Exosomes: immune properties and potential clinical implementations
- First Online:
- Cite this article as:
- Chaput, N. & Théry, C. Semin Immunopathol (2011) 33: 419. doi:10.1007/s00281-010-0233-9
- 2.9k Views
To communicate, cells are known to release in their environment proteins which bind to receptors on surrounding cells. But cells also secrete more complex structures, called membrane vesicles, composed of a lipid bilayer with inserted transmembrane proteins, enclosing an internal content of hydrophilic components. Exosomes represent a specific subclass of such secreted membrane vesicles, which, despite having been described more than 20 years ago by two groups studying reticulocyte maturation, have only recently received attention from the scientific community. This renewed interest originated first from the description of exosome secretion by antigen-presenting cells, suggesting a potential role in immune responses, and very recently by the identification of the presence of RNA (both messenger and microRNA) in exosomes, suggesting a potential transfer of genetic information between cells. In this review, we will describe the conclusions of 20 years of studies on the immune properties of exosomes and the most recent advances on their roles and potential uses as markers or as therapeutic tools during pathologies, especially in cancer.
KeywordsExosomesIntercellular communicationAntigen presentationImmunomodulation
To communicate, cells are known to release in their environment proteins which bind to receptors on surrounding cells. But cells also secrete more complex structures, called membrane vesicles, composed of a lipid bilayer with inserted transmembrane proteins, enclosing an internal content of hydrophilic components. Exosomes represent a specific subclass of such secreted membrane vesicles, which, despite having been described more than 20 years ago by two groups studying reticulocyte maturation [68, 138], have only recently received attention from the scientific community. This renewed interest originated first from the description of exosome secretion by antigen-presenting cells, suggesting a potential role in immune responses , and very recently by the identification of the presence of RNA (both messenger and microRNA) in exosomes , suggesting a potential transfer of genetic information between cells. In this review, we will describe the conclusions of 20 years of studies on the immune properties of exosomes, and the most recent advances on their roles and potential uses as markers or as therapeutic tools during pathologies, especially in cancer.
Exosomes have been purified from in vitro cultures of various cell types, especially established cell lines of various origins (intestinal epithelial cells (IEC) , Schwann or oligodendroglial cells [53, 184], and numerous tumour cell lines, but also from primary cells of the immune (reticulocytes, mast cells, dendritic cells (DCs), platelets, B lymphocytes and T lymphocytes), and nervous systems (microglia , neurons , astrocytes , as well as fibroblasts , keratinocytes , epithelial cells , endothelial cells and mesenchymal stem cells . In addition, exosomes have been purified from several body fluids, suggesting that exosome secretion is a naturally occuring phenomenon: serum [29, 175] and plasma , bronchoalveolar fluid [4, 148], urine , tumoural effusions [13, 19, 175], epididymal fluid or sperm [58, 144], amniotic fluid , saliva , milk or colostrum [6, 69] and bile . Finally, exosome secretion is not specific of mammalian cells, since it has been also shown for avian reticulocytes , salmon leukocytes , and for an intracellular eukaryotic parasite: Leishmania .
Biophysical properties of exosomes
Since they correspond to the internal vesicles of MVBs, exosomes have specific structural and morphological properties: they are limited by a bilipidic layer and characterized by their small diameter (50 to 100 nm) and a cup-shaped morphology in electron microscopy . Floatation on a sucrose gradient has defined their density as comprised between 1.13 and 1.19 g/mL, depending on the producing cell type [153, 183]. This rather wide range of densities suggests some heterogeneity in preparations of exosomes, which sometimes may contain different types of vesicles with overlapping densities.
To qualify as “exosomes”, vesicles isolated from fluids or supernatants must meet these criteria, which distinguish them from other types of secreted vesicles. For instance, “exosome-like” vesicles secreted by endothelial cells, and bearing type I tumour necrosis factor receptor, are smaller (20–50 nm), and lighter (1.1 g/mL on sucrose gradient) than exosomes . It is not clear whether “tolerosomes” purified from serum of antigen-fed mice are exosomes or not, given their small size (40 nm), and the lack of extensive characterization . Finally, exosomes are different from larger membrane vesicles (150–300 nm in diameter), shed from the plasma membrane rather than from internal endocytic compartments, such as “microvesicles” shed by activated platelets , or “ectosomes” shed by neutrophils . The term “microvesicle”, however, has been used by some authors to designate vesicles of 50–100 nm in diameter, bearing late endosome markers, and originating from the MVBs of T cells , of several tumour cell lines , or purified from the plasma of melanoma patients : these microvesicles are probably exosomes, although their density on sucrose gradients was not reported. Conversely, vesicles enriched in late endosome markers and of less than 100 nm diameter (i.e. similar to exosomes), but budding at the plasma membrane of T cell lines, have also been called “exosomes” , although this use is not accepted by other authors. A unified terminology on secreted vesicles is thus not yet achieved in the literature.
It is important to stress that, due to their small size, exosomes cannot, at least nowadays, be visualized by fluorescence or transmitted light microscopy technics (the resolution of currently available microscope is 200 nm), nor by flow cytometry (classical flow cytometers do not distinguish vesicles below 200 nm from background noise). This makes single-exosome analyses very difficult, but current developments of more sensitive techniques of microscopy and flow cytometry raise hopes that, in the near future, such analyses will become possible.
Principles of exosome isolation
Since they are released in the extracellular medium, exosomes (like other secreted vesicles) can be relatively easily purified from cell culture supernatants or biological fluids, without the need for tedious procedures of cell disruption, required for purification of intracellular compartments. The most classical protocol used to isolate exosomes was established in the original studies published in the early 1980s , and involves ultracentrifugation (for detailed protocols, see ): the cell culture supernatant is centrifuged at increasing speeds to eliminate large cell debris and aggregates, followed by ultracentrifugation at 100,000×g to pellet exosomes, which, after wash in a large volume of PBS, can be stored at −80°C. To increase purity, a further purification step should be performed by ultracentrifugation on a sucrose gradient: contaminating non-membranous material sediments at the bottom of the gradient, while exosomes float at their characteristic density. A few groups use slightly different protocols, such as size exclusion chromatography on Bio-GelR A-50 m or Sepharose 2B columns, followed by ultracentrifugation of the void volume fraction (above 50.106 Da) at 100,000×g [94, 176]. Comparative analysis of preparations obtained side by side from human serum with the size exclusion or sequential centrifugation methods indicate similar overall protein compositions , suggesting that the two techniques lead to similar vesicle preparations, although electron microscopy analysis of the vesicles was not performed.
Very recently, a compound of undisclosed composition, called “Exoquick”, has been proposed on the market to precipitate by a single step exosomes from small volumes of serum or cell culture supernatant. However, until a full extensive characterization of the vesicles thus obtained, in terms of size, floatation on sucrose gradient, and full protein composition, is provided, it is not possible to conclude with certainty on their nature, i.e. exosomes and/or other types of vesicles.
The above techniques, however, do not allow separation of exosomes from viruses or from other nanometre-sized vesicles that could be present in the medium, and have the same density as exosomes. To address this issue, a modified protocol involving separation on iodixanol (optiprep™) gradient, where exosomes (only defined as containing Acetylcholine-esterase activity) float around 10%, whereas HIV virions (containing p24) are collected around 15% , has been proposed, but the vesicles thus purified were not fully characterized as exosomes.
Other protocols involve capture of exosomes by beads (or other solid substrates) coated with an antibody recognizing a marker enriched on exosomes (see below: composition of exosomes): MHC class II molecules for APC-derived exosomes , A33  or EpCAM  for exosomes from carcinoma cell line, or CD63 which could theoretically be used for exosomes from all sources [34, 136]. These techniques are very useful to isolate exosomes for further analysis of their composition, but cannot be applied for functional studies, since elution of exosomes from the solid substrate and/or antibody without potential alterations of the surface of exosomes (e.g. during by acidic elution) seems difficult to achieve.
Finally, a good manufacturing procedure for purification of clinical grade exosomes has also been described : it involves diafiltration of the medium containing exosomes (500-kDa membrane) and ultracentrifugation on a 30% sucrose/deuterium oxide (D2O) density cushion. Sucrose can then be removed from exosome preparations by diafiltration.
It is important to note, however, that none of the established protocols provide 100% pure preparations of exosomes, and that it is very important to minimise the presence in the starting material (culture supernatant or biological fluids) of other types of vesicles. Three major caveats of exosome purification can be pinpointed. First, since culture medium contains exosomes and other vesicles present in the foetal calf (or any other) serum , it is necessary to culture cells either in serum-free medium, or in serum-containing culture medium that has been pre-depleted from contaminating vesicles, by ultracentrifugation at 100,000×g (see ). It is worth noting, however, that the culture conditions may affect the amount, composition, or even the nature of the vesicles secreted. In particular, abrupt change of culture medium for serum-free medium during the last culture step before collection of conditionned medium must be avoided: it induces stress in the cells, which results in various physiological alterations (e.g. authophagy, induction of p53 expression etc.), possibly leading to altered profiles of exosome secretion. Second, if more than 5% dead cells were present in the culture used to produce the supernatant, membrane vesicles of non-endocytic origin will contaminate the final pellet . And it is thus not worth purifying “exosomes” from supernatants of mixed live and dead cells. Third, if the supernatant or biological fluid is frozen before elimination of large membrane vesicles, these large vesicles will be broken into smaller vesicles upon thawing and thus produce artificially small exosome-like vesicles which will also contaminate the final pellet.
In the late 1990s, in addition to classical techniques of Western blotting, the development of techniques allowing identification of proteins in a complex mixture, by mass spectrometry of the peptides generated after trypsin digestion, led to the first “proteomic” analyses of the protein composition of exosomes secreted by dendritic cells [180, 183], which were soon followed by numerous similar studies performed on exosomes purified from various sources. Ten years later, systematic compilation of these data has been proposed by an academic group , and an academic website dedicated to these analyses has been opened to encourage participation of the scientific community (http://exocarta.ludwig.edu.au).
The relative reproducibility of exosome composition, and the fact that proteins found in exosomes come almost exclusively from the plasma membrane, cytosol or endocytic compartments (including proteins involved in the formation of internal vesicles of MVBs: tsg101 and AIP1/Alix) show that exosomes are the result of a specific intracellular mechanism of formation, and thus represent a true subcellular compartment. The absence on exosomes of abundant membrane proteins, such as Fc receptors for DCs, and presence in exosomes of molecules that are not expressed at the cell surface (LAMP-2) , further demonstrates that exosomes are not mere plasma membrane fragments.
The limit of proteomic studies of exosomes, however, is that they were not performed, so far, in a quantitative manner, and unless Western blotting was subsequently used, it is not possible to know the relative abundance of the identified proteins: presence of a protein at a low level in exosomes could be explained by its presence only in a subpopulation of vesicles (or even in contaminating non-vesicular materials) present in the bulk population analysed by proteomics. For instance, some nuclear proteins, specifically histones, and proteins of the translation machinery (elongation and initiation factors) have been found in different exosome preparation, and it is not clear yet whether their presence is specific (i.e. is due to an unknown mechanism of targeting of histones or ribosomes to endocytic compartments) or is a sign of some contamination of the exosome preparation by vesicles coming out from dying cells. Similarly, molecules associated with lipid rafts (specialized membrane micro-domains that are enriched in cholesterol and sphingolipids) such as flotillin-1, stomatin and prohibitin [46, 161] were found in exosomes, but whether these molecules reflect the mechanism of exosome formation, or are due to some lipid raft shedding from the cell surface is not clear.
In addition, exosomes display molecules that are specific of the cell types secreting them: B cell receptor on B cell-derived exosomes, CD11c (a specific marker of DCs) on DC-derived exosomes, A33 antigen (a molecule essentially restricted to intestinal epithelium) in IEC-derived exosomes, molecules involved in antigen presentation (MHC class I and II molecules, CD1) and T cell stimulation (CD86 and ICAM-1) on exosomes from antigen-presenting cells. Exosome molecular composition may also vary according to the activation state of the secreting cells. Exosomes secreted by IEC exposed to interferon-γ (IFN-γ) display more MHC class II molecules than non-treated IEC . Exosomes from B cells exposed to heat stress are enriched in hsp molecules . Exosomes from DCs exposed to a maturation stimulus are enriched in molecules that are up-regulated during DC maturation: MHC class II molecules, CD86 and ICAM-1 .
In addition to proteins, exosomes contain abundant phospholipids composing their limiting membrane. The few studies that have investigated the lipid composition of exosomes do not provide a completely unified view of this composition. Exosomes secreted by erythrocytes display a similar lipid composition as erythrocyte plasma membrane  whereas exosomes from mast cells, dendritic cells [103, 169], B lymphocytes , proteolipid-protein-expressing oligodendroglial cells  or melanoma cell lines  are enriched in sphingomyelin as compared to parental cells. B cell, melanoma and oligodendroglial cell exosomes are also enriched in cholesterol [139, 184, 200], but mast cell exosomes are not . Mastocyte and oligodendroglial exosome membranes are enriched in saturated (as opposed to poly-unsaturated) phospholipids (PC and PE) [169, 184]. A lipid enriched in the internal vesicles of late endosomes, LBPA (or BMP)  and involved in formation of these internal vesicles  was surprisingly present at very low  or undetectable levels  in B cell and mastocyte exosomes. By contrast, ceramide, a lipid produced by removal by sphigomyelinase of phosphocholine moiety from sphingomyeline, is highly represented in oligodendroglial exosomes , and has been recently proposed to be the motor for formation of internal vesicles of endocytic compartments eventually released as exosomes, in these cells . Unfortunately, the results published by these four groups on the lipid composition of exosomes do not clearly determine the molar ratio of the different lipids (PC/PS/SM/Cholesterol/Ceramide) present in exosomes, and it is thus difficult to suggest a composition for artificial liposomes that would mimick exosomes, but at least such liposomes should contain sphingomyeline (and/or ceramide) and cholesterol in addition to phosphatidylcholine and phosphatidylserine.
Finally, although they expose PS at their surface [128, 136] like apoptotic and dead cells, exosomes are clearly distinct from membrane vesicles released by dying cells, in terms of protein composition and density . PS exposure is probably due to the absence in exosomes of the machinery involved in retention of PS at the internal face of plasma membrane in live cells .
A recent development of the exosome field was provided by a work published in 2007, which convincingly showed the presence, in exosomes secreted by a murine mast cell line, of messenger (mRNA) and micro (miRNA) RNA , and provided a first exhaustive list of these RNA by microarray analyses. One year later, presence of mRNA and miRNA was confirmed in exosomes purified from human primary glioblastoma cells , and another comparison of mRNA present in cells versus exosomes was performed by microarray. These two analyses showed that ribosomal RNA are not present in the vesicles, but surprisingly, another study analysing exosomes (or mixed exosomes and microvesicles) purified from urine of healthy individuals reported as abundant ribosomal RNA in these vesicles as in cells : the reason for this discrepancy with results of the other groups is not clear yet. In addition, the two studies published in 2007 and 2008 showed that the mRNA sequences present at the highest levels in exosomes were not the most abundant in cells, suggesting selectivity of RNA targeting into exosomes. Two more microarray analyses of mRNA present in exosomes were published in 2010: from a rat pancreatic carcinoma cell line , and from human saliva . Comparison of the RNA identified in the four microarray studies of exosomes should provide interesting informations on the selectivity and the possible mechanisms of targeting of mRNA into these vesicles. It will be especially interesting to test the hypothesis proposed recently, that mRNA present in exosomes are resistant to miRNA-mediated regulation, because the RISC machinery works near multivesicular endosomes to degrade miRNA-sensitive mRNA .
Concerning small RNA species, published results on miRNA present in exosomes cannot be summarized in a unified model, probably due to differences in methods for isolation and analyses of miRNA. For instance, it is still not clear whether all miRNAs sequences are targeted to exosomes, or only some of them. The first analysis of miRNA by microarray showed differences in abundance of some of them between mast cells and their exosomes , but later analyses by quantitative RT-PCR of specific miRNA species suggested instead that the more abundant these miRNAs were in cells, the more abundant they were in exosomes (from glioblastoma  or EBV-infected B cells ) or in exosome-like vesicles (from a monocyte cell line ). The miRNA profile of exosomes purified from serum of lung carcinoma patients, analysed by micrarray, was also similar to the profile of the patient’s tumour cells .
The nature of miRNA present in exosomes is thus not entirely clear. In addition, although most studies have so far only analysed sequences of mature miRNA (19–22 nt), the size profiles of RNA present in exosomes suggest that larger RNA species are also present: analyses of these species, which could be precursors of miRNAs, still have to be performed.
Involvement of exosomes in immune responses
After a first chapter describing the ways exosomes interact with surrounding cells, this section will summarize the proposed roles of these membrane vesicles in various steps of the development of immune responses, as suggested by in vitro studies.
Interaction of exosomes with surrounding cells
Once outside the secreting cell, membrane vesicles encounter extracellular matrix, but also other cells, with which they interact through different means. Vesicles bearing cytokine or death receptors can bind to cells expressing their specific ligand and thus induce signal transduction. But more complex interactions, involving transfer of the vesicle-encased components into recipient cells also take place. Several mechanisms of interaction have been proposed: adhesion at the cell surface through lipids or via specific receptors, internalization via different receptors into endocytic compartments, and fusion with plasma (or possibly internal) membrane. Due to their small size, exosomes are undetectable by regular confocal microscopy and can only be reliably visualized by electron microscopy. Several studies where fluorescently labelled exosomes were followed by confocal microscopy upon encounter with recipient cells, in vitro or in vivo, have concluded that they are phagocytosed and accumulate in endocytic/phagocytic compartments of phagocytic cells (the first such study was published by ), but may remain at the surface of non-phagocytic cells: this is most probably true at least for aggregated exosomes (leading to structures larger than 200 nm visible in confocal microscopy), but these results do not elucidate the fate of single exosomes. Development of high-resolution microscopy techniques will probably help addressing this issue in the future.
Membrane vesicles of all cellular origins consistently bear adhesion molecules, which could favour their capture by recipient cells. One such interaction has been clearly demonstrated: ICAM1-bearing exosomes secreted by mature DCs are captured via LFA1 molecules expressed by CD8+ DCs , or activated T cells . Retroviruses may have used this mechanism for their own dissemination, since HIV virions display ICAM1 derived from the surface of infected T cells, which helps dissemination to other T cells bearing high affinity LFA1 [56, 172]. Other proposed molecules include galectin5 probably bound to Lamp2 at the surface of reticulocyte exosomes , or complexes of tetraspanins (such as CD9, CD151 and Tspan8) and integrins (CD49c = alpha3 and CD49d = alpha4) on tumour exosome membrane, but receptors for these ligands on fibroblasts or endothelial cells were not identified .
Since exosomes expose phosphatidylserine (PS) at their surface, they could be captured, like apoptotic cells, via direct or indirect recognition of PS by phagocytes. Blocking of PS by addition of annexinV or diannexin indeed reduces exosome capture by endothelial cells . Milk fat globule-EGF-factor VIII (MFGE8), also called lactadherin, one of the major proteins identified by proteomics on exosomes secreted by dendritic cells , and many other cell types (see exocarta website), is bound to PS exposed at the surface of exosomes [124, 191]. When bound to apoptotic cells, MFGE8/lactadherin promotes their phagocytosis by macrophages expressing αvβ3 and αvβ5 integrins : it could thus play a similar role for capture of exosomes. Other PS-ligands include Tim1 and Tim4, two recently described PS-binding molecules expressed at the surface of activated lymphocytes or phagocytes respectively, which were shown to mediate capture of exosomes (as well as of bigger vesicles) .
Finally, a recent study analysing uptake of HIV virions and exosomes by dendritic cells showed that both types of membrane vesicles were recognized by the same surface determinants on DCs, since each vesicle could inhibit capture of the other, and that this determinant were not of proteic, but rather of lipidic nature . The exosomal lipids required for this capture were not fully determined, but they required active sphingolipid biosynthesis pathway in the producing cell, thus suggesting that ceramide or sphingomyeline are important.
The fate of exosomes after binding to the surface of recipient cells is unknown. A few observations published in the past 2 years suggest that fusion with recipient cells may occur. For instance, stem cell-derived microvesicles transfer RNA to hematopoietic cells, reprograming their differentiation , and microvesicles from glioblastoma cell lines transfer reporter RNA into normal endothelial cells , suggesting that in both cases the RNA is directly delivered into the cyotosol of recipient cells. Similarly, RNA contained inside murine mast cell-derived exosomes can be transferred and transcribed into mouse proteins in human mast cells . Another recent report shows that tumour microvesicles can transfer an oncogenic EGF receptor to cells expressing a wild-type receptor, allowing intracellular signalling leading to oncogenic transformation, suggesting that the mutant receptor has been inserted into the plasma membrane of recipient cells . Indeed, a recent study using labelling of exosomes with lipids whose fluorescence is quenched when they are concentrated on small vesicles, but becomes visible when lipids fuse with a recipient membrane and thus become diluted  has demonstrated fusion of exosomes with recipient cells, especially in conditions of acidic microenvironment, possibly mimicking the situation inside tumour masses. Whether such fusion occurs at the plasma membrane or in internal endocytic ompartments, is not clear (the author propose both sites), but this article provides a strong experimental base to explain the observed transfer of materials (genetic or proteic) via exosomes.
Exosomes and antigen presentation
Vesicles carrying antigens
Exosomes purified from cultured tumour cell lines [131, 199], or from ascites of tumour-bearing patients  bear tumour antigens, and can induce activation of antigen-specific T cells in vitro in the presence of recipient DCs that did not have other contact with the antigen. Exosomes secreted by DCs also carry antigens that can be uptaken and presented by recipient DCs on their own MHC: for instance MHC class II molecules as shown in an allograft model , or male H-Y minor antigen . Antigens spontaneously present in exosomes are mainly transmembrane proteins present at the cell surface or in endosomal compartments: Mart-1, gp100 and TRP-1, which accumulate in melanosomes of melanocytes, MHC class II molecules in DCs, and transmembrane receptors Her2/Neu and CEA found on various carcinomas. Moreover, macrophages infected with Mycobacterium tuberculosis or Mycobacterium bovis , which traffic to endosomal compartments, or endothelial cells infected by cytomegalovirus  also secrete exosomes bearing pathogen-derived antigens.
Vesicles bearing MHC-peptide complexes
Activation of CD8 T cells
Exosomes secreted by virtually any cell type bear MHC class I molecules that could potentially induce CD8 T cell activation. However, it was observed that tumour-derived exosomes can only activate CD8 T cells in the presence of recipient DCs expressing the right MHC haplotype [13, 199]. Since exosomes bearing a mismatched MHC haplotype can still induce T cell activation in these conditions, the relevant antigenic entity carried by exosomes seems to be antigens, which have to be processed by the recipient APC, and not preformed MHC-peptide complexes.
The situation is different for exosomes secreted by APCs. Several groups have reported that DC-derived exosomes could induce activation of CD8 cytotoxic T lymphocytes (CTL) clones, either by themselves [5, 111, 185], or when incubated with DCs expressing allogeneic MHC class I, showing that exosomes bear functional preformed MHC I-peptide complexes [32, 76]. In addition, in all published studies, more efficient T cell activation was obtained with exosomes purified from mature, than immature, DCs, suggesting that co-stimulatory molecules present in exosomes participate in costimulation of T cells .
Activation of CD4 T cells
Because they come from the late endocytic compartments where MHC II are stored, APC-derived exosomes also bear large amounts of MHC II molecules . MHC class II molecules born by exosomes are functional, since they can be recognized by CD4+ T cells as allogeneic antigens . Exosomes secreted by peptide-pulsed DCs can transfer preformed MHC II-peptide complexes to MHC II-deficient DCs enabling antigen-specific CD4 T cell activation [161, 181]. Exosomes bearing specific MHC II-peptide complexes can be obtained by loading APCs with antigen , or peptide [127, 161, 181]. Such exosomes can activate cognate clones , T cell lines  or activated CD4 T cells  by themselves, but need to be captured by recipient DC to activate naïve CD4 T cells [130, 161, 181]. As indicated for MHC class I presentation, exosomes secreted by mature DCs induce more efficient CD4 T cell activation than those of immature DCs [127, 161]. Recipient DCs may also use MHC-peptide complexes from exosomes as a source of peptides to load on their own MHC molecules : the relative contribution of either type of antigenic information may depend on the source of exosomes. Indeed, for MHC class II-expressing intestinal epithelial cells  or macrophages , peptides, rather than MHC-peptide complexes, seem to be the major antigenic entity transferred by exosomes to recipient DCs.
Non-antigen-specific modulation of immune responses
Apart from MHC molecules, exosomes bear also many other proteins, which can affect the immune system in various non-antigen-specific ways (Fig. 3b, c).
Inhibition of immune responses
Pioneer studies in the 1980s have suggested that tumour-derived membrane vesicles are endowed with immunosuppressive properties . Subsequently, microvesicles or exosomes obtained from tumour cell lines or tumour-bearing patients were shown in vitro to induce T cell apoptosis via FasL [14, 42, 77] and galectin-9 . In addition, exosomes can inhibit IL-2-induced T cell proliferation , and/or promote differention into Tregs , reduce CD8+ T cells proliferation , decrease NK cell cytotoxicity by displaying NKG2D ligands leading to downregulation of the receptor [17, 38, 105], impair myeloid precursor differentiation into DCs , and instead induce myeloid supressor cells [31, 106, 187, 201]. Thus, tumour-derived exosomes seem to play pleiotropic actions to subvert the development of anti-tumour immune responses.
Vesicles secreted by immune cells can also display immunosuppressive properties. Like some tumours, activated T cells secrete exosomes bearing FasL, which induce apoptosis of bystander T cells, thus participating in activation-induced cell death (AICD) .
Vesicles purified from body fluids and displaying immunosuppressive activities have been also described. Exosomes in milk and colostrum inhibit T cell activation in vitro . Similarly, exosomes in plasma of pregnant women bear FasL and reduce CD3ζ expression (hence responsiveness) by T cells . Another mechanism of immunosuppression has been observed for exosomes secreted by placental explants: they bear NKG2D ligands, and induce reduced cytotoxicity of NK and CD8+ T cells in vitro .
On the other hand, exosomes released by virally infected cells have been shown to have a number of deleterious effects on the host immune response and to favour spread of infection. HIV-infected cells could release exosomes with pro-apoptotic activity on uninfected bystander T cells . Epstein–Barr virus (EBV)-infected B cells or nasopharyngeal carcinoma secrete immunosuppressive proteins such as LMP1, and galectin-9 in exosomes [54, 92], which display T cell inhibitory activities.
Promotion of immune responses
Under other circumstances, secreted vesicles can have immuno-activatory properties. For instance, macrophages infected by various pathogens (Mycobacterium and Toxoplasma) release exosomes containing pathogen-derived pro-inflammatory molecular determinants inducing the secretion of pro-inflammatory cytokines by recipient macrophages . Interestingly, cultured cells infected by Mycoplasma, a very common pathogen, also release pro-inflammatory exosomes that induce polyclonal activation of B and T cells : this observation calls for careful consideration of studies where exosomes displayed B cell mitogenic or DC maturation activities, because of potential Mycoplasma contamination of the exosome-secreting cells.
Pro-immune activities of tumour-derived exosomes have also been reported, but mainly upon exposure of secreting cells to stress conditions: heat-shocked tumour cells release exosomes or membrane vesicles exposing hsp70, which promote NK cell activity  and TNF-α secretion by macrophages . Another group  also described secretion of more immunogenic exosomes by heat-shocked than control lymphoma or carcinoma cells. These exosomes were able to induce DC maturation and exhibited an increased T cell stimulating activity in the presence of APCs, but the basal activity of exosomes from control cells in these experiments could suggest occurrence of Mycoplasma contamination.
Finally, exosomes purified from body fluids could exacerbate autoimmune diseases: in rheumatoid arthritis patients, fibroblasts obtained from synovial fluid secrete exosomes bearing active membrane-bound TNF-α, which binds to T lymphocytes and renders them resistant to AICD , and in sarcoidosis patients, abundant exosomes present in the bronchoalveolar fluid display pro-inflammatory activity, such as increased secretion of IFN-γ and IL-13 by PBMCs, and IL-8 by epithelial cells .
Transfer of proteic and genetic materials between cells via exosomes
Although they do not directly affect immune responses, we think it is important to describe here other recently described unexpected effects of exosomes on recipient cells, mediated by transfer of materials contained in the exosomes. The first report of such an effect was published in 2008 , and it showed that microvesicles secreted by glioblastomas expressing the oncogenic EGF receptor EGFRvIII, purified using a classical protocole of exosome purification but not extensively characterized, contain this mutant receptor, and transfer it to the membrane of EGFRvIII-negative recipient cells, thus inducing activation of EGFRvIII-dependent transforming signalling pathways, and increased anchorage-independent growth [7, 8]. This observation is reminiscent of older observations showing receptor transfers mediated by larger membrane vesicles: for instance CCR5-bearing microparticles secreted by monocytes could transfer this receptor to CCR5-deficient blood mononuclear cells, which thus became sensitive to infection by a macrophage-tropic HIV .
In addition to proteins, exosomes have been shown to transfer also mRNA between cells. In 2007, the first study describing RNA in exosomes  provided evidences that, at least in vitro, and when high concentrations of exosomes were provided to cells, recipient human cells could translate mRNA from the mouse exosomes into mouse proteins, and thus, somehow, acquire new genetic materials. Although presence of RNA and their transfer to recipient cells had been previously published for larger microvesicles secreted by tumour cells, stem cells or endothelial progenitors [18, 50, 154], this first observation of RNA in exosomes prompted several groups to follow this lead, and analyse nucleic acids in exosomes from various sources. In 2008, presence of mRNA and miRNA was confirmed in exosomes purified from primary glioblastoma cells , and transfer and translation of a reporter mRNA from transfected cells to naive cells via exosome was also demonstrated.
Finally, transfer of miRNA via exosomes is nowadays the focus of intense scientific interest. In 2010, using a model of EBV-infected cells, Pegtel et al. showed that miRNA of viral origin were found in exosomes secreted by infected cells, and that their transfer to non-infected cells led to repression of known-target genes of these miRNA . Although miRNA-dependent repression of artificial target genes (miRNA-target promoter sequence fused to luciferase gene) upon cell incubation with exosomes was convincingly demonstrated in this work, it still does not show that repression of natural target genes can be mediated by exosome transfer, nor that the small (although significant) gene repression observed in recipient cells upon long term co-culture with EBV-infected cells, was due to exosomes, or even directly due to the miRNA present in exosomes, rather than to altered gene transcription mediated by mRNA or signalling proteins present in exosomes. Transfer of miRNA between cells is actually the subject of several recent papers, focusing or not on exosomes [100, 155, 207], and it will be very interesting to determine, in the next few years, what is the importance of exosomes in these transfer.
In vivo outcome of exosome-mediated immune responses
The pleiotropic actions of exosomes on individual immune and tumour cells in vitro can lead to various outcomes in the complex in vivo context. Here, we will try to summarize the observed effects of exosomes from different sources, after injection in vivo.
Exosomes secreted by antigen-presenting cells
DC-derived exosomes bearing MHC class I-tumour peptide complexes can induce the rejection of established tumours , although exosomes secreted by immature DCs only allow efficient tumour rejection when co-injected with mature DCs or chemical adjuvants . Exosomes from mature (but not immature) DCs pulsed with male antigen peptide speed up male skin graft rejection by female mice, showing in vivo priming and effector differentiation of activated CD4 T cells . Exosomes secreted by DCs can also induce humoral responses against the antigens that were fed to DCs before exosome purification [9, 41, 149], leading, in the case of the parasite Toxoplasma gondii, to strong protection against acute or congenital infection . In fact, feeding DCs with intact antigen (rather than peptide) allows them to secrete more efficient exosomes, able to stimulate both T and B cells in vivo, leading to both memory Th1 and immunoglobulin responses [22, 149]. Exosomes secreted by IFN-γ exposed OVA-pulsed intestinal epithelial cells can also induce OVA-specific humoral immune responses . Thus exosomes can induce effector immune responses in vivo.
In addition, careful evaluation of immune responses in the first cancer patients treated with exosomes from their own DCs  recently suggested a promoting effect of these exosomes on the patient’s NK cell activity . Indeed, exosomes secreted by mouse DCs were shown to bear both NKG2D ligands, and complexes of IL-15 and its receptor IL-15Rα, thus allowing simultaneous activation of NKG2D and transpresentation of IL-15 to NK cells, leading to activation of the cytotoxic activity , rather than inactivation, as observed with exosomes bearing NKG2DL in the absence of IL-15 (see above).
But conversely, MHC class II-bearing vesicles called “tolerosomes” purified from the supernatant of intestinal epithelial cells exposed to preprocessed OVA, or from the blood of antigen-fed rodents, induce a tolerogenic immune response, since they prevent development of OVA-specific allergic reaction to oral antigen . The reason for the discrepancy with the study by Van niel et al. is not clear . In addition, membrane vesicles secreted by DCs with a tolerogenic profile, i.e. either immature DCs , or DCs modified to express FasL, IL-10, CTLA-4Ig or IDO [26, 97], also induce tolerogenic, rather than effector immune responses. Such exosomes promote graft survival , reduce inflammation in a model of arthritis [96, 97], or of inflammatory-bowel disease .
Exosomes secreted by normal immature DCs also reduce inflammation in a model of septic shock , by promoting phagocytosis of sepsis-induced apoptotic cells by macrophages and reducing the release of pro-inflammatory cytokines , thanks to Mfge8/lactadherin present on these exosomes. Interestingly, another group showed that phagocytosis of apoptotic bodies by macrophages via Mfge8 promotes secretion of tolerogenic cytokines, such as IL-10 and TGFβ, by these macrophages, which orients T cell responses towards tolerogenic ones (such as generation of Tregs) . In this latter work, no link with exosomes was made, but since Mfge8/Lactadherin is strongly associated with small membrane vesicles such as milk fat globules , and exosomes , it is most likely that Mfge8 expressed by transfected cells in this work  was in fact secreted on exosomes. What remains unclear in the work analysing exosomes during septic schock [118, 119] is whether Mfge8/lactadherin present on exosomes has to be transferred from the exosomes to apopotic bodies to promote their phagocytosis, and if so, how this transfer happens!
In conclusion, in vivo, exosomes secreted by antigen-presenting cells can trigger either priming or tolerance, mainly depending on the activation state of antigen-presenting cells that produce them. In addition, it is possible that other factors, such as administration route, dose of transferred antigen, or in vivo micro-environnement, may determine the nature of exosome-induced responses. These considerations are especially important when designing therapeutic strategies involving exosomes.
Exosomes secreted by tumours
The literature describes contradictory functions in immune responses in vitro of vesicles secreted by tumour cells: on one hand transferring antigens to DCs to allow cross-presentation, on the other hand inhibiting several players of the effector step of anti-tumour immune responses. A synthesis of work by several groups on the effects of such vesicles after in vivo injection in mouse models, suggests that exosomes could promote anti-tumour immune responses, provided that they are administered together with strong adjuvants , and/or that they were purified from heat-shocked [35, 43], or transgenic cells expressing IL-2 or IL-18 . In general, “normal” exosomes alone do not allow induction of effector immune responses, although exosomes secreted by glioblastoma cell lines (in the absence of heat shock or any transfection) were recently shown to bear EGFRvIII and hsp70, and allow induction of anti-tumour immune responses in vivo without the need for adjuvant . By contrast, exosomes from other mouse tumour cell lines (carcinoma, thymoma and melanoma) when injected in mice decreased NK cell activity , and promoted differentiation of myeloid suppressor cells rather than DCs [202, 204]. In mice injected with carcinoma or melanoma exosomes in the absence of inflammatory signal before tumour challenge, tumour growth and metastasis were enhanced [105, 106], and increased metastasis could also be due to effects of exosomes on adhesion and migration of tumour cells in addition to effects on the immune system .
The reasons for the discrepancies of results published with different tumours are not yet clear. It is possible that some culture conditions of tumour cell lines may confer them immune properties absent from the parental tumours. In fact, a recent report  shows that tumour cells cultured in vitro for numerous passages become able to activate TLR2 on myeloid cells, thus promoting their immunosuppressive functions, whereas tumour cells freshly passaged in animals induce immunosuppressive myeloid cells in a TLR2-independent manner.
The in vivo effects of tumour-derived exosomes are still, thus, a matter of debate in the literature. Although several groups nowadays promote the idea that tumour exosomes dampen, rather than activate, immune responses, based on numerous immunosuppressive effects observed in vitro, one must keep in mind that tumour exosomes act on several arms of the immune system, and that the net result of these actions may be different with different tumours and/or at different stages of tumour progression.
In addition to APC or tumours, other cells can secrete exosomes displaying in vivo inhibiting activities on the immune system. Exosomes (or exosome-like vesicles) from bronchoalveolar lavage of allergen-tolerized mice transfer protection to allergy induction in other mice , and interestingly, this effect is not specific of the allergen .
Finally, a very interesting recent work has shown that a eukaryotic parasite, Leishmania, secretes exosomes, especially in culture conditions mimicking its entry into the mammalian host (i.e. 37°C instead of 25°C) , and that these exosomes orient human monocytes and dendritic cells towards tolerogenic profiles, secreting higher levels of IL-10 and lower levels of TNF-α : a parasite mutant for a heat shock protein, hsp100, by contrast, secreted exosomes with a slightly different overall protein composition, and which could promote induction of TH1 responses in vivo . These results thus show a new type of exosome-mediated interaction between a parasite and its host, with a mean to affect the hosts immune system.
Physiological relevance of exosomes
Existence of exosomes in vivo
Numerous descriptions of exosomes secreted by cells cultured in vitro call for data showing that such vesicles are produced in vivo. Recent reports have claimed the isolation of exosomes or exosome-like vesicles from dissociated lymph nodes  or thymus . Although their biochemical characteristics could fit with those of exosomes, these vesicles may have been artificially generated during the manual dissociation of tissues, from the plasma membrane of destroyed cells or from forcing open intracellular compartments, and thus cannot be taken as a proof of exosome existence in vivo.
But other observations provide more compelling evidences. For instance, nanometre-sized vesicles bearing MHC class II and tetraspan molecules were observed by electron microscopy in tonsil germinal centres, where they appear attached to the surface of follicular dendritic cells (which themselves do not express MHC class II) . In addition, exosomes have been purified in absence of any mechanical dissociation from several body fluids: presence of these vesicles shows that several cells can secrete them in vivo, especially tumour cells, endothelial cells, epithelial cells in mammary glands, urine or sperm ducts, and circulating cells in these fluids (mainly hematopoietic cells). Tumour cell-derived vesicles have been evidenced in plasma from tumour-bearing patients, with increased amounts in the most advanced cancers , showing that secreted vesicles can travel far away from secreting cells.
Potential physiological roles of in vivo exosome secretion
Even if secreted membrane vesicles exist in vivo, whether they play a physiological role is still elusive. So far, only correlative observations between exosome presence in vivo and a physiological situation are available. For instance, the group of D. Taylor  has shown that placenta-derived vesicles bearing FasL and MHC class II molecules were found in greater amount in women delivering at term than delivering pre-term. These vesicles inhibited T cell activation in vitro in a FasL-dependent manner, and greater inhibition was observed with vesicles obtained from women delivering at term. These results suggest that membrane vesicles circulating in the serum could participate in the prevention of maternal immune responses against the foetus.
In the pathological situation of cancer, the presence of exosomes bearing surface molecules of tumours in the serum of patients has been observed [77, 94], and increased amounts of these vesicles in more advanced cancers has been reported [23, 174]. But whether this increase in circulating tumoural exosomes is simply the cause of increased tumour size, or actually promotes tumour growth, cannot be determined. Since tumour exosomes have been shown both to transfer tumour antigens to DCs for induction of immune responses, but also to inhibit functions of various immune cells, and to promote angiogenesis and metastasis, the outcome of these contradictory activities in vivo probably depends on the balance of each of them, and this balance could change with the general immune state of the patient, and/or during progression of the tumour.
We have recently tried to address this question by analysing the role of secretion of an antigen on exosomes by tumour cells in a purely in vivo experimental setting, i.e. in the absence of any purification and concentration step of these vesicles in vitro . We have generated tumour cells secreting an antigen bound to small secreted membrane vesicles, due to its fusion to the phosphatidylserine binding domain of Mfge8. These tumours cells growing in vivo induce efficient CD8+ T cell activation, and thus do not grow efficiently in immunocompetent animals. Thus, in this mouse carcinoma model, secretion of membrane vesicles bearing an antigen allows induction of effector, and not tolerogenic, immune responses. In this experimental system, we may have tilted the balance in favour of the antigen-carrying functions of exosomes, which thus overcame the potential tolerogenic activities of exosomes. This work does, however, not prove an immunogenic role of exosomes in vivo since we cannot exclude that the modified antigen could bind to other types of small secreted vesicles within the tumour environment.
For DC-derived vesicles, an interesting hypothesis is their potential involvement in amplification of immune responses, by transferring MHC-peptide complexes from DCs that have been exposed to an antigen to other DCs that have not been in contact with the same antigen. Indeed, in various pathological mouse models [10, 55, 59, 158], a transfer of antigen acquired in the periphery by migrating DCs to lymph node resident CD8+ DCs has been shown to be required for induction of immune responses: whether exosomes play a role there will be worth investigating. In the context of organ transplantation, host DCs invading the grafted tissue have been recently proposed to migrate to spleen, and there transfer allopeptides captured from the graft to other host DCs, via exosomes . In this mouse model, MHC class II-peptide complexes present on exosomes would thus participate in recognition of the alloantigens from the graft by the host, but the outcome in terms of tolerance or rejection was not addressed.
How to address the physiological role(s) of exosomes?
The only way to conclusively demonstrate a role for exosomes in vivo would be to specifically inhibit or increase their secretion by a given cell type in vivo, and demonstrate that this affects the physiological outcomes, for instance in terms of immune response to a tumour or for formation and growth of tumour metastases. Unfortunately, we do not have yet adequate tools to achieve this goal. Indeed, a few molecules have been in the past proposed to affect exosome secretion, by one or another cell type, but either the vesicles affected by these molecules were not properly characterized as exosomes, or the molecules were known to modulate very general intracellular functions, and thus could not affect specifically the exosome pathway.
For instance, dimethyl-amiloride, an inhibitor of H+/Na + and of Na+/Ca2+ exchangers, induces a slight (−25%) decrease in secretion of Acetylcholine esterase by an erythroleukemia cell line , and by tumour cells , but since acetylcholine esterase has not been demonstrated as an exosome marker in any other cells than erythrocytes, and since it affects an ubiquitous ion channel, amiloride cannot be considered as a specific inhibitor of exosome secretion. Proton-pomp inhibitors, such as omeprazole, have been mentioned as potential tools to decrease exosome secretion but also affect tumour viability , hence cannot be used to address specifically the role of exosome secretion.
To eventually identify drugs affecting specifically exosome secretion, it is first necessary to identify the intracellular molecular machinery involved in their generation inside the multivesicular compartments, and in the fusion of these compartments with the plasma membrane. A few intracellular enzymes have been shown to play a role in secretion of vesicles: diacylglycerol kinase  or citron kinase  have been involved in regulated secretion of exosomes by T lymphocytes, but the relevance of these molecular mechanisms to the constitutive secretion of exosomes observed in other cell types, such as tumour cells, or dendritic cells is yet unclear. A guanine-exchange factor for ARF GTPases, called BIG2, is necessary for efficient release of exosome-like vesicles bearing TNF-RI from endothelial cells : these vesicles are secreted from the Golgi apparatus rather than from MVBs, and thus use a different molecular machinery than bona fide exosomes.
More recent studies aiming at identifying the molecular machinery of constitutive exosome biogenesis and secretion have recently started appearing in the litterature. In an oligodendroglial cell line overexpressing a myelin protein (proteolipid), the lipid ceramide was shown to be required for generation of internal vesicles of MVBs, and inhibition of its production by a drug inhibiting the enzymatic function of neutral sphingomyelinase (SW4869), or by siRNA inhibiting neutral sphingomyelinase gene expression, resulted in decreased exosome secretion . Whether this result can be used for secretion of exosomes by other cell types has not yet been convincingly addressed. More classically, the molecular machinery involved in the formation of internal vesicles of MVBs had been analysed extensively in yeast , and some of the results have been confirmed in mammalian cells. Three families of protein complexes, called ESCRT I, II and III, are involved in the different steps of internal vesicle formation, but whether inhibition of components of these complexes affect exosome secretion has not been directly analysed. Although it had been generally assumed that the same MVBs could either fuse with lysosomes for degradation of their internal vesicles, or fuse with the plasma membrane for exosome secretion, what regulates this dichotomic choice is still completely unknown. In fact, another recently proposed possibility  is that the multivesicular compartments where exosomes form to be eventually secreted, and those where transmembrane receptors are segregated in internal vesicles to be eventually degraded in lysosomes are different, and thus probably form using different molecular machineries. Indeed, new ESCRT-independent mechanims of segregation of receptors into these vesicles have recently been described in melanosomes , or via ceramide in oligodendrocytes .
For MVB transport towards and fusion with the plasma membrane, various studies have identified candidate molecules of the small GTPases, SNAREs, or molecular motor families involved in the transport, docking and/or fusion of secretory lysosomes with the plasma membrane [15, 168]. MVBs, however, even if they are part of the endocytic system like secretory lysosomes, are not the same compartments, and their secretion may or may not be dependent of the exact same machinery. We have recently shown that RAB27, a small GTPase previously associated with regulated trafficking or secretion of lysosome-related organelles in various cell types, such as melanosomes in melanocytes , secretory lysosomes in mast cells , cytotoxic granules in T lymphocytes  plays a crucial role in spontaneous exosome secretion by the HeLa tumour cell line , and that the two RAB27 genes, RAB27A and RAB27B play complementary roles and each use a specific effector in these cells. These results suggest that MVBs and lysosome-related organelles use some common molecular players, hence that similar molecules can act on close but not identical compartments in different cell types. In addition, in the oligodendroglial cell line where ceramide is involved in MVB formation, Rab35 and its effectors were shown to be crucial for exosome secretion, rather than Rab27 , whereas in the erythroleukemia cell line K562, Rab11 was involved in exosome secretion . Further work in the coming years will help clarify the situation, and determine whether different cells secrete exosomes coming from different intracellular compartments, or whether all these molecules play a role at a different step of the exosome secretion pathway. It will also be important to identify the molecules involved in the final step of exosome secretion, i.e. the fusion of multivesicular compartments wiht the plasma membrane.
Exosomes as therapeutic or diagnostic agents
Advantages of exosomes for immunotherapy
Exosomes could be considered as natural nanoparticular liposomes that have the ability to resist complement lysis  and resist RNAse attack allowing in vivo stability and protection of their nucleic acid content . As described above, exosomes are able to modulate adaptive and innate immunity and preclinical studies demonstrated that antigens have higher immunostimulatory capacities when carried by exosomes [21, 32, 41, 149, 205]. Moreover, we have set-up a method to produce exosomes according to good laboratory practice (GLP) and we have shown in the two first clinical trials that we could produce several vaccines (dendritic cell-derived exosomes) from one leukapheresis [51, 102, 129]. Exosomes, at least when derived from dendritic cells, are stable vesicles and can be cryo-preserved more than 6 months at −80°C with phenotype and function preserved. This stability gives them an advantage over whole living cells. Thus, the development of exosome-based immunotherapies remains very attractive. To our knowledge, no study has explored their biodistribution (and cellular uptake) or pharmacokinetics properties after injection in vivo and this should be addressed in future studies, always keeping in mind, however, that only accumulations of exosomes can be detected by fluorescence microscopy techniques. Future prospects should aim at designing and engineering synthetic exosomes for broader therapeutic indications, as has been recently proposed .
Exosomes as vaccine in infectious diseases
Immune properties of different types of exosomes suggest that they could be used as vaccines for infectious diseases. To our knowledge, however, no clinical studies testing exosomes in infectious deseases are yet ongoing.
Exosomal vaccines could be for instance developped from bacteria- and parasite-infected macrophages (e.g. Mycobacterium tuberculosis, M. bovis or T. gondii), since, as described above, such exosomes contain pathogen-derived antigens, activate antigen-specific CD4+ and CD8+ T cells , and induce pro-inflammatory cytokine production by naïve macrophages or dendritic cells [24, 25]. In the case of mycobacterial infection, exosomes released from infected macrophages activate pro-inflammatory pathways, but further investigation will be required to determine whether these vesicles favour control and clearance of the infection by the host. For parasite infections, on the other hand, exosomes derived from DCs pulsed with T. gondii or Leishmania major antigens were found to be protective against challenge infection with, respectively, T. gondii and L. major [9, 21, 159].
Alternatively, exosomes directly secreted by parasites display effects on the host immune system [162, 163]: although exosomes secreted by L. major or Leishmania donovani inhibit pro-inflammatory immune responses in vivo, and, when used to vaccinate mice, promote parasite growth , a mutant L. donovani (deficient for the heat shock protein HSP100) in contrast produced exosomes with a slightly different protein composition, which promoted Th1 immune responses, and did not exacerbate the disease: it would be worth investigating whether such mutant exosomes could, under some vaccination conditions, lead to actual protection against parasite infection.
Finally, membrane vesicles secreted by fungi (which may be plasma membrane-derived vesicles or exosomes) can enhance the secretion of pro-inflammatory cytokines by macrophages and enhance their uptake capacity , thus altering the fate of Cryoptococcus neoformans after phagocytosis by macrophages. These results strongly suggest the potential of these vesicular structures to modify the course of cryptococcal infections through their effects on host phagocytic cells. The authors proposed that these exosomal-like vesicules or components could have utility as fungal vaccines .
Therapeutic use of tolerogenic vesicles
No studies have yet been published evaluating exosomes in the clinic for their ability to induce immune tolerance. However, we can cite various studies demonstrating that under certain conditions dendritic cells can secrete exosomes that can induce tolerance. Between 2005 and 2007, Robbins et al. demonstrated that exosomes derived from immature DC treated or transduced with adenoviral vector expressing immunomodulatory cytokines, i.e. IL-10 and IL-4, or FasL were able to suppress inflammation in a murine model of delayed-type hypersensibility and reduce the severity of established collagen-induced arthritis [26, 95–97]. Likewise, exosomes were used for their anti-inflammatory effects in a rat cardiac allograft model: when bearing donor-MHC antigens, exosomes were able to delay the appearance of chronic rejection . Another work has shown that exosomes produced from TGF-β1- and IL-10 treated DCs, were efficient inducers of immune tolerance in a murine skin transplantation model . More recently, Robbins et al. have experimented in mice the use of exosomes from DC overexpressing Indoleamine 2,3-dioxygenase (IDO) as a novel therapy for rheumatoid arthritis . Altogether these studies suggested that immature DCs are able to favour tolerance rather than immunity and could serve in the future for the treatment of inflammatory diseases. In addition, protective effects of exosomes from bronchoalveolar lavage of allergen-tolerized mice would also suggest their use in treatments of allergies [147, 148].
Exosomes and their content as diagnostic markers?
The relative facility to collect exosomes (or other vesicles) from biological fluids, has suggested their use as diagnostic markers, easily obtained in a non-invasive manner. For instance, urine is a source of vesicles whose protein composition has been extensively studied in the search of markers of pathologies of the urinary system [1, 122, 133, 143, 164, 208]. In plasma, identification of tumour-derived membrane proteins in membrane vesicles is also currently proposed as a means to diagnose cancer [83, 91, 94, 108, 165, 175]. The actual use of exosome-derived proteins as biomarkers is, however, not yet implemented in clinical practice.
In 2008, mRNA and miRNA were identified in exosomes purified from the serum of glioblastoma  or ovarian carcinoma  patients, thus promoting the use of these vesicles and their RNA content as markers of cancer. Since then, several groups have published similar observations (i.e. identification of mRNA or miRNA in vesicles purified from blood of patients), for cancer or dental or neurological diseases [48, 120, 137, 152], but it remains unclear from all these studies whether the RNA content of exosomes is in any way more informative than that of other vesicles found in serum, especially microvesicles of around 1 μm in diameter: side-by-side analysis of the RNA content of different vesicle populations found in bodily fluids has still to be performed. The potential profits generated by identification of easily obtained biomarkers of widespread diseases in western countries have recently oriented biotech companies towards the exosome field of research: we will see in a few years whether their efforts will pay.
Exosomes in anti-tumour immunotherapies
Phase I trials
To date, very few clinical trials have tested the efficacy of exosomes as cell-free vaccine in patients and these trials were all realized in late-stage cancer patients [44, 51, 129]. However, the recent demonstration that dendritic cell vaccines against prostate cancer could have a benefit rekindles immunotherapeutic approaches [74, 167].
Three phase I clinical trials have been conducted (in China, France and the USA), involving the application of exosomes to elicit an immune response against established tumours. Dai et al. reported that using ascitis-derived exosomes when coadministered with granulocyte macrophage colony-stimulating factor (GM-CSF) in 54 patients with colorectal cancer induces a tumour antigen-specific cytotoxic T cell response, and that toxicity was minimal and well tolerated in patients . Escudier et al., using dendritic cell-derived exosomes (“Dex”), demonstrated the feasibility of scaling up and purification of clinical grade exosome using a GLP process, and observed restored number and NKG2D-dependent function of NK cells in 15 melanoma patients [51, 193]. Furthermore, a study by Morse et al. (also using Dex) reported a MAGE-specific T cell response and increased NK lytic activity in 13 patients with non-small cell lung cancer (NSCLC) . These phase I clinical studies demonstrated the feasability and the good tolerance of exosomes in late-stage cancer patients.
Ongoing phase II
During the two first Dex-based clinical trials, exosomes where produced from immature dendritic cells (iDex). Recent findings in mouse studies highlighted that iDex (see above) could preferentially induce tolerance. Since low levels of antigen-specific immune responses were detected in patients and based on the discovery that maturation stage of DCs could influence exosomes’ immune properties, we produced second generation Dex (from IFN-γ-treated DCs) with boosted immune properties . We showed that IFN-γ was a suitable maturation agent for monocyte-derived DCs to produce immunogenic exosomes. Indeed, the quality of the exosomal vaccine is improved in three directions: while the previous manufacturing process provided exosomes (1) with low expression of CD40, CD86 and ICAM-1 molecules , both critical for T cell priming and/or the DC/T cell interactions , (2) requiring a transfer onto host APC to elicit T cell responses in preclinical assays [12, 32], (3) requiring a direct pulsing of exosomes with peptides after acid elution of MHC class I molecules [51, 102, 129], the novel process using IFN-γ to mature DC resulted in “immunogenic” exosomes with a significant enrichment in co-stimulatory and ICAM-1 adhesion molecules, and which could promote direct activation of a Mart-1 specific CD8+ T cell clone and of NK cells in vitro, and the priming of CD8+ T cells in vivo . Based on these results, we set-up to conduct a phase II trial (at Curie and Gustave Roussy Institutes, France) aiming at assessing progression-free survival in non operable and advanced NSCLC patients (stages IIIB and IV) after vaccination with autologous DC-derived exosomes (DEX² trial). Preclinical and pilot studies in end-stage patients indicated that metronomic cyclophosphamide is a reasonable option to enhance Dex vaccine immunogenic potential  and reverse tumour-induced immune tolerance (such as T cell anergy and NK cell deficiencies , This phase II clinical trial is testing the clinical benefit of γDex as a maintenance immunotherapy responding to or stabilized with cis-platine-based induction chemotherapy. Here, γDex are loaded with two HLA-DP04-restricted and four HLA-A2.1-restricted peptides. HLA-A2.1+ chemosensitive patients will be eligible to receive γDex-based maintenance immunotherapy consisting in 3 weeks oral therapy by metronomic cyclophosphamide  followed by i.d. injections of γDex vaccines, weekly for 4 weeks (induction immunotherapy) and every 2 weeks for 6 weeks (continuation immunotherapy) . Forty-one patients will be enrolled between June 2010 and December 2012. The primary objective is to ameliorate progression-free survival at 4 months post-chemotherapy. Secondary objectives are the clinical efficacy of γDex (assessed as overall survival, objective response rates), biomarkers of efficacy (NK cell activation, restoration of NKG2D expression and peptide vaccine-specific T cell responses) and the safety of γDex.
We hope we have convinced the reader that exosomes represent a subclass of secreted membrane vesicles with numerous specific immune functions, and numerous potential applications in pathologies. Many questions remain however, especially concerning their actual physiological roles, but the development, in the past 5 years, of the scientific community working on these fascinating vesicles will hopefully help address these issues and lead, in the next years, to major advances in understanding their functions. A major challenge will also consist in providing proper comparisons of the properties and functions of exosomes and other classes of secreted membrane vesicles.
The authors thank Fondation de France, Agence Nationale pour la Recherche, and Institut National du Cancer for funding their research.