Exosomes are lipid nanovesicles released after fusion of the endosomal limiting membrane with the plasma membrane. In this study, we investigated the requirement for CD4 T cells, B cells, and NK cells to provide help for CD8 T cell–mediated response to B cell–derived exosomes. CTL responses to Ag-loaded exosomes were dependent on host MHC class I, with a critical role for splenic langerin+ CD8α+ dendritic cells (DCs) in exosomal Ag cross-presentation. In addition, there was an absolute dependence on the presence of CD4 T cells, CD8 T cells, and NK cells, where the loss of any one of these subsets led to a complete loss of CTL response. Interestingly, NK cell depletion experiments demonstrated a critical cutoff point for depletion efficacy, with low-level residual NK cells providing sufficient help to allow optimal CD8 T cell proliferative responses to exosomal protein. Despite the potential role for B cells in the response to B cell–derived exosomal proteins, B cell depletion did not alter the exosome-induced CTL response. Similarly, a possible role for the BCR or circulating Ab in mediating CTL responses to B cell–derived exosomes was ruled out using DHLMP2A mice, which lack secreted and membrane-bound Ab, yet harbor marginal zone and follicular B cells. In contrast, CTL responses to DC-derived exosomes were significantly inhibited within Ab-deficient DHLMP2A mice compared with wild-type mice. However, this response was not restored upon serum transfer, implicating a role for the BCR, but not circulating Ab, in DC-derived exosome responses.
Exosomes are lipid nanovesicles formed via inward budding of the endosomal membrane, then released upon fusion of the endosomal limiting membrane with the plasma membrane. Exosomes stimulate potent anticancer immune responses, and in some cases, their Ag-presenting ability has been reported to exceed that of their parental cells (1–6). Ag associated with the membranes of exosomes is presented at much higher efficiency than soluble Ag, and the natural adjuvant properties of exosomes cannot be solely explained by their size or lipid content, because soluble or liposome-encapsulated Ag normally fails to induce T cell responses in the absence of adjuvants (7–10).
Surprisingly, recent in vivo studies have suggested that exosomes are more efficient at presenting intact Ag, as compared with peptide (2, 3). This is in distinct contrast with peptide-pulsed dendritic cells (DCs), which are around 50-fold more potent at activating T cells, as compared with protein-pulsed DCs (3, 11). In addition to T cell–mediated responses, exosomes induce B cell activation and Ab secretion, and it has been suggested that B cells and their Ab play a critical helper role in the induction of T cell response to protein Ag (2, 12, 13).
Despite the efficiency by which exosomes activate the immune system, the mechanisms of how this occurs are still poorly understood. The activity of peptide-pulsed exosomes is enhanced by transfer to DCs or by the addition of adjuvants, such as CpG (6, 14, 15). In addition, DCs, CD4 T cells, NK cells, NKT cells, and B cells have been proposed to be critical for optimal exosome activity in vivo (12, 14–17). Human DC-derived exosomes (DC Exo) are potent activators of NK cells, which may occur via NKG2D, IL-15Rα, or TNF superfamily ligands (15, 16). It is also possible that Ab and complement components aid in exosome Ag presentation (2, 12, 18).
In this study, we have examined cellular and molecular requirements for primary B cell–derived exosome-mediated CTL immunity in vivo. Despite the high-level expression of MHC class I (MHC-I) on primary B cell–derived exosomes, host rather than exosomal MHC-I was required for Ag presentation to CD8 T cells (3, 19). In particular, a role for endogenous splenic langerin+ CD8α+ DCs in the cross-presentation of B cell–derived exosomal Ag was observed. Interestingly, exosomal immunity showed an absolute dependence on CD4 T cells, CD8 T cells, and NK cells, with the depletion of each subset alone leading to a complete loss of CTL response. Depletion of B cells, or use of the DHLMP2A mouse model that harbors B cells, yet lacks functional Ab, demonstrated that B cells, the BCR, and their secreted Ab do not play a role in the immune response to B cell–derived exosomal Ag. In contrast, DC Exo required the BCR for optimal CTL responses to exosome-associated OVA, highlighting mechanistic differences of immune responses initiated by exosomes of different cellular origin. Overall these results suggest a complex interplay of cooperating leukocytes for exosome immunogenicity and emphasize the role of endogenous APCs in the exosome immune response.
Materials and Methods
C57BL/6 (B6; wild-type [WT]), B6.C-H2bm1/ByJ (bm1), OVA257–264 TCR-specific OT-I mice, and CD45.1+ B6.SJL-PtprcaPep3b/BoyJ were obtained from Jackson Laboratories. lang-DTREGFP or DHLMP2A mice on B6 background were kindly provided by B. Malissen (Centre d’Immunologie de Marseille-Luminy, France) or K. Rajewsky (Max-Delbrück-Center for Molecular Medicine, Germany), respectively. All mice were bred in specific pathogen-free conditions at the University of Otago Hercus Taieri Resource Unit. All immunizations and adoptive transfers were administered i.v. to the lateral tail vein in a volume of 100 μl. Animal studies were approved by the regional Animal Ethics Committee. DHLMP2A mice were genotyped by PCR using the primers (IDT): LMP2A forward 5′-CCT CTC ACT TCT ACT CTT GGC AGC-3′; LMP2A reverse 5′-GAG CAA ATC AGG AGA ACC ACA AGT G-3′; DQ52 forward 5′-ACG TCG ACG TTT TGA CTA AGC GGA GCA CCA CA-3′; JH1 reverse 5′-CCC GTT TCA GAA TGG AAT GTG C-3′. The PCR protocol was as follows: 94°C for 5 min; 45 cycles of 94°C for 30 s, 58°C for 35 s, 72°C for 40 s; 72°C for 5 min. PCR products were run on a 1.5% agarose gel at 95 V for 35 min, stained with 1× SYBR Safe (S3310; Invitrogen), and visualized at 600 nm with an Odyssey Fc imaging system (LI-COR Biosciences).
Complete media (10% FCS/RPMI complete media [R10]) was composed of RPMI 1640 (31800-022; Life Technologies) supplemented with 10% FCS (PAA Laboratories), 100 U/ml penicillin (15140-122; Life Technologies), 100 μg/ml streptomycin (15140-122; Life Technologies), 55 μM 2-ME (21985-023; Life Technologies), and 2 μg/ml NaHCO3. For exosome cultures, a 50% FCS/RPMI stock medium was exosome-depleted via ultracentrifugation at 120,000 × g for 18 h at 4°C and diluted prior to use.
Bone marrow–derived DCs
LPS-activated B6-derived bone marrow–derived DCs (BMDCs) were prepared as previously described and used at day 7 (3, 20). In brief, bone marrow cells (5 × 105 cells/ml) were cultured in GM-CSF–supplemented R10, fed with an additional 10 ml of GM-CSF–supplemented R10 on day 3, and matured overnight with 200 ng/ml LPS (Salmonella Typhimurium, L6511; Sigma) prior to harvest on day 7. Where stated, DCs were pulsed with 1 μM OVA257–264 peptide (DC257) for 4 h at 37°C on day 7, or 200 μg/ml OVA protein (DCs cultured with OVA protein [DC-pro]) for the final 2 d prior to harvest. DC257 and DC-pro were labeled with allophycocyanin-conjugated anti-mouse CD11c (1 μg/ml; clone HL3, 550261; BD) and PE-conjugated anti-mouse H2Kb:OVA257–264 (2 μg/ml; clone 25-D1.16, 12-5743-82; eBioscience) or unlabeled rabbit polyclonal anti-OVA (1 μg/ml). Unlabeled Ab was detected with goat anti-rabbit IgG (10 μg/ml). Forward scatter (FSc) and side scatter (SSc) single cells were gated and CD11c+ DCs were analyzed for anti-H2Kb:OVA257–264 or -OVA protein.
B cell–derived exosome culture
B6 or bm1 splenocytes were stimulated with anti-CD40 mAb (clone FGK45; 5 μg/ml) and murine IL-4 (50 ng/ml; R&D Systems, Auckland, New Zealand) at 2 × 106 cells/ml in exosome-depleted R10. Cells were cultured at 37°C with 5% CO2 for 3 d, followed by an additional 2 d with fresh media and exosomes isolated from day 3 and/or day 5 culture supernatant. Where stated, B cell cultures included 200 μg/ml OVA protein (A5503; Sigma) for 2 d (days 3–5), prior to supernatant harvest (exosomes derived from B cells cultured with OVA protein [BcExo-pro]). For CD19+ cell–derived exosomes, day 3 cultured splenocytes were sorted by MACS (107 cells/ml) with anti-CD19 microbeads (130-052-201; Miltenyi Biotec) as per the manufacturer’s instructions and incubated for a further 2 d with fresh media. Alternatively, cell-free “mock” preparations (anti-CD40/IL-4/OVA/R10) were treated identically as for BcExo-pro culture (mock preparation), or B cells were cultured without OVA protein and exosomes isolated from day 3 and/or day 5 cell supernatant (Empty BcExo).
DC Exo culture
BMDCs were prepared as described earlier with slight modifications; culture media used exosome-depleted FCS and included 10 ng/ml murine IL-4. On day 7, BMDCs were recultured in fresh GM-CSF/IL-4/LPS–supplemented R10, either with 200 μg/ml OVA protein (DC Exo-pro) or without OVA (Empty DC Exo), and cell supernatant was harvested on day 9 for exosome purification.
To purify exosomes, culture supernatant was centrifuged at 450 × g for 5 min, then 2000 × g for 20 min at 4°C to deplete cells and debris, respectively. The supernatant was 0.2 μm filtered, exosomes pelleted by ultracentrifugation at 120,000 × g for 1 h at 4°C, and then washed twice with 30 ml of PBS. Where stated, ultracentrifuged BcExo were pulsed with 1 μM OVA257–264 (BcExo257; Genscript) for 4 h at 37°C and washed twice with 30 ml of PBS. Washed exosomes were resuspended in 12 ml of PBS, overlaid onto 4 ml of 30% sucrose/200 mM Tris/D2O cushion, and ultracentrifuged at 100,000 × g for 75 min at 4°C. Exosomes were harvested by pooling fractions 1 ml above to 2 ml below the sucrose cushion interface (3). Sucrose was removed by washing twice with PBS by ultracentrifugation. Where stated, cell-free mock preparations were purified side by side with a BcExo-pro preparation, treated identically, and resuspended in an equivalent volume.
Exosomal protein quantification
Protein concentration was determined by pipetting 1 μl of spots of purified exosome samples or a standard of BSA/PBS onto Whatman filter paper (grade 4, 1004-185; Sigma). Samples were air-dried at room temperature (RT), flooded with 0.25% Coomassie brilliant blue G-250/10% glacial acetic acid/40% methanol/H2O, and destained twice with 10% glacial acetic acid/25% methanol/H2O for 10 min at RT with rocking. Coomassie G-250 fluorescence (800 nm) was visualized with an Odyssey Fc Imaging System (LI-COR Biosciences). Protein was quantified using Image Studio Lite software and background (PBS) signal subtracted from exosome samples to determine the final protein concentration. This method has previously been found to be comparable with that of the Bradford and BCA assay (data not shown).
Cryoelectron microscopy analysis of exosomes
Exo-pro samples (4 μl) were loaded onto plasma-glowed Quantifoil 2/2 grid and blotted to remove excess liquid. The grid was frozen by plunging into liquid ethane (−180°C) contained within a Reichert KF80 plunge freezing device (C. Reichert) and then stored in liquid nitrogen. Grids were mounted into a precooled Gatan 914 cryo holder and viewed in a JEOL 2200FS cryotransmission electron microscope with an omega filter. Zero-loss images were acquired at a filter width of 20–25 eV using a TVIPS F416 camera. The image was cropped using the software 3Dmod v4.7.9.
Flow cytometric analysis of exosomes
Exosomes were phenotyped by flow cytometry as previously described (3, 19). In brief, exosomes were bound to 4 μm of aldehyde-sulfate beads (A37304; Molecular Probes) overnight at 4°C, blocked with 0.05% BSA/PBS, and quenched with 100 mM glycine/PBS. Exo-beads were labeled with the following Abs: rabbit polyclonal anti-OVA (1 μg/ml; 23744; Polysciences); Alexa Fluor 488–conjugated goat anti-mouse IgG (5 μg/ml; A-11029; Invitrogen); PE-conjugated anti-mouse H2Kb:OVA257–264 (2 μg/ml; clone 25-D1.16, 12-5743-82; eBioscience), MHC-I (1 μg/ml; anti-H2; clone M1/42, 125506; BioLegend), MHC-II (5 μg/ml; anti–I-A/I-E; clone M5/114, 557000; BD), CD19 (2 μg/ml; clone 1D3, 115507; BD); or 5 μg/ml biotinylated anti-mouse B220 (clone RA3-6B2, 553085; BD) or CD81 (clone Eat2, 559518; BD). Unlabeled or biotinylated Abs were detected with Alexa Fluor 546–conjugated goat anti-rabbit IgG (10 μg/ml; A-11010; Invitrogen) or streptavidin-PE (1.25 μg/ml; 554061; BD). Exo-beads were analyzed by flow cytometry and subject to FSc and SSc doublet discrimination, in addition to FSc versus SSc gating.
In vivo T cell proliferation
CD45.2+ OT-I mice splenocytes and lymph node cells were stained with 2 μM CFSE (C34554; Molecular Probes) for 7 min at 20°C, quenched, and washed as previously described (3). CD45.1+ mice were adoptively transferred with 107 OT-I cells 2 d prior to immunization with PBS or 25 μg of BcExo-pro. Mice were euthanized 5 d later, spleens removed, and OT-I cells stained for CD45.2 (clone 104, 553772; BD) and CD8α (clone 53-6.7, 100743; BioLegend). FSc and SSc gated single-cell lymphocyte populations were analyzed by flow cytometry, and the percent divided of precursor cells and the division index (defined as the average number of cell divisions) calculated by proliferation analysis with FlowJo software.
In vivo CTL assay
B6 or DHLMP2A mice were immunized i.v. with PBS, 25 μg of BcExo-pro, DC Exo-pro, Empty BcExo, or Empty DC Exo, or equivalent volumes of a cell-free mock preparation treated identically as BcExo-pro, 50 μg of BcExo257, or 105 DC-pro or DC257. Where stated, mice were adoptively transferred with 107 OT-I splenocytes 2 d prior to immunization. Seven days after immunization, target cells were prepared from naive B6 splenocytes and incubated in R10 alone (unpulsed) or with 1 μM OVA257–264 in R10 (pulsed) for 1 h at 37°C. Unpulsed and pulsed cells were stained with 0.2 and 2 μM CFSE, respectively, for 7 min at 20°C, quenched, and washed as described previously (3). Where stated, targets were additionally stained with 0.5 μM CellTrace Yellow dye (CT Yellow, C34567; Molecular Probes) simultaneously during CFSE staining to distinguish target cells from GFP+ langerin+ CD8α+ DCs within lang-DTREGFP mice. Equal numbers of unpulsed and pulsed target cells were pooled (1.5 × 107 total) for i.v. injection. Mice were euthanized 18 h later, and target cell killing was determined by flow cytometry.
In vivo cell depletion and cytokine neutralization
B6 (CD45.2+) or CD45.1+ mice were depleted of CD8 or CD4 T cells with three consecutive i.v. injections (24 h apart) of 100 μg of rat anti-mouse CD8β (clone 53-5.8) or CD4 (clone YTS191.1.2) Abs, respectively; injections were performed on days −8, −7, and −6 with respect to time of immunization (day 0). Alternatively, using a single i.v. injection, mice were depleted of NK cells on day −2 (priming phase), day +5 (effector phase) with 20 μl of rabbit anti-asialo-GM1 (Wako) or, where stated, with 200 μg of BALB/c anti-mouse NK1.1 (clone PK136), or B cells on day −6 with 250 μg of mouse anti-mouse CD20 (clone 18B12; Dr. J. Browning, Biogen Idec, Cambridge, MA) Abs. As a control, mice were mock depleted as stated with 100 μg of rat or rabbit IgG, or 200 μg of BALB/c IgG on day −2, or 250 μg of mouse anti-human CD20 (clone 2B8; Dr. J. Browning, Biogen Idec) on day −6. Depletion of splenic CD8α+ CD4− CD11c+ DCs within lang-DTREGFP (or B6 control) mice was achieved via i.p. injection of 500 ng of diphtheria toxin (DT; D0564; Sigma) or PBS control at 24 h preimmunization and postimmunization. Blocking of IFN-γ was via i.v. injection of 100 μg of rat anti-mouse IFN-γ mAb (clone XMG.D6) on days −1 and +2 relative to immunization (day 0). Absolute cell numbers were determined from spleens at the conclusion of the experiment. CD4 T cells (CD4+ TCRβ+), CD8 T cells (CD8α+ TCRβ+), NK cells (CD49b+ TCRβ−), NKT cells (NK1.1+ TCRβ+), and B cells (B220+) were identified as described later.
Flow cytometry of splenocytes
Splenocytes were stained with mAb diluted in 0.1% BSA/PBS/2 mM EDTA and incubated for 15 min on ice. The following anti-mouse Abs were used: anti–TCRβ-biotin (2.5 μg/ml; clone H57-597, 553169; BD), anti–CD45.2-FITC (2.5 μg/ml; clone 104, 553772; BD), anti–CD49b-PE (1 μg/ml; clone DX5, 553858; BD), anti–CD8α-Brilliant Violet 605 (0.13 μg/ml; clone 53-6.7, 100743; BioLegend), Brilliant Violet 421–conjugated anti-TCRβ (0.33 μg/ml; clone H57-597, 109229; BioLegend) or anti-B220 (0.25 μg/ml; clone RA3-6B2, 103239; BioLegend), or allophycocyanin-conjugated anti-CD4 (1 μg/ml; clone RM4-5, 553051; BD) or anti-NK1.1 (1 μg/ml; clone PK136, 550627; BD). Biotin was detected with streptavidin-allophycocyanin (1 μg/ml; 405207; BioLegend) or -PerCP (1 μg/ml; 554064; BD).
Flow cytometric data were acquired using a BD LSRFortessa with BD FACSDiva software. Data were analyzed using FlowJo v9 software. Cells were subject to FSc and SSc doublet discrimination, in addition to FSc versus SSc gating of the live lymphocyte population. Voltages were set appropriately for negative populations of two empty channels (where no fluorophores were excited); autofluorescent double-positive cells were then excluded from the lymphocyte population by gating. For in vivo CTL or proliferation experiments, lymphocyte populations were additionally gated for the endogenous population (CFSE− or CD45.2−). Graphing and statistical analysis were performed using GraphPad Prism v6.
Exosome samples (10 or 20 μg as stated) or OVA protein standard were boiled and reduced (except those for anti-CD138 probing) prior to gel electrophoresis and transferred to nitrocellulose membrane (9, 21). Using 1% BSA/PBS as a blocker and diluent, we performed Western blotting of the membranes with goat polyclonal anti-milk fat globule-epidermal growth factor 8 (MFG-E8)-biotin (0.2 μg/ml; BAF2805; R&D Systems), rabbit polyclonal anti–14-3-3γ (2 μg/ml; sc-731; Santa Cruz), mouse anti-Alix (2 μg/ml; clone 3A9, sc-53538; Santa Cruz), rat anti-CD138 (2 μg/ml; clone 281-2, 553712; BD), mouse anti–β-actin (0.5 μg/ml; clone AC-15; A1978; Sigma), or rabbit polyclonal anti-calnexin (2 μg/ml; sc-11397; Santa Cruz) or anti-OVA (5 μg/ml; 23744; Polysciences) Abs. Primary Abs were detected, respectively, with streptavidin-DyLight800 (21851; Thermo Fisher), goat anti-rabbit IgG-HRP (A0545; Sigma) or anti-mouse IgG-HRP (A4416; Sigma), rabbit anti-rat IgG-HRP (A5795; Sigma), or donkey anti-mouse IgG-Dylight680 (SA5-10170; Pierce) or anti-rabbit IgG-Dylight800 (SA5-10044; Pierce). All HRP- or DyLight-conjugated secondary reagents were diluted 1:1000 or 1:10,000, respectively (9, 21). HRP signal was developed with diaminobenzidine or DyLight signal visualized using an Odyssey Fc imaging system (LI-COR Biosciences). Image Studio Lite software was used to quantify OVA protein and images cropped to show bands.
Anti-murine Ig ELISA
Nunc MaxiSorp 96-well plates were coated overnight at 4°C with 2 μg/ml goat anti-mouse Ig (1010-10; Southern Biotech) in PBS (50 μl/well). Plates were washed three times with 0.05% Tween 20/PBS, then blocked with 1% FCS/PBS (100 μl/well) for 10 min at RT. Blocking solution was flicked out and plates incubated overnight at 4°C with serum diluted in 1% FCS/PBS (50 μl/well). After washing, plates were incubated with goat anti-mouse IgG-HRP (50 μl/well; A4416-0.5ML; Sigma) diluted 1:1000 in 1% FCS/PBS for 1 h at 37°C. Plates were then washed, developed in the dark at RT with 3,3′,5,5′-tetramethylbenzidine (50 μl/well; 002023; Life Technologies), and the reaction stopped by the addition of 2N H2SO4 (25 μl/well). Plates were read using a Bio-Rad 559 Microplate reader at 450 nm. OD background (1% FCS/PBS control) was subtracted from sample OD using GraphPad Prism software.
Anti-OVA B cell ELISPOT
B6 mice were depleted of NK cells (or mock depleted) 2 d prior to immunization with PBS or 25 μg of BcExo-pro. Seven days postimmunization, splenocytes were analyzed for anti-OVA Ab-secreting cells as follows. All incubation steps were performed with 100 μl/well volumes unless stated otherwise. Multiscreen 96-well plates with Immobilon-P membrane (MAIPS4510; Millipore) were prewet with 70% ethanol (50 μl/well), flicked out immediately, then washed five times with sterile Milli-Q H2O (200 μl/well). Plates were coated overnight at 4°C with goat anti-mouse IgG (1010-01; Southern Biotech) diluted in PBS. Plates were washed as described earlier and blocked with R10 (200 μl/well) for 30 min at RT. Quadruplicate splenocyte samples in R10 were added (2 × 105/well) and incubated for 18 h at 37°C with 5% CO2. Samples were flicked out, plates washed five times with PBS, and then incubated with 1 μg/ml biotinylated OVA protein (A5503; biotinylated in-house; Sigma) diluted in 0.5% FCS/PBS for 2 h at 37°C. Plates were washed with PBS and biotin detected with alkaline phosphatase-conjugated streptavidin (3310-10; Mabtech) diluted 1:1000 with 0.5% FCS/PBS for 1 h at RT. After washing with PBS, plates were developed with 5-bromo-4-chloro-3-indolyl-phosphate)/NBT-plus (0.45 μm filtered; 3650-10; Mabtech) in the dark at RT and the reaction stopped with tap H2O. Membranes were air-dried overnight in the dark at RT prior to counting.
NK cell requirement for optimal CD8 T cell proliferation in response to B cell–derived exosomal-protein immunization
We have previously shown that primary B cells release high levels of exosomes (BcExo) in response to CD40 and IL-4 signaling (19). BcExo isolated via differential centrifugation and sucrose cushion purification were visualized by cryoelectron microscopy (Fig. 1A). Further characterization demonstrated the expression of B cell markers CD19, CD45R (B220), and Ig, as well as MHC-I, MHC-II, and the tetraspanin CD81 (Fig. 1B) (3, 19). BcExo preparations were free of the endoplasmic reticulum (ER) contaminants as shown by the lack of calnexin detected by Western blotting (Fig. 1C). We detected the exosome biogenesis markers Alix and syndecan-1 (CD138), the signal transduction molecule 14-3-3γ, as well as MFG-E8 (Fig. 1C) (22, 23). To investigate the immune cells responding to B cell–derived exosomal Ag, the absolute numbers of splenic immune cell subsets was determined after i.v. immunization with exosomes derived from B cells cultured with whole OVA protein (BcExo-pro). A significant increase of NK cells, B cells, CD4 T cells, and CD8 T cells within the spleen was noted postimmunization (Fig. 1D, 1E). To address whether NK cells were required for lymphocyte expansion in response to BcExo-pro, we depleted mice of NK cells with anti–asialo-GM1. NK cell depletion did not alter the absolute numbers of B cells in response to BcExo-pro; however, lower numbers of CD4 and CD8 T cells were observed (Fig. 1E). In vivo depletion of NK cells led to a marked decrease in CD8 T cell (OT-I) proliferation (Fig. 2A), but this difference failed to reach significance (Fig. 2B). However, during routine analysis of depletion efficacy (Fig. 2C), a positive correlation between NK cell number and the division index of CD8 T cell proliferation was noted (R2 = 0.9114; p < 0.0001), thus indicating a critical threshold of NK cell number within the spleen (∼750,000 NK cells/spleen) was required for optimal CD8 T cell proliferation (Fig. 2D) to BcExo Ag.
Absolute dependence of NK cells, CD4 T cells, and CD8 T cells for B cell–derived exosomal protein–induced in vivo cytotoxicity
Stimulation of whole splenocytes with anti-CD40 mAb and IL-4 led to the specific expansion of B cells [∼85–95% B cell purity by day 3; n = 10; data not shown (19)]. B cells were further expanded by culture with fresh medium containing anti-CD40 mAb and IL-4 for an additional 2 d and exosomes subsequently isolated from this day 3–5 cell culture supernatant. Exosomes derived from splenic B cells cultured with OVA protein (over the additional day 3–5 culture period) were potent inducers of in vivo cytotoxicity (Supplemental Fig. 1) (3). Functional exosomes were confirmed to be of B cell origin, with no significant difference noted in in vivo cytotoxicity from immunization with exosomes derived from CD40/IL-4–stimulated cultures (days 3–5) of whole splenocytes or CD19+ sorted cells (Supplemental Fig. 1).
To more specifically determine the role of immune cell subsets in the response to B cell–derived exosomal Ag, we used Ab-mediated cell depletion in conjunction with in vivo CTL assays. BcExo-pro–induced CTL responses showed absolute dependence on the presence of NK cells, CD4 T cells, and CD8 T cells, where the loss of any one of these subsets led to the complete abolishment of CTL response (Fig. 3A, 3C). In contrast, despite the expansion of OVA Ab-secreting B cells in response to BcExo-pro (Supplemental Fig. 2), the loss of B cells had no effect on the in vivo CTL response (Fig. 3A). CTL responses to protein-pulsed DC (105; DC-pro) immunization were also reduced after the depletion of CD4 T cells, CD8 T cells, and NK cells (Fig. 3B). BcExo-pro expressed high levels of surface OVA protein and displayed detectable cross-presented H2Kb:OVA257–264 complexes (Fig. 3D). In contrast, only a small proportion of DCs expressed intact OVA protein or H2Kb:OVA257–264 (Fig. 3D). A geometric mean of 260.0 ng ± 33.62 SD (n = 3) of OVA protein was present in 10 μg of BcExo-pro (Fig. 3E), thus indicating that a BcExo-pro immunization dose of 25 μg contains ∼650 ng of OVA protein.
It has been reported that administration of anti–asialo-GM1 depletes not only NK cells, but also basophils in vivo (24). We therefore compared the efficiency of anti–asialo-GM1 and anti-NK1.1 (clone PK136) Ab to reduce the CTL response to BcExo-pro. Significant inhibition of BcExo-pro–induced cytotoxicity was observed regardless of the Ab depletion method (Fig. 4A). Anti-NK1.1 treatment resulted in significant NK cell (DX5+ TCRβ−) depletion, but not of NK1.1+ TCRβ+ cells (Fig. 4B, 4C), the majority of which represent NKT cells (25). Interestingly, the depletion of NK cells at the effector phase resulted in the complete inhibition of in vivo cytotoxicity (Fig. 4D). However, because the observed CTL activity was targeted only toward cells expressing MHC-I:OVA257–264, this implicates NK cells with a helper rather than an effector role.
To address the possible role of IFN-γ in this NK cell help for CTL responses, we administered neutralizing anti–IFN-γ Ab to mice prior to BcExo-pro immunization. Only a slight and nonsignificant decrease in in vivo cytotoxicity was noted upon administration of neutralizing IFN-γ prior to priming (Fig. 4E). Furthermore, immunization of IFN-γ–deficient mice with BcExo-pro resulted in no reduction of in vivo cytotoxicity compared with WT mice (data not shown).
Immunization of mice with either an equivalent protein quantity of unpulsed empty BcExo (Fig. 4D) or equivalent volumes of a mock preparation (anti–CD40/IL-4/OVA protein in cell culture medium, but without B cells) as compared with identically purified BcExo-pro (Supplemental Fig. 1) failed to induce an in vivo CTL response. Thus, the observed BcExo-pro–induced cytotoxicity was in response to exosomes carrying OVA protein, as opposed to either a nonspecific exosome response or the presence of residual reagents from the cell culture medium.
Increased T cell precursor frequency overcomes NK cell–dependent B cell–derived exosome-induced in vivo cytotoxicity
B cell–derived exosomes were pulsed with OVA peptide (OVA257–264; BcExo257) and the anti-OVA CD8 T cell precursor frequency increased by OT-I adoptive transfer to compensate for the loss of potency using the peptide system (3, 12). NK cells were found to be dispensable for the in vivo CTL response (Supplemental Fig. 3A). To confirm whether the observed difference of NK cell involvement was due to the altered exosomal antigenic context or the increased frequency of OVA-specific OT-I T cells, we immunized mice with BcExo-pro in conjunction with OT-I T cell supplementation. In contrast with the endogenous system, OT-I supplementation inhibited CTL activity in mice immunized with BcExo-pro (Fig. 4A, Supplemental Fig. 3B, mock depleted). Interestingly, BcExo-pro–induced cytotoxicity in the presence of OT-I cells was restored upon NK cell depletion (Supplemental Fig. 3B, 3C).
Interestingly, the ability of peptide-pulsed DCs (Supplemental Fig. 3A) to promote in vivo CTL activity was substantially greater than protein-pulsed DCs (Fig. 3B). This response by peptide-pulsed DCs also exceeded the CTL activity induced by exosome immunization (regardless of antigenic context; Fig. 3A, Supplemental Fig. 3A). This possibly reflects the high level of H2Kb:OVA257–264 expressed by DCs directly pulsed with peptide (DC257; Supplemental Fig. 3D), as opposed to the lower level of naturally cross-presented peptide on BcExo-pro and DC-pro (Fig. 3D).
Splenic langerin+ CD8α+ DCs cross-present B cell–derived exosomal Ag for the induction of in vivo CTL responses
To determine whether MHC-I:peptide complexes on exosomes directly stimulate CD8 T cells in vivo, we used bm1 mice that form complexes of mutant H2-Kbm1/OVA257–264, which cannot be recognized by B6 or OT-I–derived OVA257–264–specific T cells (26). OVA (peptide or protein)-pulsed exosomes derived from either B6 or bm1 B cells were equivalent in their ability to induce in vivo OVA257–264–dependent CTL responses (Fig. 5A, 5B), demonstrating that exosomal Ag is cross-presented by host APCs. To determine whether the cross-presenting langerin+ CD8α+ DC subset (27) was involved in cross-presentation of B cell–derived OVA on exosomes, we used lang-DTREGFP mice. Efficient depletion of splenic langerin+ CD8α+ CD4− CD11c+ DCs within lang-DTREGFP mice after DT administration was observed (92–95% depletion; n = 2; data not shown). A significant decrease in BcExo-pro–induced in vivo cytotoxicity was noted within DT-treated lang-DTREGFP compared with both mock-treated lang-DTREGFP and DT-treated B6 mice (Fig. 5C).
In vivo cytotoxicity induced by exosomal protein of B cell origin is independent of host BCR expression and circulating Ab, whereas DC Exo require host BCR presence
Although host B cells were not required for B cell–derived exosomal protein–induced in vivo cytotoxicity (Fig. 3A), the Gabrielsson group (2, 12) reported that DC Exo induce T cell responses that are B cell dependent. These contrasting results may be caused by the differing exosome cellular origin or the techniques used; whereas our study used WT mice depleted of B cells but still possessing circulating natural Abs, the Gabrielsson group (2, 12) used B cell–deficient mice that are also devoid of circulating Ab. To further address the role of circulating Ab in the exosome-induced immune response, we used DHLMP2A mice that retain follicular and marginal zone B cells, but lack a BCR and circulating Ab. WT (B6) or DHLMP2A mice were immunized with BcExo-pro; however, no difference in in vivo cytotoxicity between the two mouse strains was noted, nor was there any enhancement in response when BcExo-pro were admixed with Ig-containing WT serum (Fig. 6A). To ensure BcExo-pro admixed with serum were not rejected by Ab-deficient mice, homozygous DHLMP2A mice used for this experiment were born to hemizygous mothers (WT phenotype), therefore ensuring tolerance of pups to Ig because of exposure via passively transferred Ig. Although a low level of circulating passively transferred Ig was still detectable at 8 wk of age within DHLMP2A mice born to hemizygous mothers compared with those born to homozygous mothers, this had no bearing on the CTL activity in response to BcExo-pro (Fig. 6B, 6C). To investigate whether the contrasting results between the Gabrielsson group (2, 12) and our own was due to the exosome cellular origin, we immunized B6 or DHLMP2A mice with DC Exo-pro. DC Exo-pro preparations were enriched in MFG-E8 and OVA protein, showed detectable levels of 14-3-3γ, but lacked the ER contaminant calnexin (Fig. 7A, 7B). Surprisingly, in contrast with exosomes of B cell origin, the ability of DC Exo-pro to induce in vivo cytotoxicity was significantly impaired within DHLMP2A mice, as compared with WT mice. However, the observed low-level cytotoxicity within DHLMP2A mice in response to DC Exo-pro remained unchanged regardless of WT serum addition (Fig. 7C), implicating a role for the host BCR, but not circulating Ab, in the immune response to exosomes of DC origin.
Although exosomes play wide-ranging roles in immunity and tolerance, the mechanisms leading to exosome uptake and Ag presentation of exosomal Ags are still poorly understood. Few studies have focused on exosomes derived from primary cells, and most use exosomes derived from either in vitro differentiated bone marrow stem cells (e.g., BMDCs) or tumor cells (28–30). We have previously established a role for primary B cell–derived exosomes in the promotion of T cell responses in vivo, and in this study we have investigated the cellular mediators required for the CTL response to exosomes (3). Although a number of studies have previously focused on the APCs and their molecules required for exosomal uptake and responses in vivo, far fewer studies have focused on the role of “helper” cells in aiding the activation of T cells to exosomes (31–36).
Our experiments utilizing exosomes prepared from OVA protein–loaded B cells (BcExo-pro) from H-2Kbm1 mice demonstrated that the recognition of peptide in the context of exosomal MHC-I was not required for the induction of CTL responses (Fig. 5A, 5B). The fact that exosomal MHC-I was not required for induction of T cell responses to either protein or peptide suggests that Ag is transferred to host APC and is consistent with results obtained using MHC-I–deficient exosomes (34) and demonstrated in vivo for MHC-II:peptide (31). To further investigate the role of host DC, we used lang-DTREGFP mice depleted of splenic langerin+ CD8α+ DCs and observed a significant decline in BcExo-pro–induced CTL response in vivo (Fig. 5C). This corresponds with other reports of observed transfer of exosomal material to CD8αα DCs (32, 35). Our results may indicate a collaborative role between splenic CD169+ macrophages that specifically capture blood-borne exosomes (3), but exhibit poor cross-presentation ability (37), with the closely associated langerin+ CD8α+ DCs that are conversely poor at capture of blood-borne particulate Ag but have high cross-presentation ability (38). Although the mechanisms of cross-presentation in this system are obscure, recent work has uncovered possible pathways of exosome uptake by cells. Heusermann et al. (39) have shown that exosomes are efficiently taken up by filopodia transported in a retrograde fashion, intersecting with potential cross-presenting cellular organelles, including endosomes and the ER. Our data mirrors that using OVA-pulsed DCs, where DCs are able to immunize hosts independently of MHC-I:peptide presentation and appear to be acting as simple vessels for Ag transfer to host DCs (40–42).
Immunization with soluble OVA protein at a 15-fold higher dose (9, 21) of that contained within a 25 μg BcExo-pro immunization dose (Fig. 3E) did not induce in vivo CTL responses. Thus, B cell and DC Exo proved to be interesting immunogens possessing classical adjuvant properties, likely because of their APC origin. The reason for the observed surface-expressed OVA protein is not clear (Figs. 3D, 7B); however, mature OVA protein possesses membrane binding capacity because of the retention of an endogenous, uncleaved signal sequence at the N terminus (43).
In this study, we have identified CD4 T cells and NK cells, but not B cells, as crucial helper cells for the induction of in vivo CTL responses to blood-borne exosomal protein (Fig. 3A). Although the requirement of CD4 T cells was not surprising and most likely represents classical Th1 help required for CD8 T cell activation, NK cell requirement was unexpected. Murine and human DC Exo have been reported to promote NK cell activation and proliferation via an NKG2D-dependent mechanism, as well as IL-15Rα (15, 44). In contrast, Gehrmann et al. (17) observed no difference in NK cell proliferation within mice immunized with DC Exo compared with PBS. Rather, the incorporation of α-galactosylceramide into DC Exo led to significant NK cell proliferation in an invariant NKT cell–dependent manner (17). To our knowledge, ours is the first report of the fundamental collaborative role between NK cells, CD4 T cells, and CD8 T cells for responses to exosomal Ag, with each subset absolutely critical for the induction of CTL responses. Strikingly, we observed a positive correlation between NK cell number and the OT-I proliferative response (division index) in response to BcExo-pro immunization (Fig. 2D), suggesting that there exists a critical threshold of NK cell numbers (∼750,000 NK cells/spleen) required to provide help for optimal T cell proliferative responses to BcExo-pro.
NK cell depletion at either priming or effector phase resulted in a loss of BcExo-pro–induced MHC-I:OVA257–264–specific CTL response (Figs. 3A, 4A, 4D). The cytotoxic abilities of NK cells are well characterized; however, they have also been attributed with helper functions with regard to CD8 T cell activation (20, 45, 46). In particular, CD137 or IL-18–stimulated NK cells exhibit helper functions (47, 48). NK:DC cross-talk are also well-established mechanisms, such as interactions via NKp30, IL-15Rα, or membrane-bound IL-15, leading to enhanced cytokine secretion, for example, IFN-γ and TNF-α and subsequent promotion of T cell proliferation (49–52). We have previously demonstrated a critical role for NK cells in T cell responses mediated by IFN-γ (20). However, in this study, IFN-γ did not appear to play a role in the NK cell–mediated helper response (Fig. 4E), nor was a role for IFN-γ observed for BcExo-pro responses using ifnγ−/− mice as part of another study (S.C. Saunderson and A.D. McLellan, unpublished observations). Alternatively, NK cells have been suggested to provide OX40L:OX40 costimulation to CD4 T cells (53). Further work will be required to elucidate the precise mechanisms for NK cell involvement.
Although NK cells were essential for endogenous in vivo CTL responses to BcExo-pro (Figs. 3A, 4), their presence was detrimental when the precursor frequency was increased by OT-I supplementation (Supplemental Fig. 3B). This was despite optimal OT-I proliferation requiring NK cells, where only weak OT-I proliferation was observed within NK cell–depleted mice (Fig. 2). Although this appears counterintuitive, one explanation is that adoptively transferred OT-I T cells expand and compete for resources, thereby limiting high-avidity, endogenous CTL responses, a proposal consistent with other studies (54, 55). Interestingly, NK cell depletion, which would have reigned in excessive OT-I T cell proliferation in response to BcExo-pro (Fig. 2), led to a return of CTL activity (Supplemental Fig. 3B).
We and others have demonstrated a role for the BCR in the uptake and presentation of Ag by exosomes (2, 12, 19, 56). However, we were able to show that B cells and Ab were dispensable for B cell–derived exosome-induced CTL responses. In contrast, studies by the Gabrielsson group (2, 12) demonstrated impaired T cell responses to DC Exo within B cell–deficient mice and proposed a helper function role for the BCR. This difference may be attributed to either the cellular origin of the exosomes or the system employed; although our own study used WT mice depleted of B cells with anti-CD20 mAb, leaving natural circulating Ab level unaffected, the former study used transgenic mice (CD19−/−, μMT, or btk−/−) lacking B cells, and thus were devoid of or exhibited marked deficiencies in circulating natural Abs (57–59). To further address this, we used DHLMP2A mice that lack circulating Ab, yet harbor follicular and marginal zone B cells that are devoid of a BCR (60). Our experiments demonstrated that neither the innate nor adaptive functions of B cells are required for immune responses to BcExo-pro. Interestingly, when immunizations were performed using DC Exo-pro, a significant impairment in the CTL response was observed in DHLMP2A mice that was unable to be restored by the coadministration of Ig-replete WT serum (Fig. 7). These results support a role for the BCR, but not of circulating Ab for immune responses to exosomes of DC origin, and thus are consistent with those of the Gabrielsson group (2, 12). The findings also emphasize the influence of the parental cell type on the functional properties of exosomes.
Our results offer clarity on the pathways involved in the immune response to primary B cell–derived exosomes and emphasize the role of multiple lymphocyte subsets in the immune response. As well as highlighting a key difference in B cell dependence for the induction of in vivo CTL responses by exosomes of different APC origin. It will be interesting to determine whether the cooperation identified in this study between NK cells, CD4 T cells, and CD8 T cells is operational for exosomes derived from other sources.
We thank Klaus Rajewsky for the DHLMP2A mice, Franca Ronchese for the ifnγ−/− mice, Bernard Malissen for the lang-DTREGFP mice, Jeff Browning for the CD20 mAb, Greg Jones for the 14-3-3γ Ab, Merilyn Hibma for assistance with mouse breeding, and the Otago Centre for Electron Microscopy for the cryoimaging of exosomes.
This work was supported by a University of Otago Division of Health Sciences Career Development postdoctoral fellowship (to S.C.S.) and a University of Otago research grant.
The online version of this article contains supplemental material.
Abbreviations used in this article:
B cell–derived exosomes pulsed with OVA257–264 peptide
exosomes derived from B cells cultured with OVA protein
bone marrow–derived DC
dendritic cells pulsed with OVA257–264 peptide
- DC Exo
- DC Exo-pro
exosomes derived from DCs cultured with OVA protein
DCs cultured with OVA protein
- Empty BcExo
unpulsed B cell–derived exosomes that do not contain OVA Ag
- Empty DC Exo
unpulsed DC-derived exosomes that do not contain OVA Ag
milk fat globule-epidermal growth factor 8
MHC class I
- OVA pro
10% FCS/RPMI complete media
The authors have no financial conflicts of interest.