ICAM-1 on exosomes from mature dendritic cells is critical for efficient naive T-cell priming

Dendritic cells (DCs) are key actors in the initiation of immune responses.^1,2  Immature DCs present in peripheral tissues constantly sample their microenvironment, thus acquiring antigens. Under steady-state conditions, antigens are eventually processed into peptides, which, when loaded onto major histocompatibility complex (MHC) molecules, can be recognized by specific T lymphocytes. Immature DCs, however, do not induce effective immune responses, but rather are involved in maintaining peripheral T-cell tolerance. By contrast, under inflammatory conditions, DCs gradually acquire high numbers of MHC-peptide complexes and costimulatory molecules, and become the most efficient inducers of T-cell–dependant immune responses.

Immature DCs, like several other cell types, secrete small membrane vesicles called exosomes (50 nm to 100 nm diameter) that form within late multivesicular endosomal compartments (MVBs).^3-5  Exosomes are released in the extracellular environment by fusion of MVBs with the plasma membrane. Immature DC–derived exosomes bear functional MHC class I– and class II–peptide complexes, which can induce activation of specific T cells in vivo.^6,7  In vitro, exosomes are used as a source of T-cell receptor (TCR) ligands, or of antigens, by recipient DCs which then activate antigen-specific T cells.^6,8,9  The costimulatory molecules, such as B7.2, required for T-cell activation, however, are contributed by DCs, and not by exosomes.

These results suggest that exosomes are involved in the spreading of MHC class II–peptide complexes between DCs. Such spreading should increase the number of DCs expressing the ligand for a given antigen-specific T cell. The functional outcome of indirect T-cell activation via exosomes remains, however, unclear. Immature and mature DCs induce different functional T-cell responses (ie, tolerance versus priming). Whether exosomes induce T-cell tolerance, or priming, or none, could therefore depend on the maturation state of DCs producing the exosomes, and/or of DCs receiving the exosomes. We asked here whether exosomes secreted by immature or mature DCs were functionally different.

We have compared exosomes produced by lipopolysaccharide (LPS)–treated (mature) and untreated (immature) DCs. Immature and mature DCs secrete morphologically similar exosomes. In functional T-cell stimulation assays, exosomes from mature DCs induced strong antigen-specific T-cell proliferation both in vitro and in vivo, even when used at very low doses. Exosomes from mature DCs conferred the ability to activate naive T cells to inefficient antigen-presenting cells (APCs) (ie, B cells), whereas exosomes from immature DCs were unable to do so. Exosomes secreted by mature DCs, but not those secreted by immature DCs, were able to prime naive T cells in vivo, as shown by their effect in a minor antigen-bearing skin graft assay. At the molecular level, mature DC–derived exosomes bear increased amounts of MHC class II and costimulatory molecules, and a decreased amount of milk fat globule–epidermal growth factor (EGF)–factor VIII (MFG-E8). We demonstrate that decreased MFG-E8 or increased B7.2 on mature DC–derived exosomes do not account for their stronger activity. By contrast, intercellular adhesion molecule 1 (ICAM-1) is necessary for the immune activity of exosomes secreted by LPS-treated DCs.

C57Bl/6 wild-type (WT) and mutant mice were obtained from Charles River (C57Bl/6; L'Arbresle, France), O. Lantz (Marilyn; Institut Curie, Paris, France), CDTA (I-Aβ^–/–; Orléans, France), B. Salomon (B7^–/–; Pitié-Salpêtrière, Paris, France), Jackson Laboratories (Icam-1^–/–; Bar Harbor, ME). Mice were housed in specific pathogen-free conditions. Experiments were done in accordance with the guidelines of the French Veterinary Department.

The DC line D1¹⁰  was cultured as previously described¹¹  in 10% fetal calf serum (FCS) complete Iscove modified Dulbecco medium (IMDM; Sigma, St Quentin, France), supplemented with 15% granulocyte macrophage–colony-stimulating factor (GM-CSF)–containing, J558-conditioned medium. Bone marrow–derived DCs (BMDCs) were generated by 10 to 14 days of culture in complete IMDM with 30% J558-conditioned medium,¹²  and used when fluorescence activated cell sorting (FACS) analysis revealed more than 75% CD11c⁺ cells. Maturation was induced by a 16-hour treatment with 5 μg/mL LPS (Sigma). DCs were purified from spleens using anti-CD11c–coupled magnetic beads, B cells were purified from the CD11c-negative fraction with anti-CD19 beads (Miltenyi Biotec, Paris, France).

Fluorophore-coupled antibodies used for FACS analysis were as follows (all from BD PharMingen, Le Pont de Claix, France): anti-CD11c (HL3 clone), I-A^b (AF6-120.1), CD40 (3/23), CD86 (GL1), B220 (RA3-6B2), CD69 (H1.2F3), CD45.1 (A20), CD4 (RM4-5), and the corresponding isotype controls. Tricolor-conjugated streptavidine was from Caltag (Burlingame, CA). Antibodies used for exosome characterization by Western blotting were: rabbit antisera anti–mouse MHC class II α chain C-terminus,¹²  MHC class I (P8; Dr Ploegh, Boston, MA), MFG-E8 (prepared against LVTLDTQRGDIFTEY peptide from mouse MFG-E8,¹³  mouse monoclonal anti–mouse clathrin (23 clone; BD Transduction Laboratories, Le Pont de Claix, France), annexin II (Dr Gerke, Münster, Germany), flotillin-1 (18 clone; BD Transduction Laboratories); goat anti–IgG 101 (M-19 clone; Santa Cruz Biotechnology, Santa Cruz, CA); rat monoclonal anti–mouse CD9 (KMC8 clone; BDPharmingen), gp96 (9G10 clone; Stressgen, Victoria, Canada), hsc70 (1B5 clone; Stressgen), and CD86 (GL1 clone; BDPharmingen). Horseradish peroxidase (HRP)–conjugated secondary antibodies were from Pierce (Perbio, Brebieres, France). Carboxyfluorescein diacetate succinimidyl ester (CFSE) was from Molecular Probes (Eugene, OR). HY (NAGFNSNRANSSRSS) and Eα56-73 (ASFEAQGALANIAVDKA) peptides were from Neosystem (Strasbourg, France).

Exosomes were recovered from the supernatant of cells cultured for 24 hours in depleted medium (ie, medium depleted from FCS-derived exosomes by overnight centrifugation at 100 000g).¹¹  Exosomes were purified by filtration on 0.22-μm pore filters, followed by ultracentrifugation at 100 000g as described.¹¹  In each exosome preparation, the concentration of total proteins was quantified by Bradford assay (BioRad, Marnes la Coquette, France). Exosomal proteins (3 μg to 5 μg) were analyzed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and Western blotting or coated on beads for FACS analysis as described previously.¹¹  To obtain exosomes bearing MHC-peptide complexes, immature or mature DCs were pulsed for 3 hours in complete medium with 250 nM HY peptide (or 50 μM Eα56-73), before the 24-hour culture period in depleted medium.⁶ 

Exosomes fixed in 2% paraformaldehyde (Carlo Erba, Rodano, Italy) were loaded on Formwar/carbon-coated EM grids. Samples were postfixed in 1% glutaraldehyde, contrasted and embedded successively in 2% uranyl acetate, pH 7, and 2% methylcellulose/0.4% uranyl acetate, pH 4. EM samples were observed at 80 kV with a CM120 Twin Phillips electron microscope (FEI company, Eindoven, The Netherlands). EM negatives were scanned at 400 dpi with a HI-SCAN drum scanner (Eurocore, Heidelberg, Germany) and processed with ColorBrain 3.1 software (Eurocore) and Adobe Photoshop 8.0 software (Adobe, San Jose, CA).

Exosomes (30 μg) were separated by a 16-hour centrifugation at 100 000g on a linear 2 M to 0.25 M sucrose gradient as previously described.¹²  Pellets recovered from 1-mL fractions were analyzed by SDS-PAGE and Western blotting.

Slices 1-mm wide were cut out from Coomassie blue–stained gels, and were reduced, alkylated, and trypsin digested as previously described.¹⁴  Extracted peptides were dried and resolubilized in 95:5 water/acetonitrile. One third of the concentrated protein digests were separated on an LC-Packing system (Dionex, Sunnyvale, CA) coupled to the nanoelectrospray II ionization interface of a QSTAR/Pulsari (Applied Biosystems, Foster City, CA). The MS/MS data were searched using MASCOT software (Matrixscience, London, United Kingdom) against the National Center for Biotechnology Information database (Bethesda, MD). MASCOT files were extracted to a software interface PROMS (Bioinformatics, Institut Curie, Paris, France), and each peptide was validated manually.

As previously described,⁶  exosomes or HY peptide were incubated in 96-well plates with 2 × 10⁴ recipient cells (ie, LPS-treated BMDCs, purified spleen DCs, or B cells) in 50 μL complete IMDM for 3 to 4 hours at 37°C. Plates were irradiated (5000 rad) and 5 × 10⁴ Marilyn lymph node cells were added in 100 μL complete IMDM. ³H-thymidine (0.5 μCi/well [18.5 KBq/well]) was added 72 hours later for 16 hours. Cells were harvested on glass fiber filters (Tomtec, Hamden, CT), and ³H-thymidine incorporation was counted by liquid scintillation. Alternatively, cells were harvested 24 hours after addition of Marilyn T cells, and stained with anti-CD45.1, anti-CD4, and anti-CD69 antibodies before FACS analysis.

Female C57Bl/6 mice received intraveneous injections of 10⁶ CFSE-labeled (5 μM in phosphate-buffered saline [PBS]–0.1% bovine serum albumin [BSA], 8 minutes, 37°C) CD45.1⁺ Marilyn lymph node cells. At 24 hours later, exosomes or DCs were injected in the hindfootpad. Draining lymph nodes were analyzed 4 days later by FACS. The amount of exosomes injected ranged from 1 μgto15 μg of total proteins.

Female C57Bl/6 mice received footpad injections of PBS (50 μL) or 5 μg HY-bearing immature or mature exosomes. The next day, 1-cm² fragments of belly skin from male C57Bl/6 mice were grafted onto the backs of female recipients. Plasters were removed after 6 days, and from then on, only mice with accepted grafts were analyzed. Grafts were macroscopically examined every other day, and were considered rejected when more than 80% of graft surface appeared necrotic.

Exosomes produced during 24 hours were purified from supernatants of LPS-treated or untreated DCs (D1 cell line or BMDCs), using the usual filtration and ultracentrifugation-based exosome purification procedure.¹¹  Exosomes from LPS-treated DCs will be called mature exosomes, as opposed to immature exosomes produced by untreated DCs. As described before,¹²  LPS-treated DCs produced on average 2- to 3-fold fewer exosomes than untreated DCs (0.2 ± 0.1 μg/million mature cells, versus 0.5 ± 0.1 μg/million immature cells). The morphology of mature exosomes observed by electron microscopy (Figure 1A), their density on sucrose gradients (Figure 1B), as well as their overall protein composition, as seen on a coomassie blue–stained gel (Figure 1C) were similar to that of immature exosomes. Thus, although mature DCs lose part of their internal multivesicular endosomes, which become elongated tubular structures,¹⁵  the remaining MVBs are still a source of exosome secretion.

We analyzed whether exosomes from mature DCs would display increased immunogenicity as compared with immature exosomes. A DC-dependent antigen-presentation assay was used, as previously described.⁶  Immature or LPS-treated DCs were incubated for 3 hours with HY peptide, derived from the male antigen Dby, before the exosome production phase. Purified exosomes were incubated in vitro with mature BMDCs for 3 hours, before addition of HY-specific naive CD4 T cells from Marilyn mice.¹⁶  T-cell activation was assessed by measuring up-regulation of the early activation marker CD69 after 24 hours (Figure 2A), and T-cell proliferation after 3 days of culture (Figure 2B). In both assays, mature exosomes induced similar or higher plateau of T-cell activation as peptide directly loaded on DCs, whereas immature exosomes were 50- to 100-fold less efficient. Mature I-A^b/HY–bearing exosomes induced T-cell proliferation when loaded on MHC class II–deficient DCs, which cannot activate T cells in the presence of peptide alone (Figure 2C). In this setting as well, mature exosomes were 100-fold more efficient than immature exosomes. Exosomes from mature DCs that had not been exposed to HY peptide (Figure 2A), or exosomes from mature MHC class II–deficient BMDCs pulsed with HY peptide (Figure 2B) did not induce T-cell proliferation. The weak immature exosome activity was not increased by addition of LPS (Figure S1A, available on the Blood website; see the Supplemental Materials link at the top of the online article) or of mature exosomes that did not bear HY peptide (Figure 2D). We conclude that T-cell activation and proliferation are more efficiently induced by mature than immature exosomes. Strong immunogenicity of mature exosomes is not due to nonspecific contaminants (ie, LPS or free HY peptide) in the mature exosome preparations, but rather to the actual composition of the vesicles secreted by mature DCs.

Immature exosomes can be used in vitro by other DCs (BMDCs or spleen-derived DCs) to activate CD4 T cells, but not by other APCs such as B cells or macrophages.⁶  It is unclear whether this lack of naive T-cell priming is due to the inability of B cells or macrophages to bind exosomes, or to the lack of the appropriate costimulation molecules in these antigen-presenting cells. To determine whether mature exosomes behaved like immature exosomes in this respect, CD11c⁺ DCs and CD19⁺ B cells from female mice spleens were used as exosome recipients in the in vitro T-cell activation assay. Both cell types expressed similar levels of MHC class II, and expressed the costimulatory molecule CD40 (Figure 3A). Spleen B cells did not induce naive CD4 T-cell proliferation in the presence of HY peptide (Figure 3B) or immature exosomes (Figure 3D) in conditions where splenic DCs induced efficient T-cell activation (Figure 3B-C). Mature exosomes, by contrast, were readily presented by spleen B cells and induced potent T-cell proliferation (Figure 3D). Thus, mature exosomes can transfer to inefficient APCs (ie, B cells) the unique ability of DCs to activate naive T cells.

The effects of mature exosomes in vivo were next analyzed as previously described.⁶  Immature or mature HY-bearing exosomes (15 μg) were injected in the footpads of female C57Bl/6 mice adoptively transferred one day before with CFSE-labeled Marilyn lymph node cells. Draining lymph nodes were harvested 4 days later, and HY-specific T-cell proliferation was evidenced by a 2-fold decrease of CFSE intensity in daughter cells at each division cycle (Figure 4A). Although both immature and mature HY-bearing exosomes induced T-cell divisions, a larger proportion of T cells underwent extensive proliferation (more than 5 division cycles) upon mature exosome injection (Figure 4A). When injected at lower doses, mature exosomes still induced proliferation of Marilyn cells very efficiently, even with as little as 1 μg per mouse, whereas 10- to 20-fold higher concentrations of immature exosomes were needed to induce similar proliferation (Figure 4B). Exosomes from mature DCs that had not been exposed to HY peptide did not induce CD4 T-cell proliferation (Figure 4B). Coinjection of immature exosomes with LPS also did not induce more efficient T-cell activation (Figure S2). Mature exosomes are therefore more efficient than immature exosomes to activate naive T cells in vivo.

To determine whether exosomes induce effector T cells in vivo, we used an autologous skin graft model. DCs loaded with HY peptide (ie, those which activate I-A^b-HY–specific CD4 T cells), have been previously shown to accelerate male skin graft rejection by female mice.¹⁷  We thus tested the effect of immature or mature I-A^b-HY–bearing exosomes on male skin graft rejection by autologous female mice. In mice treated with immature exosomes one day before graft, the kinetics of graft rejection was not significantly different from the kinetics in control mice injected with PBS (Figure 4C). By contrast, mature exosomes induced significantly faster skin graft rejection (Figure 4C). This effect was not due to contaminants (for instance residual HY peptide), since exosomes from MHC class II–deficient DCs pulsed with HY peptide did not accelerate the skin graft rejection, as compared with PBS injection (data not shown).

These results show that mature exosomes are more efficient than immature exosomes for the induction of T-cell activation in vivo. For the induction of effector CD4⁺ T cells, however, only mature exosomes are effective.

To identify the molecular bases of the strong immunogenicity of mature exosomes, we analyzed their protein composition. Extensive proteomic analysis by tandem mass spectrometry identified 93 new proteins that had not been evidenced before in DC-derived exosomes^11,18  (Table 1). Some of these proteins had also been identified in exosomes from other cell types (Table 1; see Table S1 and corresponding references for exhaustive results). All these proteins were found both in immature DC– and mature DC–derived exosomes. Their respective amounts in each type of exosome could not be assessed by the proteomic approach. Western blotting was therefore performed to compare quantitatively immature and mature exosomes. Gp96, a resident protein of the endoplasmic reticulum used as a control for endoplasmic reticulum contamination, was absent from both types of exosomes (Figure 5A). Several proteins were present in equal amounts in immature and mature exosomes (Figure 5A). On the other hand, several proteins that are modulated during DC maturation were similarly modulated in exosomes (Figure 5B): MHC class II and costimulatory molecules B7.2/CD86 and ICAM-1/CD54 were strongly increased in mature exosomes as compared with immature exosomes. MFG-E8/lactadherin, which is down-regulated in mature DCs¹⁹  (P.V., E.S., Gaël Sugano, S.A., and C.T., manuscript submitted, May 2005), was also much less abundant on mature exosomes. Finally, MHC class I and tsg101, which are respectively up-regulated and not affected by LPS treatment, were both slightly down-modulated in mature exosomes (Figure 5B). The protein composition of exosomes produced by DCs incubated with LPS for 3 hours, rather than 16 hours, was intermediate between that of immature and mature exosomes, in terms of MHC class II, CD86, and MFG-E8 content (data not shown).

Mature exosomes contain 2 to 3 times more MHC II molecules (Figure 5B; Figure S3), and only 1.5 times more specific MHC II–peptide complexes (Figure S3), than immature exosomes. Increased number of MHC-peptide complexes on mature exosomes could therefore not account for their stronger efficiency in activating T cells. Among exosomal proteins known to be involved in immune responses, 3 were strongly modulated in mature exosomes: MFG-E8, B7.2, and ICAM-1. We thus analyzed the role of these proteins in exosome-induced T-cell activation.

MFG-E8 is involved in phagocytosis and destruction of apoptotic cells by macrophages.²⁰  MFG-E8 present on exosomes could thus mediate targeting of exosomes to DCs. To test this hypothesis, exosomes were purified from MFG-E8–deficient BMDCs. MFG-E8–deficient mice obtained from embryonic stem (ES) cells bearing a genetrap vector in the Mfge8 gene²¹  are described elsewhere.¹³  DCs obtained in vitro from BM precursors of Mfge8^–/– mice presented identical differentiation kinetics and phenotype as those obtained from WT animals, and induced identical activation of CD4 T cells in the presence of HY peptide (data not shown). Apart from the absence of MFG-E8, exosomes purified from immature Mfge8^–/– BMDCs had the same protein composition as exosomes from WT BMDCs (Figure 6A). In the in vitro presentation assay, immature Mfge8^–/– exosomes did not induce stronger T-cell proliferation than WT exosomes (Figure 6B). Note that the mature BMDCs used as recipient cells do not express MFG-E8¹⁹  (P.V., E.S., Gaël Sugano, S.A., and C.T., manuscript submitted, May 2005), and therefore could not compensate for the absence of MFG-E8 on exosomes. Thus, absence of MFG-E8 on mature exosomes does not explain their increased biologic activity.

The roles of B7.2 and ICAM-1 on exosomes were analyzed next. As compared with WT DCs, DCs from B7.2/CD86 and B7.1/CD80 double knock-out mice (B7^–/– DCs) need 3 times more HY peptide (Figure 7A), and DCs from Icam1/CD54 knock-out mice need 6 times more HY peptide (Figure 7B), to activate Marilyn T cells in vitro. Both molecules thus participate in CD4⁺ T-cell activation.

Mature exosomes devoid of B7.2 were purified from B7^–/– BMDCs. Their protein composition was similar to that of exosomes from WT DCs, except for the absence of B7.2 (Figure 7C). In vitro, in the presence of T cells and recipient WT spleen DCs (Figure 7D), B7^–/– exosomes induced similar T-cell proliferation as WT exosomes. When using B7^–/– DCs as recipient of WT and B7^–/– exosomes, we also observed similar efficiencies of the 2 types of exosomes (Figure 7E), although T-cell stimulation was lower than with WT DCs. Therefore, B7.2 expressed by mature exosomes does not account for their increased immunogenicity as compared with immature exosomes. By contrast, mature exosomes produced by Icam-1^–/– BMDCs did not induce efficient T-cell activation, when fed to ICAM-1⁺ recipient DCs (Figure 7F) or to B cells (Figure S4). ICAM-1 expressed on recipient DCs did not overcome the defect of ICAM-1–deficient exosomes. Expression of ICAM-1 on exosomes is therefore critically required for stimulation of CD4⁺ T cells.

Finally, B7^–/– DCs induced less efficient T-cell activation than WT DCs in the presence of mature WT exosomes (Figure 7D-E), whereas the difference between WT and Icam-1^–/– DCs was not as striking (Figure 7F-G). We concluded that recipient DCs can use ICAM-1 present on exosomes, but need their own B7.2 molecules, for efficient T-cell activation.

The results presented here show that LPS-treated DCs (ie, mature DCs) secrete bona fide exosomes, as defined by their morphology, qualitative protein composition, and buoyant characteristics. Exosomes from mature DCs, however, are more potent for antigen-specific T-cell activation than exosomes from immature DCs. Three major findings on exosomes produced by mature DCs are reported here: (1) these exosomes induce effector T-cell responses, since they induce rapid rejection of antigen-bearing grafts, (2) they can transfer not only MHC-peptide complexes, but also T-cell–stimulating ability to poorly efficient APCs, such as B cells, and (3) expression of ICAM-1 on mature exosomes is necessary for their immunogenicity.

We have previously reported that immature exosomes induce in vivo antigen-specific CD4 T-cell activation and proliferation.⁶  However, the functional outcome of this early activation was not evaluated, and could have been either an effector or tolerogenic response. Another group has shown that immature DC–derived exosomes induce efficient CD8-dependent antitumor immune responses, but only when coinjected with an adjuvant.⁷  In a heart allograft model, repeated intravenous injections of immature exosomes led to tolerance.²²  Here, we have used a skin graft assay, which allows detection of tolerogenic and effector responses. In the protocol used here, CD4⁺ T-cell activation induced by immature exosomes did not give rise to either effector or tolerogenic responses, since skin graft rejection was not significantly affected by these exosomes. By contrast, a single injection of mature exosomes induced effector T-cell generation, and led to fast skin graft rejection in the mouse model system. Using immature exosomes to inhibit immune responses in autoimmune disease or transplantation settings will therefore need careful examination of the treatment conditions, such as injection sites, number of injections, and amount of exosomes.

Since we have first described antitumor immune responses induced by DC-derived exosomes in mice,²³  the use of exosomes in clinical antitumor immunotherapy has been a strong focus of interest. As opposed to skin rejection, tumor rejection needs strong CD8⁺ T-cell response, enhanced by efficient CD4⁺ T-cell help. Whereas MHC class II was increased on mature exosomes, MHC class I was slightly down-regulated. Experiments are being perfomed to evaluate the impact of mature exosomes on tumor growth in mice. It is also important to stress that human and mouse DC-derived exosomes are not identical. Exosomes purified from monocyte-derived human DCs express ICAM-1 and B7.2,^24,25  and probably display an intermediate phenotype between our “immature” and “mature” mouse DC–derived exosomes. Human clinical-grade exosomes,²⁵  loaded ex vivo with peptides, activate both CD4⁺ and CD8⁺ T cells in vitro.²⁶  These results suggest that exosomes from human monocyte–derived DCs are efficient to activate T cells.

The way exosomes induce antigen-specific immune responses is not yet elucidated. We and others have shown that MHC class I– and class II–peptide complexes present on exosomes can be used by recipient DCs to activate antigen-specific T cells.^6,9,26  In these studies, other APCs such as B cells or macrophages, which could not activate naive T cells unless high doses of peptide were used, could also not use exosomes as a source of MHC-peptide complexes to activate T cells. A striking property of exosomes produced by LPS-treated DCs is that they endow B lymphocytes with the ability to activate naive T lymphocytes. As described by others,²⁷  we also observed some T-cell activation by exosomes from mature DCs in the absence of recipient APCs (data not shown). This activation, however, was weak and only occurred when using high amounts of exosomes, over 3 μg, equivalent to the amount secreted by about 10 million mature DCs. In the presence of APCs, on the other hand, T-cell activation was strong with as few as 0.1 μg exosomes. These observations suggest that, rather than directly presenting MHC-peptide complexes to naive T cells, exosomes physiologically transfer them to recipient APCs, thus activating T cells. Secretion of exosomes by mature DCs in the lymph node could allow surrounding B cells to acquire MHC-peptide complexes, and so provide chances for a T lymphocyte to encounter its specific ligand and become activated in a broad area of the lymph node (ie, also at the B-cell/T-cell area junction). On the B-cell side, however, the outcome of interaction with a T cell via exosomal MHC-peptide will not necessarily lead to B-cell activation. Indeed, to induce full B-cell maturation and antibody production, repeated stimulation via MHC molecules must occur, especially on daughter B cells resulting from proliferation. Exosomal MHC molecules are probably not available in sufficient amounts to allow this repetitive B-cell activation, and would therefore not result in nonspecific antibody response.

Exosomes secreted by mature DCs are thus small vesicles carrying the exquisite ability of mature DCs to induce T-cell responses. Exosomes do not act as mere adjuvants, since only exosomes bearing both MHC-peptide complexes and other factors from mature DCs induce efficient T-cell activation. Among several molecules differentially expressed by exosomes from immature versus LPS-treated DCs, one molecule was identified as responsible for this increased immunogenicity: ICAM-1. Exosomal ICAM-1 can compensate for a lack of this adhesion molecule on APCs, whereas B7.2 has to be expressed by the APC to efficiently induce T-cell activation. These 2 molecules thus play complementary roles on T-cell activation. As a ligand for the lymphocyte function-associated antigen (LFA-1) and Mac-1 (CD11b/CD18) integrins (which are expressed by DCs, B cells, and T cells), ICAM-1 could either help with capture of exosomes by the recipient APCs, or favor T-cell binding to the recipient APC bearing exosomes at its surface. Endocytosis of immature DC–derived exosomes by immature DCs, and subsequent reprocessing of exosomal antigens by DCs has been shown.¹⁸  In the present work, however, recipient APCs that use exosomes to activate T cells are mature DCs and B cells displaying poor endocytosis activity. In addition, presence of preformed MHC-peptide complexes on exosomes allows T-cell activation by recipient APCs without a need for antigen reprocessing. Exosomes from mature DCs could thus be presented to T cells as whole MHC- and ICAM-1–enriched microdomains at the surface of recipient APCs. To do so, exosomes could either fuse with the plasma membrane, or simply stick to the surface of these recipient APCs. Analysis of interaction among exosomes, DCs, and B cells in the presence or not of T lymphocytes, will be performed to clarify this issue.

In any case, our observation that MFG-E8 associated to exosomes does not play a role in their immune function is consistent with a minor role of phagocytosis in exosomal antigen presentation by DCs. MFG-E8 has been proposed to allow phagocytosis, by macrophages and DCs, of phosphatidyl-serine (PS)–containing vesicles, such as apoptotic cells.^19,20  Exosomes also expose PS at their surface.¹⁸  Here, absence of MFG-E8 on exosomes did not alter in any way the subsequent T-cell activation. We also did not observe either impairment or improvement in T-cell immune responses to antigens associated to apoptotic cells or exosomes in the MFG-E8–deficient mice in vivo (C.T., P.V., unpublished results, September 2004). MFG-E8 is therefore not an essential targeting factor for addressing antigens to DCs in vivo.

The physiologic relevance of MHC-peptide or antigen transfer via exosomes in vivo remains controversial. Existence of exosomes in vivo has now been extensively documented, by purification from various bodily fluids (ie, serum,²⁸  malignant effusions,^29,30  bronchoalveolar fluid,³¹  and urine³² ), and by electron microscopy of human tonsils.³³  Exosomes purified ex vivo from malignant effusions are sources of tumor antigens, used by DCs for cross-presentation in vitro.²⁹  Transfer of antigens via exosomes in vivo could take place in the thymus, where DCs acquire antigens from medullary epithelial cells, and thus allow autoreactive T-cell deletion in vivo.³⁴  It could also allow uninfected CD8⁺ DCs to acquire herpes simplex virus antigens from viable Langherans cells emigrating from infected skin.³⁵  Direct demonstration that antigen transfer via exosomes actually takes place in vivo is still lacking, however. This demonstration would require a way to specifically inhibit exosome production, which is not yet possible. Further understanding of the molecular mechanisms of exosome secretion could make this possible in the future.

Transfer of preformed MHC-peptide complexes by exosomes to other APCs, is specific of exosomes secreted by DCs. Such a transfer could explain how endogenous DCs acquire peptide from injected exogenous DCs, and thus allow efficent T-cell activation.^36,37  Recent reports seem to argue against the relevance of such a mechanism. None of them, however, provides a definite proof of the absence of MHC-peptide transfer between DCs. Spörri et al³⁸  have observed that antigen-specific T cells induce full maturation of cognate DCs, but only partial maturation of bystander DCs, which do not bear the TCR-specific ligand. In the herpes simplex virus model, analysis of mixed MHC class II–deficient × MHC class I–mutant hematopoietic chimeras revealed that efficent CD8⁺ T-cell priming requires presentation of CD4 and CD8 T-cell ligands on the same DC.³⁹  Both observations could be interpreted as a proof of inefficient transfer of MHC-peptide complexes between neighboring DCs in vivo. Another interpretation, however, is that exosomes bound to recipient DCs allow antigen-specific T-cell activation, but do not lead to reverse signaling from the exosomal surface molecules to the recipient DCs, and thus would not sustain full maturation and licensing of the recipient DCs. In another work involving mixed hematopoietic chimeras, Spörri and Reis e Sousa⁴⁰  have recently shown that deoxycytidylate phosphate deoxyguanylate oligonucleotide (CpG)–matured DCs can induce indirect maturation of neighboring CpG-unresponsive DCs. The indirectly matured DCs, however, did not allow full activation of T cells specific for the MHC-peptide complexes they bore. This observation again suggested that MHC-peptide complexes were not transferred from the CpG-unresponsive to the CpG-responsive DCs, since such a transfer should have allowed full activation of T cells specific for exosomal MHC molecules. But since we show here that exosomes from mature DCs are the efficient MHC-transfer vehicles, it is likely that CpG-unresponsive DCs do not secrete such exosomes, and thus do not lead to T-cell activation in trans. Experiments designed at addressing the 2 questions raised here are currently beeing performed in our laboratory.

In conclusion, we propose that, in vivo, exosomes secreted by mature DCs are sent at medium-range distances in the lymph node, and act as MHC and adhesion molecules microdomains, which, once bound to other APCs, induce T-cell activation. Such a mechanism in vivo remains to be demonstrated. By contrast, the in vivo efficiency of exosomes produced in vitro from mature DCs is clearly established here. Identification of both MHC-peptide complexes and ICAM-1 as major exosomal proteins for induction of immune responses suggests possibilities to produce artifical “exosome-like” liposomes to stimulate immune responses against tumors or infectious agents, which could prove cheaper and as efficient as exosomes produced in vitro by DCs.

We thank Drs B. Salomon (Hôpital Pitié-Salpêtrière, Paris, France) and O. Lantz (Institut Curie, Paris, France) for providing mice; Drs V. Gerke (Institut für Medizinische Biochemie, Münster, Germany) and H. Ploegh (Harvard Medical school, Boston, MA) for providing antibodies; the Laboratory of Mass Spectrometry and Proteomics and the Department of Bioinformatics at Curie Institute for help in proteomic analysis; and Drs O. Lantz and C. Hivroz for helpful comments on the manuscript.