to drive the fusion of membranes on the neck of mature buds. This results in the fission of ILVs from the limiting membrane into the lumen of the MVB [27] . Furthermore, the endosomal recycling components ALIX and EHD1 (Table 1) facilitate the nanoscale curvature required for ILV biogenesis [21,28]. Notably, ALIX also interacts with the ESCRT III protein CHMP4B (Table 1), which forms a multimeric spiral within the neck of the developing ILV. The ATPase activity of VPS4A then induces a conformational change in CHMP4B spirals leading to the scission of ILVs from the limiting membrane of MVBs [18,28,29].

Along with providing the V-ATPases (Table 1), lysosomal hydrolases, and small GTPases required for endosomal maturation, the trans-Golgi network (TGN, see glossary) also supplies ESCRT components for cargo sorting during ILV biogenesis [19]. Additionally, Golgi-dependent MVB formation is controlled at the TGN-endosome trafficking level by selective cargo capture in the Golgi [30], and at the endoplasmic reticulum (ER)-Golgi trafficking level [31] possibly mediated by selective cargo modifications in the Golgi. For instance, the EHD1-retromer (see glossary) interaction controls the exosomal packaging of several PM proteins, including M6PR, Wntless, EphB4, LDLR, and GPRC5B, which enter the Golgi via PM-TGN trafficking to facilitate receptor replenishment on the PM [32-34]. The Golgi apparatus facilitates MVB cargo sorting through post translational modifications of ILV cargo, including phosphorylation, ubiquitination, glycosylation, and acetylation, making it an integral component of ILV biogenesis and secretion [33,35].

As endosomes mature into MVBs they interact with the ER, where mature MVBs are sorted to one of the two destinations: lysosomes or the PM. Indeed, although MVBs were initially identified as a "means-to-an-end" for the targeted uptake and degradation of the epidermal growth factor receptor (EGFR) [36], evidence points to the use of MVBs in both degradative and secretory functions. The dynamic physical interaction of the ER with many organelles, including MVBs, is crucial for cellular homeostasis [37]. Furthermore, the enrichment of the cholesterol-sensing protein Cav1 at ER-endosome membrane contact sites (MCS, Figure 1, inset 2) promotes ILV formation and cargo sorting during cancer metastasis [38,39]. Evidently disruption of these MCS leads to the accumulation of cargo in the ER followed by its degradation through the proteo-lysosomal pathway [39]. This pathway can be defective in neurons leading to pathological accumulation of misfolded prion protein aggregates [8]. The total amount of cellular cholesterol and the relative levels of cholesterol between organelles control the formation of ER-endosome MCS. These MCS promote the recruitment of several membrane-binding and cholesterol-sensing adaptor proteins that are necessary for MVB formation (Table 1) [39-42]. In addition to cargo sorting and MVB trafficking, ER-endosome MCS have been reported to regulate a switch of MVBs from a secretory to a degradative fate [43] (see below and Figure 2).

Figure 2. Exosome secretion, uptake, and release of exosomal cargo.

(A) Cartoon illustrating the requirement of endoplasmic reticulum (ER)-multivesicular body (MVB)/late endosome membrane contact sites (MCS) for regulating the secretory fate of an otherwise degradative MVB by switching the small GTPase Rab7 for Arl8b, which allows for kinesin-mediated peripheral trafficking of the MVB. A second GTPase switch from Arl8b to Rab27 facilitates the fusion of a secretory MVB with the plasma membrane (PM), allowing exosome secretion by the donor cell (left). The zoomed inset in (B) illustrates the extracellular release of exosomal cargo either through the action of exosomal transmembrane protein transporters (gray ovals) or by the action of putative extracellular lipases on the exosomal membrane represented by a dashed line.

The uptake of exosomes by the recipient cell (right) happens non-specifically by (C) micro/macropinocytosis, which usually culminates in the degradation of exosomal cargo upon fusion of exosome-containing endosome with lysosomes, or selectively by (D) clathrin (blue T-shaped structures) or caveolae (blue circles) and receptor/HSPG-mediated endocytosis. Exosomes internalization can be promoted by filopodia-mediated exosome capture. (E) Exosomes can bind to specific receptors on the PM to facilitate a signaling response in target cells. (F) Exosomes can directly fuse with the PM and release their internal cargo into the cytosol of the recipient cell. (G) Following endocytosis, the exosome-containing endosome can either (1) deliver exosomal cargo to specific organelles such as the ER, (2) release the exosomal internal cargo by fusion of the exosome membrane with the endosomal membrane, (3) release intact exosomes upon direct membrane permeabilization of an endo-lysosome that contains internalized exosomes, (4) be trafficked back to the PM for re-secretion of internalized “functional” exosomes from the recipient cell, or (5) fuse with a lysosome, permeabilizing the membrane of endocytosed exosomes, and releasing the soluble cargo of exosomes from within these endolysosomal structures, outside the recipient cell by lysosome exocytosis.

While the tetraspanins (see glossary) CD63, CD81 and CD9 are present on sEVs including exosomes [2,3], their localization on microvesicles and small ectosomes appears to be stochastic [44] (Box 1). However, for these proteins to be secreted in exosomes the neutralization of endosomal pH is required [44]. Nevertheless, in most cases, CD63-positive exosomes constitute the major fraction of secreted exosomes, including exosomes derived from non-canonical pathways [5,45] (see below and Figure 1).

BOX 1. EXOSOME HETEROGENEITY.

Cargo type defines the diversity of exosomes

Based on orientation, cargoes in ILVs/exosomes can be defined as: (i) internal, such as soluble proteins [110], micro-RNAs, ds/ssDNA, and free lipids packaged within the lumen of ILVs; (ii) membrane-associated, comprised of transmembrane proteins; [66] and (iii) exofacial, made up of the cargo associated with the outer surface/corona of the exosomes/ILVs [46,67,94,96]. Both intra-membranous and exofacial cargo are involved in directing internal ILV cargos to their targeted destination [111]. RNA and translation/signaling-associated proteins compose most of the exofacial cargo and require association with lipoproteins on the ILV surface. In contrast, the association of exofacial DNA fragments (see below) is facilitated by electrostatic interactions of positively charged histones with the negatively charged ILV membrane [111,112]. The exofacial cargo can be further subdivided into the innate corona and the acquired corona. The innate corona can either be assembled intracellularly by MVB-autophagosome fusion during amphisome (see glossary) biogenesis, or extracellularly via PM endocytosis that allows for the capture of extracellular milieu within the MVB lumen [51,111,113]. The acquired corona, however, is generated by non-specific interactions between the protein cargo and the surface of secreted exosomes. This accounts for differences observed in the acquired corona of exosomes from the same cell types in different extracellular environments such as blood plasma or cerebrospinal fluid [114]. Glycosylation of the exosomal surface during cancer can influence its zeta potential and consequently the exofacial cargo of exosomes from the same cell [13,115]. Secreted extracellular Amyloid β (Aβ) interacts with glycosphingolipids on the exosomal surface, adding to the acquired corona. However, intracellular Aβ is packaged in MVBs as an internal ILV cargo for secretion within exosomes [86]. The packaging of different types of cargo in the ILVs is a highly selective process regulated by a series of proteins listed in Table 1.

Stochastic cargo sorting contributes to exosomes diversity

Given the heterogeneity of cargo and the physical limitation that a vesicle of 50 nm can only hold up to 600 surface proteins of 4 nm each, an estimate of no more than 3000 different proteins can be present in an exosome [44]. Most of the cargo sorting machinery described in Table 1 presents a strong case for the selective sorting of exosomal cargo regulated by specific adapter proteins or lipid rearrangements. However, a recent study reported the secretion of phenotypically heterogeneous exosome populations from single cells [116] with no correlation between the abundance of different markers on exosomes, providing support for a stochastic sorting of membrane-associated cargo. Au contraire, another study used high resolution purification of exosomes (Table 2), to show selective sorting of cytosolic cargo [113]. There are a variety of stochastic parameters that could drive exosomal heterogeneity such as differences between cell types, temporal and treatment-specific variations in protein expression, variations during protein sorting at different organelles, and intrinsic nanoscale variations among exosomes [44]. On the other hand, the observation that the uniqueness of the exosomal proteome among different cell types is not associated with differential expression of those proteins in cells suggests an equally important role of selective cargo sorting [114]. It is possible that initial stochastic events control the recruitment of specific cargo sorting machineries, thereby providing selectivity to exosomal cargo. This can explain the spectrum of varied cargo abundance in exosomes secreted from individual cells cultured under identical conditions [116].

Specialized Non-canonical pathways

In addition to the canonical PM and endosome-derived exosomes, recent studies have reported that membrane budding from various organelles, including mitochondria and the nuclear envelope, can initiate non-canonical pathways for MVB biogenesis (Figure 1).

Recycling endosomes

In multiple cancer cells, inhibition of Akt/mTORC1 signaling (see glossary) induced by glutamine deprivation acts as an “exosome switch.” This results in the depletion of late-endosome-derived, CD63-enriched exosomes and the biogenesis of Rab11-positive exosomes that are devoid of prominent markers of canonical exosomes (Figure 1B). These specialized exosomes are enriched with amphiregulin, an EGFR ligand, which regulates ERK-MAPK signaling to promote tumor growth and adaptation during stress [46].

Secretory autophagy

During prostate cancer, cholesterol induces the increased recruitment of Cav1 to endosomes [38]. This, along with the concerted action of the Golgi-biosynthetic pathway and calcium-dependent secretory autophagy, leads to the biogenesis of relatively smaller, specialized Cav1-exosomes that lack conventional exosome markers [38]. Moreover, many enveloped viruses including Vaccinia [47], and Human Cytomegalovirus (HCMV) [48] hijack the ESCRT and tetraspanin cargo sorting machineries to facilitate viral secretion in exosomes. These viral exosomes are usually smaller and denser, compared to classical exosomes derived from uninfected cells. HCMV infection is also mediated via a Golgi-dependent early biosynthetic pathways in endothelial cells and leads to the release of mature viral particles via secretory autophagy. This increases infectivity compared to secretion via conventional CD63/LBPA- (see glossary) positive MVBs in fibroblasts, which have the potential to be degraded in lysosomes [48]. Additionally, SARS-CoV-2 escape from lysosomal degradation is accomplished through a partial release in exosomes via the secretory autophagy pathway [49,50]. Secretory autophagy also mediates the loading of RNA and RNA-processing enzymes in exosomes via LC3 (Table 1) dependent EV loading and secretion (LDELS) in an nSMase-dependent and ESCRT-independent manner [51] (Figure 1C).

Autolysosomes

Early apoptotic cells secrete apoptotic exosome-like vesicles (AEVs) in a caspase 3-dependent manner [52]. These are similar in size to classical exosomes but are relatively denser. In addition to the classical exosome markers syntenin, fibronectin, CD63, and CD9, the presence of LAMP1 and Rab7 in AEVs suggests a role for an endolysosomal trafficking-like route in AEV biogenesis. In contrast to apoptotic bodies (~1 μm size), AEVs lack the conventional apoptotic body markers, GM130, calreticulin and tubulin [52-54]. Evidently, in serum-starved endothelial cells, enlarged perinuclear autolysosomes facilitate AEV biogenesis [53] (Figure 1D). Sphingosine-1-phosphate (S1P) signaling enables AEV secretion from apoptotic HeLa cells by modifying cortical actin to trigger endocytosis and formation of CD63 and S1P receptor-positive endosomes. Moreover, G-protein-mediated Rho activation promotes the retrograde movement and maturation of these endosomes into MVBs [52]. Contrary to the anti-inflammatory function of apoptotic bodies, the overrepresentation of DAMPs (damage-associated molecular patterns) and proteasome (see glossary) machinery in AEVs point to their role in initiating inflammatory responses during apoptosis [52,54]. While the formation of AEVs is dependent on the proteasome [54], AEV exocytosis is caspase 3-dependent [52,53] (Figure 1D).

Mitochondria

During homeostasis, double membranous mitochondria-derived vesicles (MDVs - see glossary) fuse with canonical MVBs to release their internal vesicle as exosomes containing mitochondrial DNA (mtDNA), along with the release of canonical exosomes (Figure 1E). The dynamin-like protein, OPA1, and the dynamin-binding protein, Snx9, regulate the selective sorting of mtDNA and the IM/matrix proteins, TOM20 and mtHSP70, in MDVs [5]. While the mitochondria protein quality control protein PINK1 connects late endosomes and mitochondria to facilitate mtDNA loading into the MVB lumen [55], Parkin promotes the clearance of damaged mitochondria by regulating MDV-lysosome fusion [5]. Furthermore, ferroptosis-like conditions stimulate the Rab27-dependent secretion of mtDNA-containing exosomes from breast cancer cells [55]. Additionally, exofacial mtDNA promotes both the hormone therapy-resistant metastasis [56] and the antiviral responses in dendritic cells [57].

Nuclear envelope

To effectively infiltrate injured or infected tissues, neutrophils relay primary chemotactic signals to neighboring neutrophils by secreting the secondary chemoattractant leukotriene B₄ (LTB₄, see glossary) [58]. LTB₄ and its synthesizing enzymes are packaged in MVBs that release their ILVs as exosomes during neutrophil chemotaxis [59]. In contrast to canonical exosome biogenesis, the biogenesis of the LTB₄-containing MVBs originates at the nuclear envelop where the LTB₄-synthesis machinery is present [45]. On the inner nuclear membrane, nSMase-dependent ceramide generation leads to the clustering of the LTB₄ synthesizing enzymes, which culminates in membrane curvature and initiates nuclear membrane budding. Expansion microscopy and exosome enrichment (Table 2) data demonstrate that these nuclear envelope buds mature into nuclear envelope-derived MVBs (Figure 1F), containing ~200 nm sized ILVs that are released as exosomes in a CD63-independent fashion. The presence of ALIX (Table 1) in the nuclear envelope buds and nuclear envelope derived MVBs, suggests a role for the ESCRT machinery in these events [45].

REGULATION OF ILV BIOGENESIS

Given the diversity in MVB biogenesis pathways and the molecular machinery involved in ILV biogenesis (Table 1), the heterogeneity in ILVs size and density are reported in a context dependent manner, as described below. In addition, recent studies suggest a spatial and functional segregation of ILVs within a single MVB. Although the molecular mechanisms that regulate ILV size and functional diversity remain poorly defined, recent studies have revealed new and unexpected findings.

Dynamics of distinct ILV pools within MVBs

MVBs contain distinct pools of clustered ILVs (see Figure 1, inset 1) [60,61]. One minor fraction (11%) of ILVs is destined for release as exosomes and another pool contains cargo destined for degradation in lysosomes or cargo transfer to other organelles. This suggests the existence of distinct biochemical signatures among ILVs that are destined for either intracellular (lysosomal degradation) or extracellular (exosomal) cargo trafficking. Using a chemically tunable cell-based reporter system in human melanoma cells, Perrin and colleagues recently provided evidence for the presence of a pool of ILVs that dynamically fuse back to the limiting membrane of MVBs [61]. This process, referred to as of ILV retrofusion (see glossary), requires constitutive MVB acidification and is inhibited by IFITM3 (interferon-induced transmembrane protein 3) [61]. IFITM3 is known to inhibit viral release from MVBs by increasing cholesterol and LBPA levels within vesicular stomatitis virus (VSV)-positive ILVs [62]. The bulk (2/3^rd) of exosomes originate from the retrofusion-resistant pool of secretion competent ILVs. However, the dynamic pool of retrofusing ILVs can be diverted to a secretory fate and this pool can constitute the remaining fraction (1/3^rd) of exosomes secreted from the cell [61]. While the physiological significance of ILV retrofusion remains to be determined, it could support the membrane equilibrium of MVBs, which is required for maintaining the thermodynamic stability of MVBs [61]. Indeed, the limiting membrane volume of MVBs continually decreases during ILV biogenesis and increases during ILV retrofusion [61]. ILV retrofusion has also been reported to facilitate the delivery of ILV cargo (antigen) to the cytosol during antigen processing and facilitate MHC II (see glossary) antigen presentation [61,63,64].

A recent study reported that the binding of the housekeeping protein GAPDH (see glossary) to the surface of exosomes secreted from either mesenchymal stem or Hela cells can induce exosome clustering [12]. Similarly, the presence of tetherin (see glossary) on cholesterol rich ILVs has also been reported to generate clusters of ILVs and clusters of exosomes [22,60]. Since the length of GAPDH tethers between exosomes is close to the ones observed between ILVs [12,22,60] it has been suggested that GAPDH is present within ILVs to act as a tether protein. This discovery of ILV tethering [22,60] further strengthens the existence of spatially and functionally distinct ILV clusters within the MVBs, as demonstrated by Perrin and colleagues [61].

Cholesterol and ceramide define ILV size

In addition to controlling cargo sorting in ILVs, ER-endosomes MCS can regulate the content and size of exosomes [40]. On endosomes, Cav1 regulates the segregation of ILV populations as well as cargo sorting within ILVs in a cholesterol-dependent manner. Cav1 may act as a cholesterol rheostat on MVBs thereby restricting membrane flexibility and increasing ILV segregation. This is supported by the discovery that Cav1-deficient fibroblasts produce compact and dense cholesterol-rich exosomes This phenotype is exacerbated by the loss of the ER-endosome tethering complex ORP1L-VAPA (Table 1) [39]. Since annexins (see glossary) are enriched in exosomes derived from Cav1 KO fibroblasts [39], they may serve to compensate for the loss of Cav1 by promoting exosome secretion. Overall, the redundancy in cholesterol sensing at ER-endosome MCS is crucial for regulating ILV biogenesis during homeostasis and pathology.

Elevated levels of ALIX and Cav1 in the exosomes secreted by metastatic cancer cells correlates with an increase in exosomal cholesterol levels and a concomitant decrease in exosomal size from an average of 100 nm to ~50 nm [38,65]. A similar reduction in exosome size has been observed upon integrin β3 depletion in breast cancer cells, which is accompanied by an alteration in the cholesterol-binding exosome proteome [66]. Although, a similar increase in cholesterol-enriched smaller ILVs is observed upon Hrs-depletion in HeLa cells [22], no correlation among ILV number/size and MVB size has been established [18,22]. In contrast to CD63-enriched exosomes, nSMase-dependent, ceramide-rich exosomes are significantly larger [21,22,67]. Similarly, in migrating neutrophils, where nSMase-dependent, nuclear envelope-derived MVBs harbor relatively larger (200 nm) ILVs, the ILV number correlates with MVB size [45].

The mechanisms that regulate ILV/exosome size remain ill-defined and require further investigation. Notably, the assessment of exosome size and density is subjective to the type of method used for exosome enrichment, purification, and analysis (see Table 2). Overall, the phenotypic and functional diversity of exosomes being secreted by a cell type is dependent on multiple factors including but not limited to cholesterol levels, organelle contact sites, origin of cargo, and effector function of exosomes [5,13,42,43,45,52,65,68,69].

EXOSOME SECRETION

Once MVBs commit to the secretory pathway, exosome release requires two additional steps. First, MVBs dock to the PM, and second the MVB limiting membrane fuses with the PM. Using TIRFM and CLEM-based high-resolution imaging of cancer cells [70,71] (Table 2), distinct and relatively slower kinetics of MVB-PM fusion compared to that of soluble protein or membrane receptor exocytosis events, have been reported. These findings suggest that MVB-PM fusion is an active, multistep process. Although studies have shown that calcium is necessary for exosome secretion [70,72], Verweij and colleagues reported calcium-independent MVB-PM fusion events facilitated by SNARE (see glossary) activation and downstream of GPCR signaling using a pH-sensitive sensor for exosome release (Table 2) [71]. Regardless of the calcium requirement, small GTPases (for docking) and SNAREs (for membrane fusion) invariably regulate the fusion of MVBs with the PM. Furthermore, the secretion of exosomes is regulated by various components of intracellular trafficking as discussed below.

Cytoskeleton remodeling

The actin binding protein, cortactin, has been reported to control the number of MVBs moving towards cortical actin and to increase the number of PM-docking sites by stabilizing branched-actin structures [73]. Stable branched actin networks eventually facilitate the transfer of microtubule associated MVBs to cortical actin with the help of unconventional myosin. In addition, DAG-PKCδ signaling reportedly regulates MVB-MTOC proximity at immune synapses and facilitates cortical actin remodeling to regulate the polarized secretion of exosomes (Table 1) [74].

Recycling endosome machinery

Rab11 is a small GTPase present on perinuclear recycling endosomes that canonically regulates the replenishment of transmembrane receptors on the PM. Rab11 also promotes calcium-dependent MVB-PM fusion by regulating the docking/tethering of MVBs on the PM [75]. HSP90 is a Rab11 effector, and a well-known exosome marker that promotes exosome release by inducing membrane deformation required for MVB-PM fusion [76]. The exocyst (Table 1) subunit Sec10 interacts with Rab11 on MVBs and facilitates the assembly of the remaining exocyst subunits, Sec3 and Exo70 on the PM, thereby promoting MVB-PM fusion [77]. Another exocytosis regulator, Munc13-4, promotes MVB-PM docking by recruiting Rab11 on MVBs [70]. Rab35, a component of the endosome recycling pathway, also promotes MVB-PM docking and additionally induces MVB-PM fusion in a CD2AP-dependent manner (Table 1) [30,78]. The absence of the retromer complex during lysosome dysfunction in neuronal disorders prompts the switch of MVB trafficking from lysosomes to the PM, facilitating the clearance of misfolded proteins through exosomes [79]. Furthermore, deletion of Rab11/Rab4 reduces exosome secretion and causes cargo accumulation, suggesting a crucial role for the recycling endosome machinery in the switch from degradative to secretory MVBs [79]. A recent study of cancer cells grown on a stiff matrix reported that Akt-signaling mediates the activation of Rab8, a recycling/exocytic trafficking regulator, which in turn promotes the secretion of exosomes that regulate metastasis [11].

ER Membrane Contact Sites

Intracellular cholesterol levels regulate ER-endosome MCS by inducing conformational changes in oxysterol binding protein 1L (ORP1L) residing on late endosomes. This facilitates its interaction with the ER-resident adapter protein and cholesterol sensor, VAP (Table 1) [68]. At the same time, the presence of ORP1L at ER-late endosome MCS regulates the switch from Rab7a, which is late-endosomal and recruits dynein, to Arl8b, which is lysosomal and recruits kinesin [43]. This switch facilitates the movement of Arl8b-positive MVBs to the PM, and their transition from degradative to a secretory phenotype [43] (Figure 2A). While VAP regulates ER-secretory MVB MCS, MCS between the ER and degradative MVBs are facilitated by Annexin A1 [42]. During homeostatic conditions, ORP1L is abundant on secretory MVBs, with a scattered distribution on a few degradative MVBs. Cholesterol deprivation forces ORP1L enrichment specifically at degradative ER-MVB MCS [42].

Late-endocytic machinery

Rab27a and Rab27b, each regulate different stages of MVB maturation and exosome release [80]. Rab27b localizes to the perinuclear region and mediates the transfer of membrane from the TGN to MVBs. It also helps retain secretory MVBs at the cell periphery by regulating the transfer of MVBs from microtubules to cortical actin. Rab27a, on the other hand, localizes to peripheral CD63-positive MVBs. There, it prevents fusion of MVBs with each other or other vesicles, thereby promoting MVB fusion with the PM [80]. Rab27a also regulates cortical actin dynamics at MVB docking sites by inhibiting coronin 1b localization [73]. Counterintuitively, inhibiting mTORC1 in mouse fibroblasts caused an increase in the Rab27a-dependent release of exosomes. This phenotype can be explained by the ability of mTORC1 to regulate endo-lysosomal positioning [81]. KIBRA (see glossary) provides an additional checkpoint on MVB-PM fusion by preventing the polyubiquitination and subsequent proteasomal degradation of Rab27a [82]. In canine kidney epithelial cells, Rab27a/b mediate exosome secretion from the apical membrane while Rab39 controls exosome secretion from the basolateral membrane by modulating BORC (a multi-subunit complex that regulates lysosomal positioning) [83]. Furthermore, a significant fraction of CD63-positive MVBs is LAMP1-positive and acquires the small GTPase Arl8b to promote kinesin-mediated trafficking towards the PM. This peripheral movement of secretory MVBs is followed by a Arl8b-to-Rab27a switch, rendering these MVBs fusion competent [43] (Figure 2A).

Lysosomal machinery

The shift of MVBs from a degradative to a secretory fate is governed by several processes [84]. Lysosome membrane permeabilization during lipid overload in hepatocytes [7], and defective lysosome-MVB fusion by TRPML1 inactivation in podocytes [85] results in increased exosome secretion. Alteration of lysosome motility by the depletion of TRPML1 and the ceramide-binding protein LAPTM4B (Table 1), inhibits MVB-lysosome fusion and boosts exosome secretion [85,86]. Furthermore, it has been reported that tetraspanin 6 targets the syndecan-syntenin complex (see glossary) for lysosomal degradation, thereby inhibiting exosome secretion [87]. Finally, ESCRT proteins have been reported to be engaged in both ILV biogenesis and in lysosomal repair [17], and the unanticipated rise in the secretion of exosomes from ESCRT depleted canine kidney cells is most likely brought on by lysosomal escape of MVBs [21].

Lipid metabolism

nSMase2 activity depletes sphingomyelin/cholesterol microdomains in late endosomes that are required for the assembly and packaging of V-ATPase (Table 1) in ILVs, thereby preventing MVB acidification and increasing exosome release [60,88]. At the same time, nSMase2 depletion increases the colocalization of CD63-positive MVBs with LAMP1-positive compartments, which potentially represent non-degradative lysosomes required for the trafficking of secretory MVBs [43,88]. On the other hand, treatment with a known inhibitor of V-ATPase and endolysosomal cholesterol trafficking, bafilomycin A1 reduces PM cholesterol with a concomitant accumulation of cholesterol in ILVs, robustly increasing the exosome secretion [60]. In this context, these superfluous exosome clusters remain attached to the PM following exosome secretion in a tetherin-dependent manner [60].

Functional outcome of exosome secretion

The tethering of exosome clusters on the PM suggests that cells have evolved an internal mechanism to regulate the release of exosomes for either long range communication such as signal relay during chemotaxis [59,89], or for short range signaling such as MHC II presentation on immune synapses [63]. Furthermore, the polarized secretion of exosomes at the front of directionally migrating cancer cells has been reported to create integrin-rich exosomal tracks for both autocrine and paracrine migration, potentially creating migration tunnels by degrading collagen in vivo [4]. The secretion of cAMP- (see glossary) containing exosomes from the back of chemotaxing Dictyostelium discoidium cells generates “breadcrumb” trails for the efficient migration of these cells [89]. In the context of innate immunity, the secretion of LTB₄-containing exosomes from directionally migrating neutrophils is proposed to have a similar role [59]. Finally, several other studies show the polarized secretion of two structurally and biochemically distinct exosome sub-populations from the apical and distal ends of polarized epithelial cells and neutrophils [4,33,45].

EXOSOME UPTAKE AND CARGO RELEASE

Following exosome secretion, the exosomal cargo is often released into the extracellular milieu through various mechanisms, including transmembrane channel-mediated active translocation through the exosomal membrane [89] or exosome disruption in the extracellular space presumably by secreted lipases [90] (Figure 2B). In this context, the packaging and regulated release of exosomal cargo could provide a mechanism to protect labile cargo from harsh extracellular environments and allow effective intercellular communication.

Since one function of exosome secretion is targeted delivery of cargo to recipient cells, exosome uptake is often selective and therefore relies on receptor-mediated endocytosis, which in turn is facilitated by clathrin, caveolae, dynamin, and cholesterol (Figure 2D) [91,92]. In this context, actin-based cellular protrusions known as filopodia are known to act as “endocytic-hotspots” for exosome uptake. During this phenomenon, which was previously shown to play a role in viral entry, exosomes move along filopodia and retraction fibers, towards the cell body before being endocytosed [4,93] (Figure 2D). In this context, the exosome internalization time ranges from almost immediately [93] to as long as 20 minutes [4], depending on the length of filopodia. In addition to facilitating the signaling response in target cells by binding to specific receptors on the PM [94] (Figure 2E), the exofacial cargo (Box 1) provides specificity to receptor-mediated exosome internalization in various contexts. These include the capture of circulating exosomes by cardiac cells [95], the exosomal transport of iron to metastatic cells by GAPDH-containing exosomes [12], the transfer of anti-viral immunity from macrophages to hepatocytes [92], and the suppression of T-cells by cancer-derived exosomes [96]. Furthermore, exofacial cargo specifies CD9-mediated transfer of specific exosomes derived from cancer-associated fibroblasts to metastatic pancreatic cancer cells [97] as well as the internalization by macrophages of vimentin-containing exosomes secreted from colorectal cancer cells [98]. In line with this, proteinase K stripping of exofacial cargo ablates exosome internalization by lung epithelial cells, with no effect on non-specific bulk uptake of exosomes by peritoneal macrophages [91,99].

Since the bulk uptake of exosomes by micro/micropinocytosis is mostly non-specific, the internalized exosomal cargo by this route is usually degraded by fusion of exosome-containing endosomes with lysosomes (Figure 2D) [92,100,101]. The clearance of internalized non-specific exosomes might serve as a mechanism for the degradation of cellular antigens by immune cells in an immunologically ‘silent’ fashion [102]. Additionally, lipid raft- (Table 1) mediated endocytosis and HSPG- (see glossary) mediated internalization also facilitate bulk-uptake of exosomes. Although these are considered as inherently non-specific uptake pathways, the abundance of HSPG and lipid-rafts on some cells compared to others provides a degree of selectivity by biasing the exosome uptake in favor of these cells [66,103,104]. This can provide a functional advantage to these recipient cells.

Additionally, the release of exosomal α-synuclein cargo into the recipient cell appears to be unaffected by the inhibition of either clathrin/caveolae-dependent or bulk uptake pathways. This suggests that these exosomes do not internalize, rather they fuse with the PM to release their internal cargo in the cytoplasm [105] (Figure 2F).

Once the exosomal cargo are internalized inside cells, they are often delivered to specific compartments, such as the mitochondria, ER, PM or cytosol, of recipient cells [93,106]. Recent studies show that internalized exosomes follow ER trajectories in a stop-and-go pattern. These trajectories resemble ER-endosome MCS (Figure 2G-1) [93], where RISC (see glossary) loading, which can boost the silencing efficacy of miRNA (see glossary), has been reported to happen [93]. However, the mechanisms that mediate the release of exosomal cargo at ER-endosome contact sites remain to be determined. Furthermore, the cargo of internalized exosomes can be released into the cytoplasm of recipient cells or back in the extracellular environment by one of the many ways recently described using GFP nanobody based donor-acceptor pair technique [107], and other microscopy-based approaches [4,71,108] (Table 2). For the cytoplasmic release of the internal cargo of exosomes (Box 1), the membrane of internalized exosomes can fuse with the membrane of exosome-containing endosomes (Figure 2G-2). However, the fusion of internalized exosomes with the endosome membrane occurs only after early endocytic vesicles acquire either Rab7 or CD63 [4,92]. These fusion events are often prompted by low pH-induced conformation changes in the exosome membrane proteins, along with the acquisition of cationic lipids such as cholesterol and LBPA by endosomes [92]. This fusion can also transfer intra-membranous exosomal cargo (Box 1) to the endosomal membrane. Alternatively, the release of intact exosomes in the cytoplasm (Figure 2G-3) has also been shown using dual fluorescent tagging system (Table 2) [107]. Functionally, the permeabilization of exosome-containing endo-lysosomes can lead to the escape of tau seeds from exosomes into the cytosol causing neuropathology [109]. However, using size-exclusion chromatography and proteinase K treatment (Table 2) of dual-tagged exosomes, it has been reported that not all internalized exosomes fuse with endosome membranes. Indeed, these authors found that some of these exosomes with fully functional cargo are re-secreted (Figure 2G-4) [106] from the recipient cells following bulk uptake. Furthermore, the extracellular release of specialized exosomal cargo such as IL1β that mature in lysosomes, requires exosome uptake, followed by the fusion of internalized exosomes with lysosomes and subsequent release during lysosomal exocytosis (Figure 2G-5) [110].

CONCLUDING REMARKS

Over the past four decades there has been a growing appreciation of the role of exosomes in many aspects of biology. There is also a great deal of interest in employing exosomes as a cargo delivery system for therapies with high selectivity. The novel concept of stochastic cargo sorting along with dynamic processes like ILV-retrofusion and ILVs/exosomes tethering, add to the existing heterogeneity in exosome biogenesis. Current dogma states that the secretory or degradative fate of MVBs is determined by ILV cargo. In contrast, it was recently discovered that separate pools of ILVs with secretory or degradative fate can be present within a single MVB. ER-MVB MCS have been identified to regulate the fate of MVBs. Similarly, ER-endosome contact sites are postulated to facilitate the release of exosomal miRNA in acceptor cells.

Distinct extracellular cues can prompt non-canonical exosome biogenesis originating from mitochondria, autophagosomes, and the nuclear envelope in diverse cell types. Notably, the machinery and molecular organization required for specialized MVB formation are much the same as what is required for canonical exosome biogenesis. Overall, the redundancy of the MVB biogenesis machinery underlies the elegance of the exosome biogenesis system. This redundancy proves to be essential for generating a diversified and dynamic, yet selective exosome profile under homeostatic and pathological conditions. The recent discoveries discussed in this review open Pandora’s box of questions (see Outstanding Questions Box). The progress in the development of the tools and techniques (Table 2) required to explore these questions will enable the use of exosomes for the efficient delivery and uptake of biologically relevant cargo.

OUTSTANDING QUESTIONS BOX.

Q. Is the heterogeneity of the exosomal cargo content achieved through selective sorting or is it stochastic? What are the factors that govern the apparent stochasticity? Identifying and correlating unique exosome proteomes with pathological conditions is critical for diagnostics, but the possibility of a stochastic mechanism for exosomal heterogeneity complicates therapeutic intervention.

Q. What elements affect whether synthesized MVBs are secreted or degraded? Is it the cargo type/modifications, the levels of cholesterol/ceramide, the intra/extracellular pH gradients, the levels of nutrients, lysosome physiology or some combination of these factors?

Q. Is the spatial segregation of tethered ILV clusters, which drives selective cargo transfer to diverse organelles, enabled by a MVB kiss-and-run (see glossary) mechanism? Are these ILV clusters functionally distinct? Does the pool of ILVs that can retrofuse back to the limiting membrane play a role in kiss-and-run by supplying the machinery required for contact site persistence and fusion? Is ILV retrofusion physiologically or pathologically relevant?

Q. Do the size and membrane lipid composition of exosomes define their internalization pathway?

Q. The identification of exosome re-release raises several questions, including whether exosome composition dictates the fate of internalized exosomes, whether the macropinocytic recycling route is implicated, whether exosomes recharge in acceptor cells prior to re-release, and, most importantly, what are the biological ramifications of exosomal re-release?

Q. The development of high-resolution real-time imaging tools to study exosome biogenesis, secretion and uptake can help answer several puzzling questions. For example, is there a preference between various routes of exosome uptake and subsequent cargo release and are there universal or specific signals for cargo sorting during ILV biogenesis? How do cells determine which type of exosomes to secrete?

HIGHLIGHTS BOX.

• Specialized, non-canonical exosomes originate from the nuclear envelope, mitochondria, autolysosomes, and autophagosomes.

• During canonical MVB biogenesis, wave-like, repetitive, and coordinated recruitment of clathrin and ESCRT enable the production of a single ILV at a time.

• ILVs exist within MVBs as separate populations of secretory and degradative pools, as well as a distinct pool that stays in dynamic equilibrium with the limiting membrane of MVBs through ILV retrofusion.

• Cholesterol and ceramide levels in the PM, ER, and endosomal membranes influence both the content and size of exosomes.

• Exosomes are secreted both as individual structures that diffuse into the extracellular environment as well as in clusters that remain attached to the PM of donor cells.

• Uptake of exosomes is facilitated by filopodia, bulk micropinocytosis, and by receptor - mediated endocytosis.

• ER-endosome contact sites facilitate both the packaging of ILV cargo during MVB biogenesis and the unpackaging of exosomes following their uptake in acceptor cells.

• The bulk of the exosome cargo is exofacial, which determines the selectivity of exosome uptake.

GLOSSARY BOX

Amphisome

Hybrid organelles generated upon fusion of autophagosomes with endosomes. They are required for lysosomal degradation of cargo captured by autophagosomes.

Annexins

A class of calcium-dependent, cholesterol-sensing proteins with the ability to generate lipid-ordered membrane microdomains.

cAMP

Cyclic adenosine 3’,5’-monophosphate is a small hydrophilic molecule widely known as an intracellular second messenger that plays a role in signal transduction. It also acts as the main chemoattractant that mediates the chemotaxis and development of the social amoebae Dictyostelium discoideum.

Endocytosis

The process by which PM invagination forms an intracellular membrane-bound vesicle containing extracellular material.

ESCRT

Endosomal Sorting Complexes Required for Transport that function during membrane bending and scission processes.

Exofacial

Present on or associated with the external surface of an exosome.

GAPDH

GlycerAldehyde 3-Phosphate DeHydrogenase, is a glycolytic enzyme that is also involved in DNA repair, RNA binding, membrane fusion, and cytoskeletal dynamics.

HSPG

Heparin Sulfate ProteoGlycans such as syndecan and CD44 are expressed on the PM and serve as endocytosis receptors for growth factors and chemokines.

KIBRA

An elusive protein expressed in Kidney and BRAin associated with cognitive decline. It interacts with retromer and actin and modulates PKC signaling.

Kiss-and-run

A phenomena where vesicles come in contact briefly through a small fusion pore that allows for the exchange of small cargo. This is then followed by unsuccessful fusion and the dispersion of the vesicles.

LBPA

LysoBisPhosphatidic Acid is involved in regulating intracellular protein and lipid transport including cholesterol homeostasis.

LTB₄

LeukoTriene B_4, is a lipid mediator generated by cPLA2-dependent arachidonic acid release, involved in directional migration of neutrophils and macrophages in tissues.

MDV

Mitochondria-Derived double membranous Vesicles generated through the selective incorporation of mitochondrial contents, mostly from the outer membrane, and involved in mitochondria quality control.

MHC II

Major histocompatibility Complex II is involved in antigen-presentation by immune cells such as monocytes, macrophages, and dendritic cells.

miRNA

microRNAs are a class of non-coding RNA, usually transcribed as precursor miRNA, which upon maturation through RISC, modulate the expression of specific genes.

mTORC1

mammalian Target Of Rapamycin Complex 1 is a set of proteins that integrate various growth factor and nutrient signals to regulate cell growth during energy sufficiency and starvation.

Proteasome

Multi-subunit protein complex that regulates cellular homeostasis by selective and efficient degradation of damaged and surplus proteins.

Retromer

Protein complex required for the endosome-to-Golgi retrieval pathway that is necessary to maintain an active pool of hydrolase receptors in the TGN.

RISC

RNA-Induced Silencing Complex, comprised of dicer, double stranded RNA-binding protein and argonaute2, and involved in miRNA maturation and target-RNA cleavage.

Retrofusion

A phenomenon where vesicles fuse back to the membrane from which they originate. Alternatively known as back-fusion.

Secretory autophagy

A process where autophagosomes fuse with the PM rather than lysosomes. This allows for the unconventional secretion of cytosolic proteins that lack signal sequences required for conventional secretory pathways.

SNARE

Soluble N-ethylmaleimide-sensitive factor-Activating protein REceptors constitute a large superfamily of membrane-bound proteins that facilitate fusion of two membranes.

Syndecan-syntenin complex

HSPG-adapter protein pair that facilitates ILV biogenesis by recruiting ALIX.

Tetherin

a transmembrane protein with a GPI-anchor known to anchor enveloped viruses on the PM.

Tetraspanins

Four-pass transmembrane proteins involved in membrane protein scaffolding.

trans-Golgi network

The plasma membrane facing side of the Golgi apparatus that constitutes the major secretory pathway sorting station and directs newly formed proteins to various target organelles.