Context-specific regulation of extracellular vesicle biogenesis and cargo selection

Abstract

To coordinate, adapt and respond to biological signals, cells convey specific messages to other cells. An important aspect of cell–cell communication involves secretion of molecules into the extracellular space. How these molecules are selected for secretion has been a fundamental question in the membrane trafficking field for decades. Recently, extracellular vesicles (EVs) have been recognized as key players in intercellular communication, carrying not only membrane proteins and lipids but also RNAs, cytosolic proteins, and other signaling molecules to recipient cells. To communicate the right message, it is essential to sort cargoes into EVs in a regulated and context-specific manner. In recent years, a deluge of lipidomic, proteomic, and RNA sequencing studies have revealed that EV cargo composition differs depending upon the donor cell type, metabolic cues, and disease states. In many cases, analyses of these distinct cargo “fingerprints” have uncovered mechanistic linkages between the activation of specific molecular pathways and cargo sorting. In addition, cell biological studies are beginning to reveal novel biogenesis mechanisms regulated by cellular context. Here, we review context-specific mechanisms of EV cargo sorting, focusing on how cell signaling and cell state influence which cellular components are ultimately targeted to EVs.

II. Introduction

Once regarded as an essential waste disposal pathway¹ or a process that only takes place under specialized circumstances^2,3, secretion of extracellular vesicles (EVs) is now recognized as a bona fide mechanism to exchange molecules and convey signals between cells. EVs can circulate through the blood to affect distant tissues or remain near the site of secretion to promote autocrine or paracrine signaling⁴. Two major subtypes of EVs – exosomes and ectosomes – have been categorized based on their origin in the cell⁵. Exosomes are small (~50–150 nm diameter) EVs that form when the endosomal membrane buds inwardly to create intralumenal vesicles (ILVs). These ILVs are secreted as exosomes when the endosome, now called a multivesicular body (MVB), fuses with the plasma membrane (Fig. 1a,^5–7). On the other hand, ectosomes arise from outward protrusions of plasma membrane that are excised and shed into the extracellular space (Fig. 1b,^5–7). Ectosomes range in size from less than 100 nanometers to several micrometers in diameter and include a variety of vesicle types, including large oncosomes. Recently, additional EV subtypes have been described, such as migrasomes (Fig. 1c) that fit less well into these two main categories and are still being characterized⁸. Apoptotic bodies that form when cells fragment during programmed cell death are also categorized as EVs, in some cases forming from protrusions called apoptopodia (Fig. 1d)⁹. While EV subtypes are defined by biogenesis mechanism, they can be difficult to distinguish from each other in typical biochemical preparations, leading to recent proposed nomenclature standards in which purified EVs are described as small and large EVs (Box 1)¹⁰. In this review, we will use these nomenclature standards, except in cases where the biogenesis mechanism has been defined by other methods, such as imaging.

Figure 1: A road map of EV biogenesis.

a | After endocytosis, early endosomes undergo maturation into multivesicular bodies (MVBs) containing intraluminal vesicles formed through inward membrane budding into the endosome lumen. Intraluminal vesicles can be directed towards lysosomal degradation by MVB-lysosome fusion or towards secretion into the extracellular space through MVB-plasma membrane fusion. The immunoelectron micrograph (i) shows gold-labelled transferrin receptor on the surface of exosomes². b | A variety of biogenesis pathways give rise to ectosomes, but in all cases stem from surface blebs or protrusions that are cut off from the plasma membrane. Bladder cancer cells release large oncosomes (ii), shown by staining with cholera toxin B and visualization by confocal microscopy³²³. Smaller “classical” ectosomes immunostained for annexin A1 are visible by structured illumination microscopy budding off of the plasma membrane of colorectal cancer cells (iii)¹⁷⁸. Much smaller ectosomes can also bud off of the plasma membrane through a pathway depending on arrestin domain-containing protein 1 (ARRDC) (iv). These ARMMs (ARRDC1-mediated microvesicles), are detected on the surface of cells expressing ARRDC1-mCherry by immunoelectron microscopy with gold-labelled anti-mCherry⁹³. Small ectosomes may also bud off of the tips and sides of surface protrusions (v). Inset (v) exemplifies ectosome budding off the tips of microvilli in the brush border of an ATP-treated rat small intestine³²⁴. c | Other types of extracellular vesicles include migrasomes, EVs containing internal vesicles that arise from retraction fibers (transmission electron micrograph (vi))⁸, and vesicles originating from amphisomes that are formed when when the outer autophagosome membrane fuses with a late endosome. d | Apoptotic bodies arise from orderly fragmentation of apoptotic cells. Caspase-3 substrates include key players in apoptotic body formation, including ROCK1 and other regulators such as Pannexin-1 and Plexin-B2^9,172,173. ROCK1 activates actomyosin contractility, resulting in blebbing from the plasma membrane itself or from the tips of surface protrusions termed apoptopodia^9,173. Apoptopodia can undergo beading along their length, visualized by DIC microscopy of apoptotic THP-1 monocytes in (vii)⁹.

Box 1: EV nomenclature describing EV size and origin.

Historically, EVs have been described by a variety of names, including exosomes, microvesicles, microparticles, and ectosomes. Most typically, the term exosomes has been used to describe small EVs that are formed within MVBs in cells, whereas microvesicles and ectosomes have referred to larger EVs that bud from the cell surface. However, many early papers used these terms interchangeably, making it difficult to identify what type of EV was being studied without digging deeply into the methods. Moreover, EVs are often studied after isolation from biofluids or cell culture conditioned media, which contain mixtures of EVs from a variety of membrane or organelle sources. Finally, it has become clear that not all small EVs are exosomes, since EVs less than 150 nm in diameter have been shown to both bud from the plasma membrane and be formed within endosomes as ILVs. Due to these issues, the EV field has recently sought to define vesicles based on physical characteristics rather than the mode of biogenesis¹⁰, often to the confusion of many readers. Generally, material obtained by lower-speed ultracentrifugation (10,000–16,000xg) contains “large EVs,” while material obtained by high-speed ultracentrifugation (≥100,000xg) contains “small EVs”. However, this method cannot be used to separate ectosomes from exosomes because some ectosome biogenesis pathways produce small EVs of equivalent size and density to exosomes⁹³.

Although it is experimentally practical to categorize EVs by size, it is important to bear in mind that EV subtypes (e.g., exosomes, ectosomes, apoptotic bodies) can also be useful terms when there is evidence to support their use. These biological subtypes are defined by where they form in the cell and further classified by their biogenesis mechanism (e.g., ectosome blebbing vs. virus-like budding; and ESCRT-independent exosomes vs. ESCRT-dependent exosomes). EV subtypes can share some similarities with one another – for instance, ESCRT subunits mediate both exosome budding in endosomes and ARMM-type ectosome budding on the plasma membrane^21,27,93. This biological terminology is necessary for a field-wide discussion of EV biogenesis and cargo sorting from a molecular, mechanistic and cell biological point of view. Attributing changes in EV composition or function to specific EV subtypes will help identify key regulatory pathways that control cellular communication. As an analogy, if cell biologists decided to define cell types (neurons, immune cells, red blood cells) based solely on their size, it would be difficult to discern anything about the logic of transcriptional programs and differentiation underlying cellular heterogeneity. Thus, there is a need for the EV field to move beyond a purely size-based nomenclature.

Microscopy and genetic perturbation studies have already brought progress in defining EV subtypes based on the subcellular location from which they originate^17,87,88,327. New technologies for single-EV molecular analysis (e.g., EV flow cytometry or super-resolution microscopy^328,329) may facilitate the identification of EV subtypes within complex mixtures of EVs, with the caveat that there is no current consensus on how molecular markers currently correspond to EV membrane of origin. For example, CD63 is typically a good exosome marker in most (but not all) cell types^87,89,327, but it is probably not present in all exosomes; therefore, its absence is not evidence for plasma membrane origin of a small EV. Thus, subtypes would need to be distinguishable based on multiple characteristics – perhaps by multiple surface markers in combination with size – much as how single-cell technologies are used to distinguish cell types. If successfully employed, single-EV methods in combination with cell biology investigations could facilitate identification of EV subtypes within purified EV preparations. For instance, one could use microscopy to assess how a genetic perturbation affects EV cargo sorting or biogenesis at the relevant subcellular compartments, and then use single-vesicle techniques to identify the affected EV subpopulations out of a complex mixture and characterize how their cargo compositions change.

The first EV cargo to be discovered was the transferrin receptor, which reticulocytes[G] expel on exosomes as they mature into red blood cells^2,3. Since then, the list of EV cargoes has expanded to include a complex assortment of proteins, RNAs, and signaling lipids that differ across cell types, cell signaling states, and disease conditions. Autocrine responses to EVs modulate many cellular processes, such as cell migration and serum- and anchorage-independent growth^11–15. In other cases, EVs secreted by a “donor” cell induce phenotypic changes in a distinct type of “recipient” cell. To effect such changes, EV cargoes on the external face of EVs may be presented to recipient cell surface receptors as ligands, and induce cytoplasmic signaling events either at the plasma membrane or following endocytosis^16–18. By contrast, internal EV cargoes require delivery to the cytoplasm through fusion with cellular membranes, including the plasma membrane and endosomal membranes. Endocytosed EVs may also undergo degradation in the lysosomes without delivering cargoes to the cytoplasm.

Central to the EV communication hypothesis is the ability of cells to control EV cargo selection and thereby convey specific messages in a regulated and selective manner. In this review, we discuss the existing evidence for different pathways of EV biogenesis, with a focus on cargo sorting. We will also review how cell signaling, metabolic and disease state regulate EV cargo sorting, in many cases through post-translational modification of the cargo itself.

III. A cargo-centered view of EV biogenesis

In conventional membrane trafficking[G] pathways, cargoes destined for a given organelle recruit the machinery required for their own sorting and trafficking¹⁹. As a general outline, “trafficking machineries” sequester cargoes onto patches of membrane, remodel the surrounding membrane into the shape of a vesicle, and ultimately sever the vesicle from the membrane source⁶. EV cargoes appear to follow the same basic itinerary, binding trafficking effectors (Table I) that enrich cargoes in endosomal and plasma membrane patches and recruiting membrane bending and scission machineries to generate an EV. One outcome of having cargo-specific biogenesis pathways could be to produce multiple subpopulations of EVs, each of which is regulated independently. Cargo-initiated biogenesis is thus one potential mechanism to explain the increasingly appreciated heterogeneity of EVs (Box 2).

Table I.

Effectors of EV cargo loading^*

Sorting Effector(s)	Cargo	Ref.
Hrs	Ub-modifications	22
	PD-L1	270
TSG101	CD63	27
	MHC-II
ARRDC1	TSG101	93,94
	NOTCH2
CD63	Pmel17	79
	LMP1	47
CD9	CD10	312
	β-catenin	81
CD82	β-catenin	81
	Ezrin	97
CD81	Rac	75
Syntenin	CD63	39
	Syndecan-1
	lysyl-tRNA synthetase	313
Syndecan-1	β-integrin	43
	Fibronectin
ALIX	Syntenin	39
	PAR1	314
	PD-L1	271
Arf6	MHC-I	45,136
	pre-miRNA
	Exportin-5
	Dicer
	Ago2
VPS4 CHMP4	β-catenin	315
Caveolin-1	sortillin	301
hnRNPA1 Caveolin-1	miR-27a-3p miR-27b-3p	162
	miR-92a-3p
	miR-221–3p
	miR-21
hnRNPA2B1	miR-198	148
	miR-601
	miR-451
	miR-575
	miR-125a-3p
	miR-887
	miR-17	137
	miR-93
	miR-122–5p	149
	H19 lncRNA	165
	AGAP2-AS1 lncRNA	167
	AFAP1-AS1 lncRNA	166
YBX1	miR-133	138
FMRP	miR-155	153
SYNCRIP	miR-3470a	151
	miR-194–2-3p
hnRNPU	miR-30c-5p	316
MEX3C/AP-2	miR-451a	148
La protein	miR-126	257
	miR-145
	miR-486
	miR-122
	miR-142

^*Accompaniment of cargo by the listed effector to sites of EV loading is not inherently implied. Although direct binding between effector and cargo to mediate secretion in EVs have been demonstrated in some cases, cargo loading regulation may also represent changes in intracellular trafficking or other upstream processes.

Box 2: Identifying selective EV cargo sorting: Active vs. passive processes.

The EV field has long contemplated whether cargoes are selectively sorted into EVs, or whether EV cargoes represent nonselective samples of cytoplasm and source membranes³³⁰. EV cargo “selection” and “sorting” are defined as a process that drives the local increase in cargo concentration within a nascent EV compared to the surrounding concentration. This process is driven through recognition by the biogenesis machinery and by cargo-specific interactions. Cargoes increase in local concentration because they are captured and clustered by EV biogenesis machineries. For instance, the ESCRT machinery binds specifically to ubiquitinated transmembrane proteins (e.g., integrin α5³³¹, EGFR²³), on the endosome membrane, creating patches of membrane with higher cargo concentrations that are then incorporated into budding EVs. Similarly, cytosolic cargoes may also be captured and locally concentrated by adaptor proteins, although in many cases the adaptors are less defined. In one proposed mechanism, FMRP acts as an adaptor between the ESCRT machinery and cytosolic microRNA cargoes¹⁵³. Several studies have shown that post-translational modifications provide a mechanism for regulating the selective loading of certain EV cargoes. For example, O-GlcNAcylation of hnRNPA2B1 under oxidative stress enhances its interactions with microRNAs secreted in EVs¹³⁷.

In contrast to selected cargoes, “passively” or “randomly” loaded EV cargoes are defined as those that do not undergo local increases in concentration by clustering at EV biogenesis sites. The extent to which passive loading occurs is unknown and would presumably occur upon nascent EVs randomly sampling the cytoplasm and source membrane. In principle, levels of passively loaded cargoes in EVs could increase or decrease depending on cellular expression or the rate of EV production; which could also impact the loading of selectively sorted cargoes. Therefore, changes in EV composition do not necessarily prove selective sorting at the site of biogenesis in the absence of other information. As a hypothetical example, manipulation of factors involved in nuclear import or export of RNAs could conceivably affect EV RNA levels indirectly by changing cytoplasmic RNA levels, without a specific capture step by the EV biogenesis machinery or associated adaptors. For this reason, it is important to determine at what point in a cargo’s trafficking itinerary (from synthesis to secretion) a proposed selection event occurs.

This section will describe the field’s current understanding of basic EV cargo sorting mechanisms. Section IV will then discuss how cell, tissue and disease context regulates those sorting mechanisms. As most EV cargo sorting mechanisms have been described in exosomes, we will begin by describing those, noting that many of the same principles apply to cargo sorting into certain types of ectosomes. In fact, inward budding of the endosome is topologically equivalent to outward budding of the plasma membrane, so in principle similar biogenesis machineries could be activated at different places in the cell to generate exosomes and ectosomes²⁰. For instance, the ESCRT machinery[G] is involved in both exosome and ectosome biogenesis²⁰. We will review situations in which sorting mechanisms appear similar for ectosomes and exosomes, and other situations in which sorting mechanisms appear unique.

IIIa. Biogenesis of exosomes

Many EV cargo trafficking effectors have been identified (Table I) that are involved in binding cargo at the early endosomal membrane. The endosomal membrane then buds into the lumen of the endosome, and an intraluminal vesicle (ILV) is freed by scission of the neck of the bud. The process of ILV formation, combined with removal of other cargoes to vesicles destined for other cellular locations, leads to endosome maturation into late endosomal multivesicular bodies (MVBs)⁶. Thus, biogenesis of exosomes is typically defined as ILV formation, and exosomes are by definition the subset of ILVs that undergo secretion upon MVB fusion with the plasma membrane.

IIIa1. ESCRT-mediated exosome biogenesis

The first exosome biogenesis pathway discovered was the ESCRT pathway, involving the action of all four ESCRT complexes (ESCRT-0, -I, -II, and -III) along with disassembly and deubiquitinating enzymes on the endosome membrane (Fig. 2a). In mammalian cells, ESCRT-0, -I, -II, -III are sequentially recruited to maturing endosomes. The “early” ESCRT machinery, comprised of ESCRT-0, -I, and -II, recruit and sequester transmembrane proteins such as epidermal growth factor receptor (EGFR) to ILVs by binding multivalently to lysine-63-linked polyubiquitinated residues or multiple monoubiquitinated residues (Table I)^21,22. Since transmembrane receptors are frequently ubiquitinated and endocytosed upon ligand-induced signaling²³, this mechanism of ESCRT recruitment specifically leads to capture of ligand-receptor signaling complexes and their incorporation into ILVs. Specifically, the Hrs subunit of the ESCRT-0 dimer (Hrs and STAM1/2) coordinates early steps in ILV biogenesis by binding to ubiquitinated cargoes and recruiting clathrin to the early endosome²⁴. This interaction between Hrs and clathrin is essential for clustering Hrs into a distinct type of endosomal microdomain and for sorting cargoes into ILVs^22,25. Roles for ESCRT proteins in ILV biogenesis are conserved, and deletion of ESCRT proteins in budding yeast results in prevacuolar, endosomal compartments that lack ILVs^21,26. An shRNA screen of ESCRT components in HeLa cells showed that depletion of ESCRT-0 proteins Hrs or STAM1, or the ESCRT-I subunit TSG101 inhibits small EV secretion²⁷. By contrast, depletion of ESCRT-II and -III proteins had little effect, suggesting that human cells express redundant isoforms of ESCRT-II and -III components.

Figure 2: Mechanisms of EV biogenesis and cargo sorting.

a | In the ESCRT pathway, ESCRT-0 (Hrs:STAM dimers) and other ESCRT components underly flat clathrin domains on the endosome and cluster ubiquitinated cargoes for incorporation into ILVs. Current models suggest that ESCRT-I and -II mediate membrane bending, and ESCRT-III filaments promote membrane scission in concert with VPS4. ESCRT also recruits deubiquitinating enzymes (DUbs) that remove ubiquitin from cargoes. b | An ALIX-syntenin pathway of ILV formation bypasses early ESCRT machinery. Acting as a scaffolding protein through its Bro1 domain and PRR^33,39,42, ALIX interacts with LBPA, phosphatidic acid (PA), other phospholipids (PLs), CHMP4B (ESCRT-III) and TSG101 (ESCRT-I). The V-domain of ALIX also engages short LYPX(n)L motifs on syntenin, which serves as a cargo adaptor for exosome cargoes like CD63 and syndecan³⁹. Syntenin captures cargo through its tandem PDZ domains, which can engage syndecan³²⁵. c | Lipids and membrane proteins may promote ILV generation by acting as cone-shaped wedges that bend the endosome membrane. nSMase2 removes the head group of sphingomyelin (SM) to produce ceramide (Cer), a cone-shaped lipid sufficient for in vitro ILV formation. nSMase2 is activated by factor associated with nSMase (FAN) and is pharmacologically inhibited by the small molecule GW4869⁵³. Cargoes sorted through this pathway localize to lipid rafts enriched in flotillin and cholesterol. Tetraspanins are also cone-shaped and can promote membrane bending. d | Large ectosomes originate as plasma membrane blebs that decouple from the cortical actin cytoskeleton (1). Myosin-II sliding promotes drawstring-like closing of the bleb (2). The bud is cut off from the plasma membrane (3), releasing an ectosome into the extracellular space (4).

In addition to their cargo-sequestering activities, the ESCRT-I and -II complexes play a role in bending the endosomal membrane into an ILV^19,28. In reconstituted systems, ESCRT-II association with cholesterol-rich membranes promotes the formation of liquid-ordered domains (L_o domains)²⁹ and ILVs enriched with L_o lipids³⁰, suggesting that the ESCRT pathway may depend on the L_o domain microenvironment, and potentially specific lipid interactions therein, to mediate EV biogenesis. It is possible that L_o domains themselves contribute to membrane bending when bound by ESCRT-II. L_o domains may also contribute to cargo selection due to enrichment with certain membrane proteins³¹. If L_o domain formation does underlie the ESCRT pathway generally, it would help explain why small EVs are often enriched in L_o domain lipids like sphingomyelin and cholesterol compared to cells³². Advances in the lipid imaging field have now enabled L_o domain visualization in a close-to-native context³¹, making this longstanding hypothesis more accessible to experimentation in the future. Capping off the process of exosome biogenesis, ESCRT-II induces the formation of ESCRT-III filaments that sever the neck of the nascent exosome from the endosome membrane^19,28. The ESCRT-III complex can also be recruited for this process by ALIX, an ESCRT “accessory” protein that binds lysobisphosphatidic acid (LBPA) on the MVB membrane³³. It has been suggested that ESCRT-III is directed to the vesicle bud neck by sensing negative membrane curvature³⁴ or promotes membrane curvature to drive the fission event³⁵; however, these roles are still being investigated³⁶. The AAA ATPase Vps4 then removes ESCRT-III filaments from the membrane to reset the ESCRT system and possibly facilitate scission^37,38.

IIIa2. Variations on the ESCRT pathway

Exosomes carrying different cargoes exhibit discrete requirements for ESCRT proteins during biogenesis, presumably reflecting differences in how the cargo is recruited into the ESCRT pathway. These requirements can be difficult to delineate in mammalian cells due to the expression of multiple ESCRT protein isoforms with partially redundant functions and the pleotropic effects of ESCRT depletion²⁷. However, studies have identified one variation on the ESCRT pathway – the syndecan-syntenin-ALIX pathway – that is specifically compromised by knockdown of the ESCRT-I protein TSG101, the ESCRT-II subunit VPS22, or the ESCRT-III filament protein CHMP4 (Fig. 2b)³⁹. Syndecan-1 is a transmembrane heparan sulfate proteoglycan that oligomerizes and promotes signaling through multivalent binding by growth factors and chemokines. On endosomes, the scaffold protein syntenin interacts with the syndecan-1 intracellular domain, linking it to CD63, and the ESCRT accessory protein Alix^39,40. Likewise, syndecans 2–4 can control EV biogenesis^39,41. ALIX serves as a second organizing protein at the MVB surface, binding to the lipid LBPA and the ESCRT proteins TSG101 and CHMP4B (Fig. 2b)^33,42. Syntenin, syndecan-1 and ALIX all appear to play a role in sorting cargo during this process (Table I).

Syndecan-syntenin-ALIX-mediated exosome biogenesis is regulated by activation of the oncogenic tyrosine kinase Src[G]. Src phosphorylates syndecan-1 on its intracellular domain, inducing syndecan-1 endocytosis. Src also phosphorylates syntenin and Alix, which together with syndecan-1 phosphorylation stimulate exosome biogenesis⁴³. Syntenin-mediated exosome biogenesis is also dependent on the GTPase ADP ribosylation factor-6 (Arf6), a regulator of endocytosis and vesicle trafficking, and phospholipase D2 (PLD2)⁴⁴. While their molecular contributions have not yet been defined in this context, Arf6 and PLD2 act downstream of Src⁴³ and could potentially impinge on ILV budding by producing phosphatidic acid to bind syntenin^44,45. The syndecan-syntenin-ALIX pathway is also hijacked by Epstein-Barr virus to load exosomes with latent membrane protein 1 (LMP1), the major EBV oncogene^46,47.

Another variation of the ESCRT pathway involves accessory ESCRT III proteins CHMP1, CHMP5 and IST1 that that are important for formation of a unique class of exosomes in cells subjected to glutamine deprivation and/or mTOR/Akt inhibition^48,49. Surprisingly, these exosomes are formed within Rab11-positive recycling endosomes as opposed to the classical Rab5/Rab7-positive endosomes. These interesting studies reveal functions for the poorly studied accessory ESCRT III proteins as well as their role in stress-induced EV biogenesis.

IIIa3. Lipids in exosome biogenesis

EVs are rich in cholesterol, phosphatidylcholine, phosphatidylserine, sphingomyelin and ceramide, which play diverse roles in EV biogenesis, uptake and functional outcomes in recipient cells⁵⁰. Lipids control many aspects of endosome biology, including cholesterol-regulated endosome positioning⁵¹ and size⁵² and ILV formation through ceramide⁵³. There are reports that certain cargoes sort to ILVs regardless of ESCRT expression, and that ILVs containing such cargoes are formed through a lipid-dependent pathway (Fig. 2c). For instance, proteolipid protein (PLP), an abundant membrane protein in the central nervous system⁵⁴, is still able to sort into endosomes in oligodendrocytes even when Tsg101, ALIX, or Hrs is siRNA-depleted or a dominant-negative VPS4 is expressed⁵³. Endosomal PLP localizes not to Hrs-containing membrane domains but rather to domains containing flotillin-2, an L_o domain marker, raising the hypothesis that unique raft-associated lipids might facilitate the formation of ILVs containing PLP^53,55. Indeed, the cone-shaped structure of certain lipids such as ceramide and phosphatidic acid can induce spontaneous negative membrane curvature, leading to invagination into the endosome membrane if cone-shaped lipids are produced on the outer leaflet^53,56. Supporting this idea, inhibition of cellular neutral sphingomyelinase-2 (nSMase2), a ceramide-producing enzyme, prevents PLP from entering ILVs⁵³. In addition to PLP, nSMase2 controls the EV-mediated release of other proteins (including TDP-43⁵⁷, interleukin-33^57,58, vacuolar H⁺-ATPase⁵⁹, prion proteins⁶⁰) and miRNAs^61–66, although it is not always clear that the affected EVs are exosomes. This distinction is important because nSMase2 inhibition could also affect ectosome release⁶⁷ or apoptotic body formation (rev. in⁶⁸). nSMase2-dependent and ALIX-dependent pathways can act concurrently to generate unique subsets of EVs from the same cell⁶⁹. For instance, a polarized epithelial cell line was found to release nSMase-2 dependent EVs from the basolateral surface along with ALIX-dependent EVs from the apical surface⁶⁹. These EV subpopulations contained unique molecular markers, supporting the premise that different EV cargoes follow different biogenesis pathways, resulting in EV heterogeneity⁶⁹; however some of those cargoes could conceivably follow those pathways simply due to their enrichment in apical or basal membranes (Box 2). In addition to nSMase2, exosome biogenesis is also impacted by ceramide through trafficking of existing ceramide from the endoplasmic reticulum (ER) to the endosome membrane by the ceramide transporter (CERT)^70–72 and receptor-mediated signaling at MVBs by the ceramide metabolite S1P⁷³. CERT-mediated transfer from the ER is also likely to affect ectosome biogenesis at ER-plasma membrane sites⁷¹.

IIIa4. Tetraspanins in exosome biogenesis

Tetraspanins[G] are integral membrane surface proteins that are frequently enriched in small EVs. At the cell membrane, tetraspanins interact with integrins and other associated proteins to form highly ordered tetraspanin enriched microdomains (TEMs)⁷⁴. As such, classical tetraspanins that are present as EV cargoes, including CD63, CD81, and CD9, can themselves promote EV biogenesis, incorporating and directing the sorting of other TEM-associated proteins into EVs (Table I). In fact, one study using tetraspanin pulldowns determined that as much as 45% of the EV proteome from lymphoblasts interacts with TEMs⁷⁵.

Reflecting the role of tetraspanins in membrane bending and vesicle formation⁷⁶, crystal structures show that the four transmembrane helices common to all tetraspanins form an inverted conical wedge with a pocket that binds lipids^77,78. Given the structural similarity between tetraspanins and ceramide, it is intriguing to speculate that both molecular classes utilize a common biophysical mechanism to promote membrane curvature. However, efforts to define the specific process by which tetraspanins contribute to exosome or ectosome biogenesis have been confounded by conflicting knockdown effects across studies. For instance, while knockdown of CD63 in melanoma cells decreases the number of ILVs per endosome⁷⁹, opposing effects on total secreted small EVs are seen in other cell types⁸⁰. Similar discrepancies have been reported for CD9 depletion and knockout studies^81,82. One possible reason for such discrepancies may be that multiple tetraspanins can be present in the same tetraspanin-enriched membrane domains^83,84, and may act upon exosome biogenesis within the same pathway. For example, syntenin-1 directly interacts with both CD63 and tetraspanin-6 during exosome biogenesis^39,41,85,86. It is also possible that downregulation of exosome biogenesis could lead to upregulation of ectosome biogenesis, leading to an overall increase in EV number.

The tetraspanin CD63 is considered a classical marker of small EVs arising from exosome biogenesis, localizing primarily to intracellular endolysosomal compartments and incorporating into exosome-destined ILVs^87,88. However, plasma membrane localization of CD63 has been observed⁸⁹. Conversely, CD9 and CD81, which reside mostly at the plasma membrane, also associate with late endosomes and are often identified with CD63 in MVB-derived EVs^88,90. Thus, the presence or absence of any given tetraspanin in a mixture of small EVs cannot be used to classify the EVs as exosomes or ectosomes.

IIIb. Biogenesis of Ectosomes

Unlike exosomes, ectosomes originate as outward buds of the plasma membrane that undergo fission to release the vesicles^91,92. Compared to exosomes, a larger variety in size has been observed with ectosomes, which may also reflect differences in biogenesis mechanisms (Fig. 1b).

IIIb1. Small ectosome formation

Formation of small ectosomes has many similarities to exosome biogenesis, utilizing much of the same machinery and forming vesicles of a similar size as exosomes (~100 nm diameter). These machineries include both ESCRT proteins and tetraspanins. For instance, the ESCRT-I protein TSG101 is recruited to the plasma membrane through interaction with the Hrs-mimicking PSAP motif of arrestin domain-containing protein 1 (ARRDC1), where it participates in VPS4-dependent small ectosome production (Fig. 1b)⁹³. Ectosomes formed through this pathway are called “ARMMs” (arrestin domain-containing protein 1-mediated microvesicles)⁹³. Overexpression of ARRDC1 increases levels of ESCRT and Notch signaling proteins in EVs, suggesting that these are cargoes of ARMMs (Table I)⁹⁴. Entry of ARRDC1 and Notch2 into EVs depends on expression of their respective E3 ubiquitin ligases^93,94, similar to the ubiquitin-mediated ESCRT sorting that occurs on the endosome. Tetraspanins also play roles in ectosome budding and cargo sorting. CD9 and CD81 link to the actin cytoskeleton through interacting ERM[G] and EWI proteins, and thereby participate in plasma membrane organization. This interactome is thought to influence signaling, cargo sorting, and vesicle budding^95,96, as exemplified by CD82’s recruitment of the ERM protein ezrin to membrane blebs[G] for release in ectosomes⁹⁷.

Various types of membrane protrusions have now been shown to shed small ectosomes, including filopodia^98–103, cilia^104–107, and microvilli^108–111 (Fig. 1b). For example, filopodia-like protrusions at sites of plasma membrane damage undergo Vps4B-dependent severing to produce ectosomes (Fig. 1b)¹¹². The ESCRT-III component CHMP4B localizes to damaged sites and may play a role on the plasma membrane similar to its role in exosome formation, at least when overexpressed¹¹². Crowding of glycocalyx[G] complexes on the outer plasma membrane of cells also promotes ectosome shedding from filopodia in response to DNA damage, producing semi-attached chains of “pearled” small EVs¹¹³. Small ectosome biogenesis from non-damaged cellular protrusions has also been described, through mechanisms involving hyaluronan[G] production, the actin cytoskeleton, inverse BAR domain[G]-containing proteins and cholesterol^{99,102,108,109,114}. HIV-1 particles can also assemble at the tips of filopodia, suggesting that filopodia formation may contribute to ectosome/retrovirion biogenesis in ways that are not yet fully understood (Fig. 1b, Supplementary box 1)^115,116.

Like filopodia, cilia also shed small ectosomes^104–107. In many cases these cilia-generated EVs carry cargoes related to diseases known as “ciliopathies”, including polycystic kidney disease and Bardet-Biedl syndrome¹¹⁷. In C. elegans, release of ectosomes containing GFP-tagged polycystins (associated with polycystic kidney disease) has been observed at the base of cilia sensory neurons and those ectosomes regulate worm mating behavior¹¹⁸. Ciliary release of EVs is also supported by studies that impair ciliogenesis by knocking out cilia-related genes, which reduces EV-associated secretion of specific cargoes, including hedgehog and Wnt signaling proteins^119–121 . Altered ectosome release is also associated with the pathogenesis of retinal ciliopathies. The specialized cilia of rod photoreceptor cells are capable of producing vast quantities of ectosomes, but are inhibited from doing so by the tetraspanin peripherin-2¹²². This suppression enables development of the ciliary outer segments; consequently, peripherin-2 gene knockout leads to retinal degeneration¹²².

IIIb2. Large ectosome formation

The formation of large ectosomes is less well understood than that of exosomes and small ectosomes. It is unclear whether the process involves early ESCRT machineries or tetraspanins, or any other features in common with exosome or small ectosome biogenesis. Instead, actin cytoskeletal rearrangements have been shown to underlie the process of plasma membrane blebbing and scission to release large EVs (Fig. 2d)¹²³. Molecular reorganizations within the plasma membrane are also implicated in the process, including alterations in proteins, lipids and electrolyte levels. For example, altered levels of Ca²⁺ activate a family of lipid scramblases that disrupt membrane lipid asymmetry^124,125. This process is thought to yield greater exposure of phosphatidylserine on the outer leaflet, one of the main features of ectosomes^45,126. Phosphatidylethanolamine exposure on the outer leaflet may also promote ectosome formation since in early C. elegans embryos, formation of ~200 nm sized ectosomes is inhibited by the phosphatidylethanolamine flippase TAT-5¹²⁷. Loss of TAT-5 activity on the plasma membrane leads to the accumulation of large ectosomes in spaces between cells, and loss of TAT-5 disrupts gastrulation^127,128.

Both large and small EVs contain sphingomyelin and ceramide^32,129 and the lipid raft marker caveolin-1¹³⁰, suggesting that EVs either contain or are derived from lipid raft-associated membrane domains. Caveolin is also known to regulate small EV biogenesis through binding cholesterol¹³¹; that principle may also apply to ectosomes. However, studies to systematically address the role of cholesterol and lipid rafts in EV biogenesis are lacking, and it seems possible that the ectosomal membrane is simply derived from the plasma membrane and associated lipid rafts.

Beside alterations in membrane composition, local disassembly of the cortical actin cytoskeleton combined with actomyosin contractility can promote plasma membrane blebbing and subsequent formation of large ectosomes (Fig. 1b, Fig. 2d)^45,132–134. This mechanism is used by abnormally large blebs on the plasma membranes of non-apoptotic cells shifting from a highly metastatic, mesenchymal state to the even more migratory and metastatic “amoeboid” phenotype. In prostate and breast cancer cells, loss of the actin nucleator Diaphanous-Related-Formin 3 (DIAPH3) induces the formation of large oncosomes (LO)¹²³. The small GTPase RhoA, together with its downstream targets ROCK[G] and the LIM kinase (LIMK), also regulates release of large ectosomes from breast cancer cells by respectively enhancing contractility and downregulating the activity of the actin severing protein cofilin¹³⁵.

In addition to their role in the syndecan-syntenin-ALIX pathway of exosome biogenesis (section IIIa2), Arf6 and PLD2 also regulate ectosome formation. At the plasma membrane, ARF6 activates phospholipase D, which in turn leads to recruitment of Erk[G] and downstream phosphorylation of myosin light chain kinase and myosin light chain, increasing actomyosin contractility and fission of the membrane bleb⁴⁵. Arf6 and PLD2 also facilitate the selective incorporation of pre-miRNAs into ectosomes. Following pre-miRNA nuclear export by exportin-5, Arf6 in conjunction with GRP1 mediates transport of the pre-miRNAs and exportin-5 to the cell periphery and into ectosomes¹³⁶. Overall, it seems that the biogenesis of ectosomes is highly dependent on a combination of plasma membrane phospholipid redistribution and activity of the actomyosin contractile machinery

IIIc. RNA recruitment to EVs

While cargo selection of transmembrane receptors via ubiquitination and recognition by ESCRT components has been well established in both exosome and ectosome biogenesis^21–23,93, the mechanisms by which cytoplasmic cargoes such as RNAs and RNA-binding proteins (RBPs) are incorporated into EVs are much less clear. Many reports have found that EVs – especially small EVs – are enriched in specific RNAs relative to cellular levels, suggesting that selective sorting mechanisms exist^137–144. In addition, the RNA composition of small EVs is distinct from large EVs, with enrichment in select subtypes of small RNAs, including miRNAs, snoRNAs[G], tRNA fragments, and mRNA fragments, and a nominal presence of short full-length mRNA transcripts (≤1 kb in length)^90,145–147. In sections IV and V, we will describe how cell signaling state and cell type can regulate the RNA content of EVs. However, for both small and large EVs, the prevailing model is that RNA-binding proteins (RBPs) localize to EV biogenesis sites and act as adaptors between the RNA cargo and EV biogenesis machineries (Table I)^148–153. Confocal fluorescence microscopy supports this model, revealing the localization of RBPs such as YBX1¹³⁹, hnRNPK¹⁵⁴, hnRNPQ⁷¹, SafB¹⁵⁴, and Ago2[G]⁷¹ in the lumen of enlarged Rab5Q79L+ multivesicular endosomes.

miRNAs in small EVs are the most thoroughly studied RNA cargo in terms of both biogenesis and function. A number of studies have described selective sorting of miRNAs into small EVs^148,155,156, based on recognition of cell type-specific but mostly GC-rich specific sequence elements by RBPs^144,157 . Such sequence preferences have been defined for hnRNPA2B1 (GGAG/UGCA, AGG/UAG)^148,158, SYNCRIP (GGCU)^151,152, and FMRP (AAUGC)¹⁵³. Among these, the most extensively reported effector is hnRNPA2B1, which has been linked to the regulated loading of numerous miRNAs (Table I)^137,148,149. While the short miRNA length lends itself more to recognition by sequence motifs than structural elements, one notable exception is a noncanonical upstream structural motif that contributes to SYNCRIP miRNA recognition¹⁵². In the case of FMRP, its loading of AXXGC-containing miRNAs into exosome-destined ILVs is triggered by inflammasome[G] activation¹⁵³. hnRNPA1 also interacts with a large number of miRNAs in various cancer cell types through an as yet unknown targeting mechanism and regulates their sorting to EVs (Table 1)^159–162. It should be noted that interactions with specific RBPs could also sequester miRNA and inhibit their loading into small EVs, as has been described in endothelial cells for miR-503 regulation by hnRNPA2B1 and Annexin A2¹⁶³.

Long noncoding RNA, short mRNAs, and mRNA fragments are also sorted to small EVs through interaction with RBPs¹⁴⁷. The lncRNAs AFAP1-AS1, AGAP2-AS1, H19 and LNMAT2 are targeted to small EVs by hnRNPA2B1 (Tables I and II), with its canonical GGAG binding motif validated in the latter two transcripts^164–167. One study in HEK293 cells identified sequence elements that direct mRNA to EVs. EV secretion of mRNA fragments is linked to the presence of three sequence motifs that are bound by SYNCRIP and NSun2. Interestingly, the location of those sequence motifs within the mRNA transcript dictates the specificity of binding by the two RBPs and the efficacy of mRNA targeting to small EVs¹⁶⁸. Unlike small EVs, large EVs can also contain full-length mRNAs >1 kb, in addition to mRNA fragments and other RNA subtypes¹⁴⁷.

Table II.

Context-dependent regulation of cargo sorting^*

Cargo	Context/Disease Association	Ref.
miRNAs
miR-17	Hyperoxia-stressed lung epithelial cells; regulated by caveolin-1 and hnRNPA2B1	137
miR-93
miR-221	Hyperoxia-stressed lung epithelial cells	229
miR-223	Irradiation-induced senescence in human dermal fibroblasts	232
miR-15b-5p	Hydrogen peroxide-induced senescence in human dermal fibroblasts Hydrogen peroxide-induced senescence in human dermal fibroblasts	233 233
miR-30a-3p
miR-23a-3p	Hydrogen peroxide-induced senescence in human dermal fibroblasts	317
miR-155	Inflammasome activated THP-1-differentiated macrophages; regulated by FMRP	153
miR-100	DKO-1 mutant KRAS colorectal cancer cells expressing Ago2-S387A	63,196
miR-320a	DKO-1 mutant KRAS colorectal cancer cells expressing Ago2-S387A, hyperoxia-stressed lung epithelial cells	196,229
let-7a	DKO-1 mutant KRAS colorectal cancer cells expressing Ago2-S387A	196
miR-122	Metabolically stressed Huh7 cells; regulated by HuR ubiquitination	318
miR-193a	Liver metastasis of mouse colon cancer; regulated by major vault protin	260
miR-503	VEGF, bFGF-treated HUVEC cells; regulated by hnRNPA2B1	163
Other RNA
LNMAT2 lncRNA	Bladder cancer 5637 and UM-UC-3 cells; regulated by hnRNPA2B1	164
NFAT mRNA	TGF-β2 stimulation of cardiomyocytes	187
HDAC5 mRNA
Proteins
Ago2	DKO-1 mutant KRAS colorectal cancer cells; regulated by Ago2 S387 phosphorylation	196
ApoE	Amyloid-β protofibrils exposure of primary differentiated cerebral cortical cells	319
bHLHE40	TGF-β1 stimulation of pulmonary artery smooth muscle cells	185
Palladin
β-integrin-1	Lysophosphatidylcholine-induced inflammation in hepatocytes	197
	Metastasis of breast cancer and melanoma	238
Caspase-3	Hyperoxia-stressed lung epithelial cells	230
Caveolin-1	Hyperoxia-stressed endothelial progenitor cells; regulated by caveolin-1 Y14 phosphorylation	137
c-Src	Onco-Dbl expressing mouse embryonic fibroblasts (MEFs)	191
FAK	MDAMB231, BT-549, Hs-578T breast cancer cells, onco-Dbl expressing MEFs	191
Fibronectin	UVB irradiation of melanocytes	320
	Migratory fibrosarcoma cells; dependent on binding to integrin receptors	11
	TGF-β2 stimulation of HCT116 colorectal cancer cells	186
HMGB1	LPS and hydrogen peroxide-induced activation of MEFs, THP-1 and RAW264.7 macrophages; LC3- and RAGE receptor-dependent	201,321
lysyl-tRNA synthetase	Serum starvation-induced release by HCT116 colorectal cancer cells; regulated by caspase-8	313
MHC-I	LOX melanoma cells, regulated by Arf6	45
MHC-II	T-cell activated dendritic cells	267
Survivin	Paclitaxel-induced in MDAMB231 breast cancer cells	322
TDP-43	ALS temporal cortex postmortem tissue	57

^*Cargos listed represent selected examples of RNA and proteins that are enriched in EVs in a context-regulated manner, not solely due to altered cellular expression.

While the importance of RBPs in mediating RNA sorting to EVs is clear, how those RBPs connect to the EV biogenesis machineries is largely unclear. One exception is FMRP, which is recruited to MVBs through its binding to Rab interacting lysosomal protein (RILP) and Hrs, thus linking FMRP-mediated miRNA sorting with the ESCRT-mediated process of exosome biogenesis¹⁵³. Another mechanism for RBP sorting to exosomes involves recognition of LC3-interacting regions in cargo RBPs by the autophagy protein LC3B¹⁵⁴. RBPs might not be captured as individual RBP-RNA complexes but rather as larger assemblies; for instance, one recent study identified a possible role for RNA-containing membraneless granules in mediating sorting of miR-233 to MVBs¹³⁹. Another study identified membrane contact sites[G] between the ER and MVBs as a key event in the formation of RNA-containing EVs. This work identified roles for the integral membrane ER tether protein VAP-A and ceramide transport protein CERT in the biogenesis of both large and small EVs enriched in RNA and RBPs⁷¹. As the ER serves as a scaffold for many types of RNA granules, it seems likely that regulation of ER-localized RNA-RBP complexes is a key component to cargo-driven biogenesis of RNA-containing EVs.

IIId. DNA in EVs

DNA can be detected in the extracellular space and in cell culture supernatants in both non-EV and EV-associated forms. In the circulation, much of the EV-contained DNA has been assumed to come from apoptotic bodies, but this is an active area of investigation as increasingly non-apoptotic EVs have been described to contain DNA¹⁶⁹. A comparative study on large and small EVs purified from prostate cancer cells and patient plasma revealed that most of the DNA is contained in the large EVs¹⁷⁰. Another study demonstrated that the DNA abundance in non-apoptotic large EVs increases in cancer cells undergoing nuclear shape instability¹⁷¹, suggesting that ectosomes may enclose and export free cytosolic DNA generated through genomic instability in rapidly dividing cancer cells.

For apoptotic bodies, disassembly of the cell and formation of apoptotic bodies is an orderly, regulated process¹⁷². In apoptotic cells, caspase-3 cleaves an autoinhibitory domain of ROCK1 to increase actomyosin contractility, resulting in blebbing from the plasma membrane directly or from the tips of surface protrusions termed apoptopodia^9,172,173. Apoptopodia can also form beads of apoptotic bodies along their length, which is regulated by actomyosin contractility and the caspase-3 targets Plexin-B2 and Pannexin-1 (Fig. 1d)⁹. It is unknown whether apoptotic bodies are released by shear force or in a regulated manner¹⁷². Pannexin-1 also regulates the incorporation of nuclear material, including DNA, into apoptotic bodies^9,174.

Exosomes and other small EVs may also contain fragmented nuclear DNA^175,176, although in many cases the DNA may simply co-isolate with EVs, potentially adhering to the EV surface^177,178. The mechanisms regulating DNA recruitment to exosomes and other small EVs are largely unknown, and it has been difficult to fathom how DNA might enter MVBs, although immunoelectron microscopy has localized DNA to the interior of ILVs within MVBs¹⁷⁹. A few reports have successfully demonstrated the presence of DNA in small EVs^175,176,180; this may reflect the context-specific nature of DNA sorting to exosomes/small EVs. Indeed, senescence (including oncogene-induced senescence) and DNA damage upregulate the levels of DNA found in MVBs and small EVs^179,181,182, which may explain why the presence of DNA in small EVs has been primarily reported for cancer cell-derived EVs. Furthermore, DNA-containing small EVs appear to be a subpopulation of the small EVs released from cells¹⁷⁷.

IV. Cargo regulation based on cell state

As the EV field has evolved, it has become clear that there are no absolute rules with respect to what is or is not an EV cargo. In many cases, it is studies on the regulation of EV cargoes by cell signaling or other cell state changes (Table II) that have led to this realization, identifying cases in which cargoes found infrequently in EVs are upregulated in a specific context. Some common themes are highlighted here.

IVa. Growth factor and oncogene signaling

A consequence of growth factor signaling, which may propagate its effects to recipient cells, is modulation of EV cargo loading. For example, fibroblast growth factor signaling in cultured hippocampal neurons, which regulates neuronal plasticity and wound repair in vivo, greatly alters the cargo of secreted EVs and stimulates MVB-plasma membrane fusion events^183,184. EV cargoes are also altered by growth factor signaling in lung tissue, where EVs mediate cross talk between endothelial and smooth muscle cells. Excessive stimulation of pulmonary artery smooth muscle cells with transforming growth factor-β (TGF-β1) both induces secretion and enhances uptake of EVs enriched in RNAs encoding cytoskeletal factors and the transcription regulator bHLHE40¹⁸⁵. Similarly, TGF-β2 signaling influences the levels of proteasome subunits, fibronectin, and histones carried by EVs from cultured colorectal cancer cell lines¹⁸⁶, and induces the loading of mRNAs encoding nuclear factor of activated T-cells 5 (NFAT5) and histone deacetylase 5 (HDAC5) into cardiomyocyte EVs¹⁸⁷ (Table II).

The transforming capacity of cancer cells is greatly expanded by the effect that oncogenic signaling has on redefining EV cargo content. A prime example is in glioblastoma tumors, in which tumor heterogeneity is maintained by different oncogenic cell subtypes through the exchange of EVs with distinct pro-tumorigenic protein cargos¹⁸⁸. One of these subtypes is driven by expression of the EGFRvIII oncogene, whose signaling profoundly alters both the expression of EV biogenesis components and sorting of EV cargoes, including invasion-promoting factors and adhesion molecules^143,189,190. Similarly, the oncogenic form of the guanine nucleotide exchange factor Dbl drives production of ectosomes that transfer focal adhesion kinase[G] (FAK) to recipient cells, promoting apoptotic evasion and anchorage independent growth¹⁹¹. In breast cancer cells, signaling by the oncogenic nonreceptor protein kinase Src enhances loading of the syndecan binding partners β1 integrin and fibronectin into small EVs that stimulate migration of recipient cells⁴³. At the molecular level, phosphorylation of both syndecan and syntenin by Src promotes endocytosis and exosomal sorting of these cargoes through the ARF6-PLD2 pathway (Fig. 3a)⁴³.

Figure 3: Examples of context-specific regulation of cargo sorting.

a | Src directly phosphorylates syndecan and syntenin to regulate integrin cargo sorting through ARF6-PLD2^43,44. b | Oncogenic mutant KRAS signaling in colon cancer cells inhibits Ago2 entry into small EVs by promoting MEKI/II-dependent phosphorylation of Ago2. Phosphorylation of Ago2 inhibits its incorporation into ILVs and increases its association with processing bodies (PBs), resulting in altered RNA and protein content in small EVs compared to small EVs secreted from isogenic colon cancer cells with wild-type KRAS¹⁹⁶. Nonetheless, certain miRNAs are secreted at enhanced levels from mutant KRAS-expressing cells⁶³, and are presumably bound by other yet to be identified RNA-binding proteins. c | Lipidated LC3 assists in nonselective engulfment of autophagy cargoes (denoted as stars) by the phagophore, a double-membrane organelle that fuses to itself to create an autophagosome. In an independent process termed LC3-dependent EV loading and secretion (LDELS), lipidated LC3 captures RNA-binding proteins (via LC3 interaction motifs) and associated RNAs onto the endosome membrane to promote nSMase2-dependent exosome biogenesis¹⁵⁴. Secretion of LC3-induced exosomes may occur via normal fusion of MVBs with the plasma membrane or after fusion of MVBs and autophagosomes to form amphisomes. Under conditions of lysosomal dysfunction, such as with inhibition of acidification by Bafilomycin A1 (BafA1), there is enhanced secretion of autophagic proteins (blue and pink stars), both inside and outside of EVs, presumably due to fusion of both amphisomes and autolysosomes with the plasma membrane³²⁶. Arrows represent membrane fusion events. d | Hypoxia activates HIF-mediated transcriptional upregulation of targeted circRNAs, miRNAs, and mRNAs, which enter EVs along with the protein products of the affected mRNAs. e | Oxidative stress from reactive oxygen species leads to caveolin-1 (Cav-1) phosphorylation, stabilizing the RNA-binding protein hnRNPA2B1 and promoting its O-GlcNAcylation. This post-translational modification enhances hnRNPA2B1 binding to microRNAs that are trafficked to large EVs¹³⁷. f | Leptin signaling activates transcription of Hsp90, which stabilizes Tsg101 to increase the number of endosomes and EVs. However, Hsp90 also antagonizes this process through its interaction with AMPKα1³⁰⁸.

Oncogenic signaling can also affect sorting of RNA cargoes into EVs. Oncogenic KRAS mutations are present in many cancers, including more than 30% of colorectal and lung adenocarcinomas and more than 80% of pancreatic cancers¹⁹². Isogenic colorectal cancer cell lines expressing wild type or mutant KRAS produce small EVs with vastly different cargo (Table II). Cells expressing mutant KRAS generate small EVs enriched in cell signaling proteins and metabolic enzymes, whereas cells expressing wildtype KRAS produce small EVs laden with RBPs¹⁹³. Similar effects of mutant KRAS on EV sorting have been reported in non-small cell lung cancer cells¹⁹⁴. Certain miRNAs such as miR-100 are also enriched in small EVs from mutant KRAS-expressing cells and can be functionally transferred to recipient cells⁶³. Moreover, mutant KRAS-derived EVs promote growth of tumor cells in soft agar; however, the specific cargoes responsible for this pro-tumorigenic behavior remain unconfirmed¹⁹⁵. Finally, RBP and RNA sorting in mutant KRAS-expressing cells are modulated by downstream MEK/Erk signaling, which inhibits Ago2 sorting into EVs by shifting its subcellular distribution away from endosomes and into RNA processing bodies (Fig. 3b, Table II)¹⁹⁶. The functional consequence of these changes in RNA-RBP trafficking remain to be defined but likely include altering the RNA expression profile of both donor and recipient cells.

IVb. Metabolic regulation of cargo sorting

IVbi. Fatty acid metabolism

As discussed in the EV biogenesis section, lipids regulate EV formation by facilitating membrane curvature and forming lipid ordered domains that provide a platform for sorting proteins. Excess fatty acid metabolism, such as occurs in non-alcoholic steatotic hepatitis[G] (NASH), also imposes dramatic changes in EV sorting, provoking the selective release of integrin β₁ as EV cargo from hepatocytes¹⁹⁷. Similar alterations in EV cargoes are brought on by fatty acid treatment of hepatic and hepatic carcinoma cells, including by myristic acid[G]¹⁹⁸ and palmitic acid[G]⁷⁰, the latter of which was shown to stimulate EV biogenesis at ER membrane contact sites. Palmitic acid may also promote EV biogenesis and cargo loading via palmitoylation of EV cargoes, potentially targeting them to the membrane at EV biogenesis sites¹³⁰.

IVbii. Autophagy

Autophagy is the process whereby cells sequester cytoplasmic components and deliver them to lysosomes for metabolic recycling. Autophagy encompasses both macroautophagy, in which cargoes are delivered to lysosomes via an intermediate organelle termed the autophagosome, and microautophagy, in which cargoes are directly internalized by endosomes or lysosomes.¹⁹⁹ It is now understood that some autophagy machinery also functions in secretory pathways, including EV biogenesis (Fig. 3c)^154,200. A recent breakthrough study found that RNA sorting to exosomes is closely linked with secretory autophagy¹⁵⁴. In this proposed mechanism, the autophagy adaptor protein LC3B and related proteins associate with the endosome membrane and recognize RBPs containing a four amino acid-long LC3-interacting region (LIR)¹⁵⁴. To couple ILV budding with cargo sorting, LC3B also recruits FAN (factor associated with nSMase) through its LIR to activate localized ceramide production¹⁵⁴. LC3-dependent sorting into EVs has also been shown for the nuclear protein HMGB1 (Table II)²⁰¹. This new role of LC3 is distinct from its classical role in capturing substrates into the autophagosome, a double-membrane organelle that directly fuses with the lysosome, since knockout of genes important for autophagosome formation downstream of LC3B lipidation do not inhibit LC3-induced exosome biogenesis¹⁵⁴. Nonetheless, the two roles of LC3 may interconnect in cases where the outer autophagosome membrane fuses with MVBs to create an “amphisome,” which itself can fuse with the plasma membrane to release conventional exosomes along with inner autophagosome contents^202,203 (Fig 3c). In another variety of endosomal microautophagy, the chaperone HSC70 captures protein cargoes containing a KFERQ peptide and not only delivers them to lysosomes but also to endosomes for secretion in exosomes^204,205.

IVbiii. Oxidative Stress

In response to environmental stress like hypoxia, hyperoxia, and chemotherapeutic agents, the production of excess reactive oxygen species (ROS) triggers an oxidative stress response that alters metabolic signaling and EV content. For example, chemotherapy results in the massive accumulation of oxidized and ubiquitylated proteins that are cleared as EV cargo²⁰⁶. ROS generated as metabolic byproducts or during immune defense also alter EV content and convey either cytoprotective or injurious effects upon recipient cells.

Under conditions of hypoxia such as in the tumor microenvironment, fibrotic lung, or ischemic myocardial tissue, the protein composition of EV cargo is dramatically altered (Fig 3d). For example, glioblastoma cells exposed to hypoxic conditions release EVs enriched in anti-apoptotic and pro-metastatic proteins²⁰⁷. In other hypoxic cancer models, EVs are differentially laden with adhesion proteins and factors involved in epithelial mesenchymal transition^208,209. Hypoxia likely exploits secretory autophagy pathways to regulate EV content, since hypoxia-induced increases in the loading of EVs with pro-angiogenic factors depends on the LC3-like autophagy factor GABARAPL1²¹⁰.

Hypoxia also alters the nucleic acid content of EVs (Fig 3d). Expression of many miRNAs are transcriptionally upregulated by hypoxia inducible factor 1a[G] (HIF-1a) and therefore more highly represented in hypoxic EVs under hypoxic conditions. These include miR-210 and miR-21^62,211–215; in models of myocardial hypoxia, both confer anti-apoptotic cardioprotective effects^216,217. In numerous cancer and cardiac disease models, hypoxia-induced EVs contain increased levels of pro-angiogenic miRNAs into EVs and promote angiogenesis in vivo^{216,218–220}. Interestingly, these types of miRNA functions can be abrogated by circular RNAs[G], and hypoxic signaling appears to fine-tune the downstream effects of EVs by modulating both miRNA and circular RNA cargos. For instance, in hypoxic pancreatic cancer cells, the circular RNA circZNF91 is upregulated by HIF-1a and transferred as EV cargo to recipient cells, where it acts as sponge to inhibit miR-23b-3p and further propagate HIF-1a-mediated transcriptional reprogramming²²¹. Hypoxic upregulation of circular RNA cargo in EVs is associated with ischemic cardiac disease, colorectal cancer and diabetic retinopathy^222–226. Hypoxic conditions also induce HIF1-mediated upregulation of mRNAs (and their cognate proteins) in glioblastoma EVs (Fig 3d)²²⁷ and altered levels of cardiac fibrosis-promoting long noncoding RNAs in cardiomyocyte EVs²²⁸. Importantly, HIF-1a also amplifies nSMase2 expression, effectively exerting its own selective process to drive the production of EVs laden with hypoxia-induced cargoes⁶².

Hyperoxia is an iatrogenic form of oxidative stress that lung cells may experience due to the formation of oxygen free radicals following administration of excess supplementary oxygen. In models of hyperoxia-induced acute lung injury, robust secretion of EVs enriched in caspase-3, miR-221 and miR-320a activates macrophage pro-inflammatory responses that stimulate tissue repair (Table II)^229,230. Conversely, hyperoxia stress in a rat model of bronchopulmonary dysplasia releases EVs containing the pyroptotic p30 fragment of gasdermin D[G], inciting widespread inflammatory, lung and brain injury²³¹. Studies of hyperoxia have also revealed important mechanistic insights into the recruitment of RNA to EVs. Specifically, hyperoxia or H₂O₂ exposure trigger phosphorylation of the lipid raft protein caveolin-1 at tyrosine-14, which then binds hnRNPA2B1. This interaction protects hnRNPA2B1 from degradation and increases its O-GlcNAcylation in the RRM1 domain (Fig. 3e)¹³⁷. hnRNPA2B1 is also sumoylated in its RRM2 domain. Both sumoylation and O-GlcNAcylation enhance the interaction of hnRNPA2B1 with specific miRNAs, allowing hnRNPA2B1 to sort miRNAs into EVs (Table II)^137,148.

IVbiv. Senescence:

Unchecked ROS can also lead to DNA damage, leakage of DNA into the cytoplasm, and induction of cell senescence or apoptosis. EV-mediated clearance of the offending cytoplasmic DNA provides a way to evade this potential cell fate^179,182. Senescent macrophages and fibroblasts secrete more, small EVs and different microRNAs compared to non-senescent counterparts, with macrophages downregulating miR-223 sorting and fibroblasts upregulating miR-15b-5p and miR-30a-3p sorting to EVs (Table II)^232,233. Senescence in primary fibroblasts also increases release of EVs carrying interferon-induced transmembrane protein 3, which leads to a more senescent phenotype in recipient cells²³⁴.

V. Consequences of disease on EV cargoes

Since EVs were discovered, the idea that their cargo could serve as disease biomarkers has excited many fields. In some cases, the cargoes are unique to specific cell types and thus serve as a marker of altered activity in a specific tissue. Stage-dependent changes in cargo content have also been documented in many diseases, demonstrating the value of EV cargo profiling for monitoring disease progression²³⁵. In many cases, these changes arise simply as a consequence of altered transcriptional activity driven by the disease state. However, in some instances, bonafide differences in the EV content of certain cargos as compared to cellular levels have been carefully documented. These disease-specific influences on EV cargo sorting warrant underscoring as they represent fascinating examples of how EV biogenesis and secretion may be targeted to modulate cell biology. Here, we highlight the effect of certain disease pathologies on cargo sorting in the associated cell types.

Va. Cancer

Perhaps the best characterized examples of context-specific regulation of EV cargo sorting come from cancer biology. Across model systems, EVs released by tumor cells and their neighboring untransformed cells have been shown to both support and inhibit cancer progression and metastasis²³⁶. Tumor-derived EVs contribute to formation of premetastatic niches by delivering proteins and miRNAs that reprogram target cells toward a pro-metastatic phenotype^237–243. Accordingly, proteomics and transcriptomics have shown major differences in the EV cargo of untransformed and cancer cells²⁴⁴

As noted in earlier sections, upstream oncogenic signaling and alterations in the tumor microenvironment can potentially act in concert to regulate EV biogenesis in cancer cells. This paradigm is illustrated well in the findings of a recent study of glutamine deprivation in colorectal, prostate and cervical cancer cell lines. Glutamine deprivation leads to an altered metabolic state and reduced Akt/mTOR signaling that promotes biogenesis of a unique class of exosomes from Rab11-positive recycling endosomal compartments⁴⁹. These exosomes have growth promoting activities and unique cargoes. Intrinsic effects of tumor metabolism may also influence EV cargo. Thus, the Warburg effect[G] on cancer cell metabolism is reflected in the cargo of cancer cell EVs and passed on to recipient cells. Small EVs released from colorectal cancer cells expressing mutant KRAS are enriched in a subset of metabolic enzymes and metabolites, including glucose transporter 1 (GLUT1)^193,245,246; transfer of these EVs to recipient cells promotes a metabolic shift toward aerobic glycolysis²⁴⁵. Similarly, free amino acids and other metabolites present in prostate cancer cell small EVs confer metabolic reprogramming in recipient cells²⁴⁷. Similar findings have been reported across diverse cancer cell types for other glycolytic enzymes, including pyruvate kinase-2, glucose-6-phosphate dehydrogenase, transketolase and transaldolase 1^248–250.

Large EVs from tumor cells are also enriched in metabolic enzymes and can drive glutamine metabolism in recipient cells^130,251. In some cases, this may reflect the presence of mitochondrial enzymes, as quantitative proteomics has shown enrichment of mitochondrial proteins in large EVs^252,253. Consistent with that idea, electron microscopy imaging has revealed whole mitochondria and mitochondrial fragments inside various types of large EVs^254–256 .Whether this represents specific recruitment of mitochondria or the bulk engulfment of cytoplasm, including whole or fragmented organelles, into ectosomes is currently unknown.

Particular attention has been paid in cancer cell biology to the transformative potential of miRNAs released in EVs. It has been broadly demonstrated that tumor cells release EVs containing altered levels of specific miRNAs relative to cellular levels, whose uptake impart significant downstream effects on tumorigenesis^257–260. In colorectal cancer (CRC), miR-100⁶³ and miR-410–3p²⁶¹ are selectively sorted in EVs. Both miRNAs have been shown to transform cells through EV uptake, with transfer of miR-410–3p by colorectal cancer cell EVs exerting PI3K-driven tumorigenic effects on recipient cells²⁶¹ and miR-100 transfer in lung cancer cell EVs imparting recipient cells with cisplatin resistance²⁶². EV loading with miR-100 in CRC is enhanced by constitutively activated KRAS⁶³ and downregulated by Ago2 phosphorylation (Table II)¹⁹⁶. Additionally, hnRNPK has been shown to load miRNAs containing the AsUGnA motif into exosomes, with trafficking of hnRNPK to MVBs occurring through a process dependent on caveolin-1 and ceramide¹⁵⁰.

Vb. Immune cells

One of the earliest functions identified for EVs is in facilitating antigen presentation. Antigen-presenting cells such as dendritic cells and B lymphocytes load antigen peptides onto MHC-II[G] inside specialized endosomal structures called antigen-processing compartments²⁶³. Fusion of these specialized MVBs with the plasma membrane releases MHC-II-containing ILVs as exosomes and also places MHC-II on the plasma membrane where it can be released in ectosomes^264,265. One study found that EVs of all sizes from unstimulated dendritic cells contain MHC-II and can activate CD4+ T cells, with small and medium EVs evoking more of a Th1 phenotype and larger EVs evoking more of a Th2 phenotype²⁶⁶. Loading of MHC-II into EVs released by dendritic cells is upregulated in response to antigen stimulation, a process that is dependent on their cognate interaction with T cells²⁶⁷. Uptake of the EVs carrying antigen-bound MHC-II by recipient dendritic cells leads to further activation of T cells and powerful modulation of the immune response^268,269. Immunoelectron microscopy studies have also shown clusters of exosomes carrying MHC-II at the T cell plasma membrane²⁶⁷.

Since interactions between tumor cells and immune cells play a central role in cancer, immune cell EVs have been studied in the context of tumorigenesis. One mechanism by which transformed cells navigate immune escape is through exosomal secretion of the transmembrane protein programmed cell death ligand 1[G] (PD-L1). In melanoma cells, stimulation with interferon-gamma triggers the loading of PD-L1 into exosomes, dependent on the ESCRT-0 protein Hrs²⁷⁰. In breast cancer cells, the ESCRT accessory protein ALIX is necessary for loading of PD-L1 into exosomes (Table I)²⁷¹. Binding of exosomal PD-L1 to its receptor PD-1 inactivates T lymphocytes by concurrently blocking MHC-II/T cell receptor signaling and costimulatory receptors required for full T cell activation^272,273. Similarly, downregulatory effects on natural killer cells and CD8+ cytotoxic T lymphocytes are conferred by exosomes presenting the NKG2D ligand to its lectin-like activating receptor^274,275. Other exosomal cargoes can have the opposing effect, activating immune defense mechanisms, such as exosomes carrying pigment epithelium derived factor (PEDF) that activate patrolling monocytes at metastatic sites²⁷⁶.

EVs selectively loaded with non-coding (nc)RNAs also mediate tumor immune surveillance. tRNA and tRNA fragments are sorted preferentially into cancer cell- and immune cell-derived EVs^257,277. Similar in length to miRNAs, tRNA fragments can regulate gene expression^278–280, apoptosis²⁸¹, and the immune response. In activated T cells, tRNA fragments are the most abundant form of EV RNA cargo. EV-mediated secretion of tRNA fragments and other ncRNAs generated during the immune response has been shown to thwart the development of chronic inflammation^277,282. Intriguingly, immunosuppressive vitamin D3 and immunostimulatory lipopolysaccharide [G] (LPS) treatments of primary dendritic cells induce opposing changes in Y-RNA[G], miRNA, and small nucleolar RNA levels in EVs²⁸³.

Vc. Neurons and glial cells

Synaptic plasticity in the central nervous system is regulated by context-specific EV loading in both normal physiology and disease pathogenesis. For instance, oligodendrocyte-derived EVs contain myelination factors, astrocyte-derived EVs contain protein cargoes that function in neurite outgrowth, neuroprotection, and synapse formation, and neuronal EVs contain proteins related to synaptic function^284–287. In Drosophila, neuronal EVs transfer Evenless Interrupted – a transmembrane protein that binds to Wingless (Wnt in mammals) – across neuromuscular junctions, facilitating synapse development²⁸⁸. External stimulation can affect the number and cargo content of the released EVs. Thus, neurotransmitters promote secretion of small EVs by oligodendrocytes, whereas glutaminergic activity induces small EV secretion by neurons^289,290. In astrocytes, ethanol-induced activation of the inflammatory response produces EVs enriched in neuroinflammatory proteins and miRNA²⁹¹. Furthermore, when exposed to trophic, inflammatory, or anti-inflammatory stimuli, astrocytes produce EVs loaded with functionally distinct cargoes²⁹².

EVs are also vectors for transmission of pathological proteins, which is thought to mediate spreading of neurogenerative disease pathology through tissues in diseases such as Parkinson’s, prion-related diseases, and Alzheimer’s²⁹³. In Alzheimer’s disease, EV-associated Tau and amyloid-β may play a central role in this process as exosome-associated Tau released from synapses can induce accumulation of additional Tau in an amyloid-β-dependent manner²⁹⁴. Tau has also been reported to be secreted in neuronal large EVs²⁹⁵. Tau expression in iPSC cells dramatically alters cargo loading to small EVs, raising the possibility that it might also have a role in EV biogenesis in neurons²⁹⁶. Likewise, the prefrontal cortex of AD patients exhibits robust upregulation of tetraspanin-6, which stimulates secretion of amyloid-β-loaded exosomes by multiple mechanisms, including altering the turnover of Alzheimer’s precursor protein and recruitment of syntenin to MVBs⁸⁵.

Vd. Cardiac cells

Several highly discriminant alterations in the cargo of serum EVs have been defined in patient sera following specific types of cardiac events. In addition to the aforementioned hypoxia-triggered alterations in long noncoding RNA cargo of cardiomyocyte EVs, CD31/annexin V double positive large EVs²⁹⁷ and VE-cadherin positive large EVs²⁹⁸ are released by injured endothelial cells and correlate with poor prognosis in coronary artery disease. A similar study of disease-specific cargo in circulating EVs following an acute coronary syndrome identified factors with known cardiac disease connections as well as a previously unknown association for polymeric immunoglobulin receptor (pIgR)²⁹⁹. After myocardial infarction in mice, large EVs carrying cardiac muscle markers transiently increase in abundance in the left ventricle³⁰⁰. While many such studies have uncovered valuable disease biomarker candidates with the potential to avert recurrent cardiac events, few have provided mechanistic insight to the processes that translate disease-specific signaling cues into altered EV sorting. One noteworthy exception is an elegant analysis of vascular calcification factors that revealed an active role for EVs in directing vascular calcification³⁰¹. The primary effector of calcification is the ubiquitous enzyme tissue nonspecific alkaline phosphatase (TNAP). Remarkably, TNAP-mediated calcification of smooth muscle depends on the deposition of EVs carrying both TNAP and the transmembrane glycoprotein sortillin. TNAP and sortillin are trafficked to lipid rafts and are dependent on Rab11, Cav-1 and phosphorylation of sortillin’s C-terminal tail by FAM20 and CK2 for loading into EVs³⁰¹.

Ve. Adipocytes and obesity

Adipocytes regulate metabolism in other tissues directly through lipogenesis, lipolysis, and fatty acid oxidation and indirectly through the immune response via release of cytokines and adipokines. In addition to these conventional secreted factors, white and brown adipocyte cells also secrete small EVs containing enzymes involved in glycolysis and fatty acid production³⁰². The small EVs secreted from white adipocytes are notably enriched in cholesterol compared to large EVs, while large EVs are marked by phosphoserine on the outer membrane leaflet³⁰³. Within white adipose tissue, adipocytes intermingle with adipose stromal cells, epithelial cells, and immune cells (macrophages, mast cells and eosinophils), which together make up the stromal vascular fraction (SVF). The EVs released from this tissue impact metabolic homeostasis in both an autocrine and paracrine fashion, most notably exerting powerful influences on tumor cell growth and pancreatic β-cell function.

Adipose dysfunction is primarily associated with obesity, as hypoxia within the expanding fat pads induces increased EV secretion of fatty acid synthesis factors that trigger adipocyte hypertrophy³⁰⁴. In obese patients, adipose-derived EV size, number and cargo composition are all altered (rev. in³⁰⁵), in part mediated through metabolic increases in the cellular levels of palmitic acid and ceramide^306,307, which escalate EV biogenesis and secretion of obesity-associated cargo. Obesity also is associated with increased levels of the fat-regulatory hormone leptin[G], which among its many effects includes stimulation of EV release from adipocytes. Mechanistically, leptin signaling through its receptor drives increased exosome biogenesis through a process dependent on Hsp90 interaction with Tsg101 (Fig. 3f). The secreted EVs are enriched in Hsp90 and other leptin signaling factors, further propagating the activation of leptin signaling in recipient cells (Fig. 3f)³⁰⁸. This pathway may be counteracted by metformin, which acts to mitigate non-alcoholic fatty liver disease by activating the metabolic sensor kinase AMPKα1, which then interacts with Hsp90 and inhibits Tsg101-mediated exosome biogenesis in adipocytes³⁰⁹. Higher circulating leptin levels are also associated with increased breast cancer risk, with leptin signaling stimulating EV release from breast cancer cells and promoting angiogenesis, invasiveness, and migration³⁰⁸. Indeed, recent studies have uncovered extensive cross talk via EV transfer between adipocytes, cardiac cells, β cells and cancer cells (rev. in³¹⁰), suggesting that EV dysregulation may underlie the development of metabolic comorbidities, e.g., obesity, cardiovascular disease, diabetes, and cancer (rev. in³¹¹).

VI. Conclusions and Perspectives

While EVs were once thought to be fairly uniform, we now know that cells from any organism release EV populations that are highly heterogeneous, with diverse cargoes and multiple biogenesis mechanisms. Indeed, EVs do not just differ because of their origin as ILVs within MVBs or plasma membrane blebs, although this is a convenient way to classify many types of EVs. The central role of cargo in recruiting various machineries illustrates how such heterogeneity can be generated, even within EV subtypes. The example of specific recruitment mechanisms through posttranslational modifications of cargo speaks to the selectivity of the process. This is particularly clear for such cargos as transmembrane cargoes that are ubiquitinated, recruiting the ESCRT machinery; however, the more recently described alternate mechanisms by which cargo engage the vesicle formation machinery (e.g., via syntenin-Alix, tetraspanin, and ceramide formation pathways) appear nonetheless specific. As another example, RNAs are selectively sorted into EVs in that the levels in EVs is not simply reflective of their cellular levels⁶³ and may depend on their location at ER membrane contact sites with EV biogenesis membranes⁷¹. However, the principal architects, conductors, and/or couriers that steer RNA-RBP complexes to these sites and into the vesicular lumen remain largely unknown. The location of cargo on the plasma membrane versus the endosome is also a likely determining factor in whether a certain cargo is incorporated into an ectosome versus an endosome.

A deeper understanding of the molecular mechanisms that drive biogenesis of distinct EV populations carrying diverse cargoes is important to drive understanding of how EVs drive cell-cell communication, as well as for biomarker and therapeutic vesicle development. In addition, a greater appreciation for the role of cell and tissue context in generating EV cargo diversity is needed. For example, to what extent are the EV cargoes of different cell types driven by the levels of expression of certain cargoes and stimulation of overall EV biogenesis pathways? And are certain cargoes more tightly controlled than others, for example by signaling and/or metabolic state (Table II)? Future studies taking in account the role of cellular context and regulation in the generation of EV diversity will be important to address such questions and to enable better engineering of therapeutic EVs loaded with specific cargoes (Box 3). It will also be important to perform comparative studies with appropriate models that allow tracking of EV cargo sorting in donor cells and EV cargo delivery to target cells.

Box 3: Engineering EVs with specific cargoes.

Although the rules for driving cargo selection and EV biogenesis are not yet entirely understood, some efforts have already been made to engineer EVs with specific cargoes. One example is the use of the C1C2 peptide of lactadherin/MFG-E8, which binds to phosphatidylserine and has been used to display a variety of proteins on the surface of phosphatidylserine-containing EVs^332–334. Other investigators have fused surface cargoes to EV resident proteins, such as CD63, to facilitate targeting to specific cells³³⁵. For targeting of luminal cargoes, several approaches have been utilized. Fusion of a palmitoylation motif to fluorescent proteins was used as a method of labeling EVs for in vivo tracking; this approach links proteins to the inner leaflet of the EV bilayer membrane³³⁶. Fusion of optogenetically controlled plant protein module CIBN to the cytoplasmic tail of the tetraspanin CD9 allowed light-dependent recruitment of CRY2-fusion proteins and EV targeting³³⁷. While lipid or tetraspanin targeting has been successful, it is clear that only a subset of vesicles is targeted. Most recently, enriched proteins identified via a proteomics screen were used to engineer fusion proteins targeting EV surfaces (PTGFRN) or lumens (MARCKS family proteins)³³⁸. While this approach seems to target a larger fraction of vesicles, more validating studies are required to determine if this holds across a variety of cell systems.

For packaging of RNAs into EVs, both brute force overexpression and engineered targeting have been employed. Cre mRNA can be transmitted into EVs via overexpression in donor cells, but the efficiency is generally low^339–341. Small noncoding RNAs are often enriched in EVs and one group recently leveraged the packaging selectivity of miR-451 to drive packaging of miRNAs and siRNAs into EVs. Reshke et al.³⁴² identified miR-451 to be a highly enriched miRNA in EVs and utilized the structural backbone of pre-miR-451 to incorporate a variety of siRNAs, including siRNAs targeting GFP and the disease genes superoxide dismutatase or transthyretin. This approach for identifying naturally selected cargoes to use as a platform for engineering may turn out to be fruitful for additional RNA cargoes, despite the paucity of knowledge about their trafficking mechanisms.

Glossary:

Reticulocyte

Nucleated precursor cell type to red blood cells.

Membrane trafficking

the process of sorting molecular cargoes, especially lipids and proteins, into different compartments within the cell

ESCRT

endosomal sorting complex required for transport, comprising five protein complexes (ESCRT-0, ESCRT-I, ESCRT-II, ESCRT-III, and Vps4) that function in cargo sorting, membrane remodeling and scission

Src

membrane-associated non-receptor tyrosine kinase that localizes to endosomes and regulates exosome biogenesis and cargo sorting

Tetraspanins

a conserved family of four-pass transmembrane proteins that organize lipid and signaling domains in membranes. Tetraspanins are enriched in many EVs and also in some cases regulate EV biogenesis.

Glycocalyx

from Greek, sweet husk; a tough casing consisting of glycoproteins, proteoglycans, and glycolipids that covers a cell

Hyaluronan

a long, linear polymer of disaccharides that forms a part of the extracellular matrix and resists compression

BAR domain

BIN, amphiphysin, and Rvs161 and Rvs167 domain; a conserved protein domain that promotes and/or senses positive membrane curvature; conversely, inverse BAR domains associate with negatively curved membranes.

ERM proteins

ezrin, radixin, and moesin proteins; proteins that link the actin cytoskeleton to transmembrane proteins in the plasma membrane, playing roles in cytoskeleton organization at the cell cortex and signal transduction

Blebs

Bulbous protrusions of plasma membrane that form when the cortical cytoskeleton decouples from the plasma membrane

ROCK

Rho-associated coiled coil containing serine/threonine kinase that acts downstream of RhoA small GTPase in cytoskeletal remodeling and membrane blebbing

Erk

extracellular signal-regulated kinase, also known as MAPK (mitogen-activated protein kinase); a protooncogene and serine/threonine protein kinase activated by EGFR and RAS-RAF-MEK signaling

snoRNAs

small nucleolar RNAs; function to perform direct covalent modification of ribosomal RNAs, snRNAs, and other RNAs. Two classes of snoRNAs include C/D box snoRNAs, which direct 2’-O-ribose methylation, and H/ACA box snoRNAS, which direct pseudouridylation.

inflammasome

a large molecular weight protein complex that activates pyroptosis, a pro-inflammatory and lytic type of programmed cell death, especially as a response to infection

Membrane contact sites

regions where two organelle membranes are tethered together (typically at a distance of ~10–40 nm) to perform a function other than fusion.

Focal adhesion kinase

a tyrosine kinase activated by integrin and growth factor signaling that controls the actin cytoskeleton to influence cell shape, focal adhesion assembly and disassembly, and cell migration.

Ago2

Argonaute 2; the central component of the RNA-induced silencing complex. Ago2 is guided by a microRNA to the open reading frame or 3’ untranslated region of a target mRNA, where it recruits other factors for translational silencing, depolyadenylation, and/or degradation of the target mRNA. Alternatively, Ago2 activates its “slicer” (endonuclease) activity when directed to an mRNA target by an exogenous small interfering RNA (siRNA) that is perfectly complementary to the target mRNA.

Non-alcoholic steatotic hepatitis

a chronic, severe inflammatory liver disease caused by metabolic alterations associated with obesity and typified by fat accumulation in the liver.

Myristic acid

a saturated fatty acid with fourteen carbons (14:0)

Palmitic acid

a saturated fatty acid with sixteen carbons (16:0)

HIF-1a

hypoxia-inducible factor 1a; a transcription factor activated under low oxygen conditions that regulates angiogenesis, apoptosis, and many other cellular processes

Circular RNAs

single-stranded, circularized RNAs arising when the 3’ end of one exon back-splices to its own 5’ end or the 5’ end of another exon.

Gasdermin D

effector of pyroptosis activated by caspase-1-mediated proteolytic cleavage downstream of the inflammasome. Cleaved gasdermin D forms pores in the plasma membrane and organelle membranes, resulting in cell death.

Warburg effect

the tendency of cancer cells to upregulate glucose uptake and aerobic glycolysis

MHC-II

Major Histocompatibility Complex II, a transmembrane protein complex found on professional antigen presenting cells that binds to exogenous antigen peptides and presents these peptides to immune cells

Programmed cell death ligand 1

a ligand that activates its receptor (PD-1) on T cells to inhibit T-cell activation.

LPS

lipopolysaccharide; a potent immunostimulatory glycolipid present on the outer membrane of Gram-negative bacteria

Y-RNA

A subtype of RNA transcribed by RNA polymerase III that is conserved among vertebrates but has poorly understood functions. Four Y-RNAs are present in humans.

Leptin

a hormone secreted by fat cells that regulates hunger and body mass