Multivesicular bodies contain membrane vesicles which either undergo lysosomal digestion or are released in the extracellular environment as exosomes. Evidence is accumulating that supports a physiological role for exosomes in, for example, antigen presentation or removal of transferrin receptor during reticulocyte development. Here, inspired by observations on exosomal release from reticulocytes, we discuss the potential involvement of the so-called ESCRT mechanism in the entrapment of both lysosomal and exosomal cargo within the intralumenal vesicles of multivesicular bodies. We propose that this mechanism operates at different sites in the endocytic itinerary in different cells, thereby providing a tool for directional sorting. We also explore the possibility that the efficiency of sorting of molecules into exosomes increases when the recycling kinetics of molecules decreases, exosomal sorting being favored by intermolecular interactions occurring within lipid domains, or with protein webs, that slow lateral mobility. These considerations are mirrored in the context of current knowledge on the mechanism of protein sorting for degradation in lysosomes, and the hijacking of such mechanisms by some retroviruses for particle budding.
Exosomes are small (50–100 nm) membrane vesicles that are secreted by various cells, in particular, cells of hematopoietic origin (1). They arise from the inward budding of endosomal membranes, which generates so-called multivesicular bodies (MVBs). The intralumenal vesicles can be released as ‘exosomes’ following fusion of the MVBs with the plasma membrane. Exosomes were first described in connection with the maturation of reticulocytes some two decades ago (2,3), and provide a means to sort obsolete proteins, such as transferrin receptor (TfR), for secretion as the cells differentiate into erythrocytes. More recently, other hematopoietic cells were also shown to secrete exosomes. These include antigen-presenting cells (APC), which might exploit this mechanism to regulate the immune response. Indeed, APC exosomes contain peptide-loaded major histocompatibility complex (MHC) class I and II molecules and the costimulatory factor CD86 (4), which activate T-lymphocytes in vitro (5) and suppress the growth of established murine tumors when injected into mice (6). These findings prompted a reappraisal of the exosome from a garbage sac, releasing obsolete proteins, to a device involved in triggering intercellular communication. In addition, the ‘exosomal pathway’ can be hijacked by certain viruses. Thus, in macrophages, HIV buds directly into the MHC class II-enriched multivesicular compartment (MIIC), and is subsequently released into the extracellular space together with exosomes (7). Possibly, in this manner the virions bypass the organism's defense and increase their infection efficiency (8), although the concept of ‘viral exosome’ is still a matter of debate (9).
The generation of MVB is a well-defined event in the endosomal pathway and it is apparent that they play a multifunctional role since, in addition to their involvement in exosomal release, proteins and lipids destined for lysosomal degradation are temporarily stored in MVB (10,11). In both cases, the molecules sequestered into intralumenal vesicles are usually ‘lost’ from the cell. However, in APC the intralumenal vesicles may also fuse with the MVB-limiting membrane from which they originated. This process can restore molecules for intracellular processes, as reported for MHC class II molecules in the MIIC of maturing dendritic cells (DCs) (12). The MVB compartment thus constitutes a highly dynamic system, acting as both a reservoir for molecules and an intermediate station for components en route to degradation, intracellular recycling and secretion. It is evident that a careful regulation of targeting and trafficking must be instrumental in these various pathways, the mechanisms of which have yet to be elucidated.
Exosomes Originate from Different Endosomal Compartments
During each round of recycling and/or transport of the TfR and MHC class II molecules in reticulocytes and APCs, respectively, minor proportions of the cycling protein are sorted into exosomes (3,5,13). However, as TfR and MHC class II molecules are major cargo proteins in the exosomes released from either cell type, it can be inferred that the vesicles may originate from compartments that are differentially positioned along the endocytic pathway. Although certainly capable of endocytosis, reticulocytes only possess a partly developed endocytic pathway; lysosomes are virtually absent and the multivesicular structures display biochemical characteristics typical of early sorting endosomes (3,14). By contrast, in APC the MIIC is located more distal in the endocytic pathway, as suggested by colocalization with CD63 (4), which cycles between prelysosomal compartments and the plasma membrane (15,16). Interestingly, the guanosine-5′-triphosphatases (GTPases) rab4 and rab5, regulating early endosomal trafficking, have been detected in reticulocyte exosomes (17), whereas the late endosomal marker rab7, and rab11 which is particularly rich in perinuclear recycling endosomes, have been identified in exosomes secreted by DCs (18). In this context, overexpression of the dominant-negative mutant Rab11S25 N in the erythroleukemic cell line K562 inhibited exosome release, whereas secretion of exosomes was stimulated in cells transfected with wtRab11 (19). Most importantly, although a direct role for rab proteins in sorting into exosomes remains to be determined, the presence of different rab proteins in distinct exosomal populations may indicate differences in intracellular sites where exosome formation can occur.
How does the cell discriminate between protein cargo in MVBs, destined for either exosomal secretion or lysosomal degradation? Although this issue remains to be resolved, it is possible that the site of MVB formation along the endocytic pathway, in conjunction with specific sorting cofactors, plays a role in this process. A machinery responsible for sorting proteins in the intralumenal vesicles of MVB has recently been identified (20). It consists of a series of proteins, including three protein complexes − ESCRT-I, -II and -III (Endosomal Sorting Complex Required for Transport) − that sequentially recognize mono-ubiquitinated transmembrane proteins and are believed to induce their inclusion into membrane domains that generate the intralumenal vesicles of MVB. Once completed, the ESCRT machinery dissociates from the MVB membrane and is reutilized for another round of protein sorting (21). Most proteins sorted in this manner have been shown to be destined for degradation, as accomplished by fusion of MVBs with lysosomes. Since exosomes are similar in nature to the intralumenal vesicles of MVB, the mechanism for sorting into exosomes may also involve a ubiquitination-dependent sorting via the same ESCRT machinery but at locations that differ from those involved in generating vesicles destined for degradation. Thus, driven by specific features on distinct endosomal populations, the same ESCRT machinery could be recruited for vesicular formation, culminating in either exosomal release or lysosomal degradation. Although direct evidence for ESCRT involvement in exosome biogenesis is lacking, its involvement is suggested by observations that proteins required for ESCRT recruitment to endosomal membranes, in particular Tsg101 (Tumor susceptibility gene 101), a component of the ESCRT-I complex that can bind to ubiquitin (18,22,23), can be recovered from exosomes (18,24). Moreover, Alix (ALG-2 interacting protein X), a protein that can bridge ESCRT-I and ESCRT-III by binding simultaneously to Tsg101 and CHMP4 (25–27) has also been found in exosomes (18). Interestingly, following AP2 degradation by the proteasome during red cell differentiation, Alix interacts with the YxxΦ motif of the TfR in reticulocyte exosomes (22). TfR sorting into MVBs occurs in early endosomes in these cells, as described above, suggesting that the ESCRT-driven sorting and budding machinery can also operate at early endosomal membranes as a sorting device for recruitment of exosomal cargo as well as the surface of late endosomes. Thus, cargo protein sorting by ESCRT might be accomplished through Alix binding, as for TfR in maturing reticulocytes, or through recognition of ubiquitin.
In support of such a site-specific generation of exosomes along the endocytic pathway is the observation that Tsg101, which primarily localizes to late endosomes, can be relocalized to early endosomes by overexpression of Hrs (28), which associates with early endosomes via its FYVE domain binding to PI3P. Hrs binds ubiquitinated cargo, and the protein simultaneously recruits the ESCRT-I complex through binding Tsg101. Thus, by assembling on endosomal compartments that display an intrinsic capacity for recycling, like early sorting and perinuclear recycling endosomes, the ESCRT machinery may ‘drive’ sorting of cargo into exosomal entities rather than mark it for degradation. The presence of different rabs in distinct exosome populations, as noted above, would be entirely consistent with this notion.
Finally, it cannot be excluded that different endosomal membrane microdomains, localized in the membrane of a given endosomal compartment, could be instrumental in sorting exosomal from degradation cargo. However, we consider this possibility highly unlikely since it is difficult to envision how efficient sorting and a redirection of transport of entire vesicles, trapped within the same membrane-bounded compartment, would be accomplished.
Frustration of Recycling Is Instrumental for Sorting into Exosomes
Molecular sorting along the endocytic pathway has been intensively investigated over the last decades. Following its incorporation into the plasma membrane, the fluorescent phospholipid analog C6-NBD-SM (N-(N-[6-[(7-nitrobenz-2-oxa-1,3-diazol-4-yl)amino])caproyl])-sphingomyelin) is endocytosed and rapidly recycled to the cell surface (29,30). By contrast, the poorly recycling lipid analog N-Rh-PE (N-(lissamine rhodamine B sulfonyl) phosphatidylethanolamine) reaches lysosomes after internalization (31). When internalized by reticulocytes, which are essentially devoid of lysosomes, N-Rh-PE accumulates in an intracellular compartment before being released into the medium in exosomes, whereas C6-NBD-SM is very efficiently recycled to the plasma membrane (32). Interestingly, N-Rh-PE is also readily sorted into the intralumenal vesicles of the MHC class II-enriched multivesicular compartment in human melanoma cell line Mel Juso (33). This indicates that irrespective of the underlying mechanism, exclusion from a recycling pathway promotes sorting into MVBs. By studying trafficking of lipids carrying acyl chains of various lengths and degrees of unsaturation of their acyl chains, insights have been obtained into molecular parameters that govern the intracellular fate of a given lipid (34). Thus, lipid probes that partition into fluid membrane domains readily engage in recycling pathways, whereas those in a less fluid environment predominantly translocate to late endosomes, and eventually lysosomes. In a more physiological context, specific lipid partitioning has also been described in the case of MVB, since lyso-bis-phosphatidic acid (LBPA), a rare negatively charged lipid, is enriched in the intralumenal vesicles of late endosomes (35), where it may play a direct role in MVB vesiculation (36). Interestingly, while colocalizing with MHC class II molecules in luminal vesicles of MIIC in B lymphocytes (37), no enrichment of LBPA was found in vesicles secreted by mast cells, dendritic cells and B-lymphocytes (38–40). This distinction supports the notion that exosomes can be released from MVBs, which display characteristics that may differ from those destined for degradation.
The rate of recycling of glycosylphosphatidyl inositol (GPI)-anchored proteins to the plasma membrane is relatively slow, which presumably relates to their relative retention in lipid rafts in recycling endosomes (41). Importantly, several GPI-anchored proteins such as acetylcholinesterase, prion protein, CD55, CD58 and CD59 are enriched in various types of exosomes (42–44). In addition, lipids like cholesterol and the ganglioside GM1, which often localize to raft domains, are enriched in exosomal membranes (45,46). Indeed, Triton X-100 insoluble fractions, an experimental criterion indicative of raft microdomains (47), containing acylated proteins such as GPI-anchored protein, have been isolated from exosomes of different origin (46). Thus, GPI-anchored proteins and other raft-associated proteins such as flotillin, stomatin and the tyrosine kinase Lyn (46) are likely sorted into exosomes due to their preferential partitioning into these lipid domains. This is highly reminiscent of another protein family, the tetraspan proteins (TSP), which are enriched in exosomes secreted by B-lymphocytes (4). The TSP are acylated, like stomatin and flotillin, and they interact with other proteins such as integrins and MHC class II molecules to form a protein web, which displays affinity for detergent-resistant domains (48,49). Consistently, a pool of MHC class II molecules can also be co-pelleted in detergent-resistant fractions isolated from exosomes, secreted by B-cells (40,46). Apart from providing a (co)signal for specific sorting into exosomes, antigen recruitment into rafts, and their clustered appearance on the surface of APC may also have functional significance in that it may increase the local density of specific class II-peptide complexes, thereby facilitating T-cell activation (50).
Thus microdomains, as a means to specifically concentrate cargo, may operate in exosomal sorting, and act in concert with the ESCRT-based sorting machinery. The overall principle may be exemplified by the mechanism by which the TfR is sorted into exosomes of reticulocytes. In these cells, a fraction of the TfR is likely concentrated or clustered like an aggregate, as revealed by SDS-PAGE analysis (31). Recruitment into microdomains could be instrumental in its clustering. This feature may
• promote, through Alix binding, the recruitment of the ESCRT machinery by increasing the avidity of the interactions;
• increase the recycling half-life of the receptor, as has been shown using dimerized (51) or oligomerized Tf (52);
• increase the sorting efficiency by a cascade effect, since one ESCRT-sorted receptor could drag along other bound receptor.
Consistent with these possibilities, TfR cross-linking with antibodies or lectin enhanced the amount of receptor released by exosomes from reticulocytes (32), in accordance with receptor down-regulation by interactions with a multivalent ligand, as documented in other cell types (53,54).
In summary, frustration of molecular recycling, as induced by various causes such as interaction with lipid raft domains and/or protein networks, likely favors the sorting of cargo into exosomes (or MVB intralumenal vesicles).
Reutilization of Released Exosomes
Once secreted into the extracellular medium, exosomes can be captured and reprocessed by other cells. Intriguingly, exosomes have recently been demonstrated to contain the GPI-anchored cellular prion protein (PrPc) and to act as potential vectors of the infectious form (23). However, particularly in eliciting an immune response, exosomes may play an important physiological role. Indeed, secreted exosomes containing MHC class I and/or II molecules have been shown to stimulate the immune system either directly by an interaction with T-cells, as revealed in vitro for B-cell exosomes (5), or indirectly through interaction with DCs as demonstrated in vitro (55) and in vivo after exosome injection (56). Dendritic cells also secrete exosomes, which may sensitize adjacent DCs, thereby amplifying the immune response (57). Moreover, tumor cells also secrete exosomes containing functional peptide/MHC class I complexes in their membrane and cytosolic tumor antigens (58). As such, tumor-derived exosomes collected from ascites are considered a potential approach in cancer immunotherapy, the principle of treatment relying on exosomal reprocessing by DCs (59).
Intestinal epithelial cells also secrete exosomes carrying MHC class II molecules with bound antigenic peptides (60,61). The basolaterally released vesicles may be involved in priming an immunogenic (62) or a tolerogenic (60) response. Finally, during the last stage of erythropoiesis, CD55- and CD59-containing exosomes released by maturing reticulocytes may transfer their GPI-anchored proteins to bystander cells. Indeed, in transgenic animals, human complement restriction factors expressed only by the erythroid lineage were expressed on the surface of vascular endothelial cells, demonstrating in vivo transfer of CD55 and CD59 from differentiating red cells to vascular endothelial cells (63,64).
Viruses Have Hijacked the Various MVB Sorting Possibilities
In Rous sarcoma virus (64) and other RNA enveloped viruses, so-called late domain (L-domain) signals, located in the Gag or matrix proteins, have been described. These motifs act to recruit the cellular machinery necessary for viral budding and membrane fission (28,65). Interestingly, this machinery includes the ESCRT complexes, which implies that at least some viruses have copied the molecular principle of ubiquitin-based sorting in MVB. In the case of HIV, a tetrapeptide motif (PTAP) located within the L-domain (p6) of Gag recruits the ESCRT-I complex through binding to Tsg101 (66,67). ESCRT components can then be sequentially recruited and assembled in the protein network involved in MVB biogenesis (68). In this sense, Gag contributes to the strategy of hijacking the ESCRT machinery since:
• Its membrane attachment by a myristic acid moiety mimics the membrane binding of the cellular Hrs FYVE domain to phosphatidylinositol 3-phosphate (28);
• a PTAP motif instead of a PSAP motif on Hrs allows the recruitment of Tsg101;
• like other ESCRT-sorted cargo proteins, Gag can be mono-ubiquitinated allowing interaction with Tsg101 (65) (Figure 1).
These three characteristics are common to other enveloped RNA viruses, which possess PT/SAP motifs (69) stressing the convergence of virus evolution to hijack the ESCRT cellular machinery. However, contrary to other retroviruses, budding of equine infectious anemia virus (EIAV) does not seem to require Gag ubiquitination since EIAV particle release is insensitive to proteasome inhibitors that deplete intracellular stores of free ubiquitin (70). Moreover, EIAV assembly is dependent on a YxxL motif (instead of PT/SAP) contained in the Gag p9 l-domain (71). It was recently shown that interaction of Alix with this YxxL motif allows Gag to bypass the need for ubiquitination (25,72) (Figure 2). Interestingly, the YxxL motif of Gag p9 was reported to interact with AP-2, inducing a colocalization of the α-adaptin subunit of AP-2 with the EIAV p9Gag protein at sites of virus budding at the plasma membrane (71). Furthermore, data on TfR sorting in reticulocyte exosomes (22) suggest that EIAV may mimic physiological sorting of YxxΦ-based receptors, involving Alix. Thus, Alix may bind to YxxΦ motifs on receptors, thereby redirecting their trafficking to degradation by MVB sorting (Figure 2). This could provide a means by which the cells could control the lifespan of an YxxΦ-based endocytic receptor. Intriguingly, in addition to the YxxL/F motif, a sequence similar to the surface-exposed helix of ubiquitin is present on EIAV p9 (70) and the TfR cytosolic domain.
Although the main site of HIV-1 particle assembly is the plasma membrane in most cells (73), in macrophages HIV-1 budding can take place in an intracellular compartment such as MIIC (7). The reason for these distinct viral assembly sites is not yet clear, but the occurrence itself is in perfect agreement with the notion described above for the recruitment and assembly of the ESCRT complexes at sites different from late endosomes. Moreover, Gag oligomerization in virus budding (74) may parallel the role of aggregation in TfR sorting into reticulocyte exosomes. Similarly, the likely involvement of lipid raft domain in exosome biogenesis might also represent a means by which viruses increase their budding efficiency (75,76).
To summarize, exosomes may well provide a mechanism to evacuate components from the endocytic/recycling pathway. This clearance process may be a quality control system to verify the recyclability of molecules. Physiologically, the molecular cause for divergence from this pathway could be diverse (e.g. molecular aggregation, protein aging, down-regulation) and the means used to do so more or less specific (e.g. cellular sorting machinery, intrinsic molecule diffusion kinetics) explaining the heterogeneity of proteins classes secreted via exosomes. In this concept, the exosome ‘function’ is not regarded as a secretory mechanism for cellular content as such, but rather as a sophisticated and specific means for reprocessing/reutilization of the (specific) exosomal cargo by bystander cells. Moreover, the sorting principle of exosomal release seems to have been selected during evolution by various pathogens, including retroviruses and the prion protein, providing an escape from intracellular surveillance and hence favoring their spread.
Work from our laboratories has been supported by the CNRS, the Ministère de la Recherche, and the Association pour la Recherche sur le Cancer (ARC #3444).