Allogenicity Boosts Extracellular Vesicle–Induced Antigen-Specific Immunity and Mediates Tumor Protection and Long-Term Memory In Vivo

The latest progress in cancer immunotherapy has renewed the interest in novel approaches to induce cancer-specific immune responses. Today, dendritic cell (DC)–based therapies are used and have been extensively explored in clinical trials reviewed in Ref. 1. DC-based vaccines rely on the host DCs for Ag presentation, as they do not present the Ags themselves directly to the CD8⁺ T cells for immune activation (2). Many of the cell-based vaccines are not of high enough quality to be reinfused into the patient and thus are both expensive and inefficient. However, these obstacles can be circumvented by the use of “off the shelf” therapies such as extracellular vesicles (EVs).

Small EVs, mainly exosomes, are 30–150-nm membrane vesicles, and those secreted by APCs have shown immune stimulatory functions in vitro (3–5) and in vivo (6, 7). Even though the physiological role of EVs to a large extent remains unknown, they are currently being investigated for their potential as therapeutic delivery tools both for immune regulating and activating applications (8, 9). In cancer immunotherapeutic applications, EVs from immature dendritic cells (DCs) have shown safety but only minor immune stimulatory effects in phase I clinical trials (10, 11). A recent phase II clinical trial showed that IFN-γ–matured DC-derived EVs loaded with MHC class I (MHCI) and MHC class II (MHCII) tumor peptides boosted NK cell function in lung cancer patients but only induced minor T cell stimulation (12), demonstrating the need for more immunogenic EVs in cancer therapy. We previously demonstrated that the addition of α-galactosylceramide (αGC) further improves immunogenicity of bone marrow–derived DC (BMDC)–derived EVs by the activation of invariant NKT cells (iNKTs) (13). This beneficial effect by αGC has been confirmed recently using tumor-derived EVs in a murine model of glioblastoma (14). Moreover, we have shown that both CD4⁺ and CD8⁺ T cells activation in vivo is dependent on B cells (15, 16). Our findings also indicated that the presence of whole Ag on the EVs rather than MHC-peptide complexes were essential for in vivo T cell priming, removing the need for MHC compatibility of EVs. Indeed, we have shown that MHCI^−/− EVs were as efficient as the syngeneic ones in priming CD8⁺ T cell responses. Moreover, a single injection of both MHCI^−/− and allogeneic EVs inhibited tumor progression (17). In line with this, DCs have been shown to be crucial for EV-induced immune responses in vivo (18), supporting that Ag-loaded EVs are processed by host cells (i.e., DCs in vivo rather than directly stimulating T cells.) This has opened up for the possibility to use impersonalized EVs for immunotherapy, which also has been explored recently for DC-derived EVs using the α-fetoprotein mouse model treating hepatocellular carcinoma (19).

Injection of allogeneic material may be associated with a multiclonal T cell response. In this study, we therefore tested allogeneic EVs for their potential to act as adjuvants due to bystander activation in a boost regimen for treatment. Our findings suggest that mainly humoral responses were enhanced by allogenicity. Long-term memory experiments demonstrated that two injections compared with a single injection induced long-lasting Abs with higher avidity and a general increase in the number of Ab-forming cells (plasma cells). In contrast, syngeneic and allogeneic EVs were equally potent stimulators of CD8⁺ T cells and germinal center B (GC B) cells, whereas allogeneic EVs induced significantly higher numbers of T follicular helper (Tfh) cells and Ag-specific Abs. Furthermore, two injections of allogeneic EVs delayed tumor progression in B16mOVA melanoma tumor-bearing mice as efficiently as their syngeneic counterparts, thus providing an alternative for immunotherapeutic purposes in the clinic.

C57BL/6NTac and BALB/cJBomTac female mice (Taconic) were kept under specific pathogen-free conditions at Karolinska Institutet animal facility. All experiments were approved by the Stockholm Regional Ethics committee.

BM cells from C57BL/6N and BALB/c mice were isolated and cultured as described previously (15). On day 6, cells were incubated with 300 μg/ml OVA (Sigma-Aldrich) and 100 ng/ml α-GC (KRN-7.000; AdipoGen). At day 7, the culture medium was removed, and the cells were washed with PBS and plated out at 0.5 × 10⁶ cells/ml in medium containing EV-depleted FCS (produced by spinning 30% FCS at 100,000 g for 15 h) and 30 ng/ml LPS (Sigma-Aldrich). At day 9, the supernatant was removed and centrifuged at 300 × g for 10 min followed by a 3000 × g spin for 30 min to remove cells and cell debris, respectively. Then, the supernatant was filtered through a 0.22-μm pore-sized filter (Techno Plastic Products), and subsequently EVs were pelleted for 2 h by ultracentrifugation (rotor type 45 Ti; Beckman Coulter) at 100,000 g. After a PBS wash, the EV pellet was resuspended in a small volume of PBS (400 to 600 μl), and protein concentration was determined by DC protein assay (BioRad). For the in vivo injections 40 μg of EVs was used per mouse, representing EVs isolated from approximately one BM (EV recovery of the BMDC culture was for the C57BL/6N 40.3 ± 6.3 μg of SEM/BM and for BALB/c, 23.3 ± 5.6 of μg SEM/BM). EVs were stored at −80°C until further use. BMDCs were stained on day 9 for CD81 (Eat-2), PD-L2 (TY25), CD9 (MZ3), CD63 (NVG-2), CD86 (GL-1), PD-L1 (10F.9G2), I-A/I-E (M5/114.15.2), CD1d (1B1), CTLA-4 (UC10-4B9), CD80 (16-10A1), CD54 (YN1/1.7.4), CD83 (Michel-19), H-2kb (AF6-88.5), H-2Kd (SF1-1.1), CD11c (N418), CD11b (M1/70) (all from BioLegend), and CD40 (3/23) and CD14 (rmC5.3) (both from BD Biosciences) and analyzed using FACS Canto II (BD Bioscience).

EVs were subjected to electron microscopy (EM) with negative ion capture. A small aliquot was added to a grid with a glow-discharged, carbon-coated supporting film for 3 min. Excess solution was removed by filter paper, the grid rinsed with distilled water for 10 s, dried using filter paper, followed by staining with 5 μl of 1% uranyl acetate in water from 7 s. Excess liquid was removed and the grid left to air dry. Samples were examined at 80 kV using a Hitachi HT 7700 EM (Hitachi) and images obtained using a Veleta camera (Olympus).

The size distribution of EVs was determined by nanoparticle tracking analysis (NTA) (LM10HSB system; NanoSight) equipped with a 405-nm laser running an NTA 3.0 analytical software package. EVs were diluted in PBS to a particle concentration of 1 × 10⁸ to 9 × 10⁸ particles/ml. When the right dilution was established, the EVs were run five times for 60 s with a syringe speed of 50 and camera level of 14. At least four independent EV batches were run for both B6 and BALB/c EVs.

Sulfate-aldehyde latex microsphere beads 4 μm, 1.3 × 10⁹ beads/ml (Invitrogen) were incubated overnight at room temperature (RT) with 10 μg of anti-mouse CD9 Ab, clone KMC8 (BD Pharmingen). The beads were washed and blocked with 100 mM glycine for 30 min at RT followed by an additional washing with 0.5% BSA/PBS. After washing, 2 μg of EVs per microliter of anti-CD9 coated beads were incubated overnight at RT. After incubation and washing, the EV-bead complexes were stained with previously described Abs (same as for cells). The EV-bead complexes were visualized on FACS Canto II (BD Bioscience) and analyzed by FlowJo software (Tree Star).

EV proteins were extracted by diluting EVs in a 5× RIPA buffer (Cell Signaling Technology) and subjecting to multiple rounds of vortex and sonication. Protein concentration was determined by DC protein assay, and 10 μg of EV proteins were added to a 4× reducing Laemmli Sample buffer (Bio-Rad) and subsequently boiled for 5 min at 95°C. Then, the proteins were loaded on a Mini-Protean TGX precast gel (any kDa; Bio-Rad) and blotted to Trans-Blot Mini PVDF membranes using the Trans-Blot Turbo Transfer system (Bio-Rad). The PVDF membranes were blocked for 2 h in a 5% nonfat milk/PBST buffer at RT. After blocking, the membranes were washed with PBST and incubated with the primary Ab overnight at 4°C as follows: anti-OVA (clone 3G2E1D9; Nordic BioSite) in combination with a sheep-anti-mouse (NA9310V; GE Healthcare), or anti-MHCII (ab180779; Abcam) and anti-Hsp70 (clone 5A5; Abcam) combined with a donkey-anti-rabbit (NA9340; GE Healthcare). The bands were visualized by ECL (GE Healthcare) and read and analyzed by ChemiDoc MP Imaging System and Image Lab software version 4.1 (Bio-Rad).

Six-week-old C57BL/6 mice were injected i.v. with either 100 μl of PBS, 40 μg of B6 EVs, or 40 of μg BALB/c EVs on day 0 and day 7. The mice were bled from the tail vein at day 7. At day 14, the mice were sacrificed, and blood and spleens were collected. Total IgG, IgG1, IgG2b, IgG2c, and IgG3 serum levels of day 14 were determined by ELISA. Furthermore, anti–OVA-IgG, -IgG1, -IgG2b, -IgG2c and -IgG3 serum Abs of day 14 and anti-OVA–specific IgG of day 7 were measured by ELISA. Single-cell suspensions were created using a 100-μm cell strainer, followed by the removal of RBCs by using ACK lysis buffer. Splenocytes were counted and divided for the IFN-γ ELISPOT and flow cytometry analysis by LSR Fortessa III (BD Bioscience) and FlowJo software (Tree Star).

Six-week-old C57BL/6 mice were randomly divided in six groups and injected i.v. with either 100 μl of PBS (two groups) or 40 μg of B6 EVs (two groups) or BALB/c EVs (two groups). On day 21 the groups that were previously injected with either B6 or BALB/c EVs received either a boosting injection of EVs or a PBS injection as a control. Starting from day 7 the mice were bled every second week from the tail vein. At day 103 all groups except one of the PBS groups were injected i.v. with 40 μg of OVA. After the OVA boosting the mice were bled at day 4 and 9. At day 14 after OVA boosting the mice were sacrificed, and blood and spleens were collected. Total IgG, IgG1, IgG2b, IgG2c, and IgG3 serum levels of the end day were determined by ELISA. Furthermore, anti–OVA-IgG, -IgG1, -IgG2b, -IgG2c, and -IgG3 serum Abs of the end day and anti-OVA–specific IgG of all the bleeding days were measured by ELISA. Single-cell suspensions were created using a 100-μm cell strainer, followed by the removal of RBCs by using ACK lysis buffer. Splenocytes were counted and divided for the IFN-γ ELISPOT and flow cytometry analysis using Fortessa ×20 (BD Bioscience) and FlowJo software (Tree Star).

Thymocytes from naive C57BL/6 or BALB/c mice were collected, and a single-cell suspension was created by using a 100-μm cell strainer. After RBC lysis, thymocytes were incubated with Fc-block (BD Biosciences) for 30 min at 1 μg per 1 × 10⁶ cells. After washing, 2 μl of serum (day 14) from the booster experiment was incubated with 100,000 thymocytes for 1 h. This was followed by secondary total IgG polyclonal F(ab′)₂ Ab (eBioscience) incubation for 30 min and analyzed by FACS Canto II (BD Bioscience).

B16mOVA from Dr. T. Tedder (Duke University School of Medicine) was cultured in complete medium (RPMI 1640 supplemented with l-glutamine, 10% heat-inactivated FCS, 2-ME, and 400 μg/ml Geneticin. All tumor cells used for one experiment were derived from the same passage of the cells. A total of 50,000 cells dissolved in Matrigel (Corning) were injected s.c. in the right flank of female B57Bl/6N mice 6 wk old. After tumor establishment 100 μl of PBS or 40 μg of EVs B6 or BALB/c were i.v. injected at day 5 and day 12 after tumor inoculation. Mice were sacrificed when the tumor size reached 1000 mm³. Tumor tissue was incubated with Liberase (Roche) for 30 min at 37°C. The tumor tissue was then passed through a 100-μm cell strainer to create a single-cell suspension, and RBCs were removed by using ACK lysis buffer. Tumor cells were incubated with anti-CD16/CD32 Fc block (BD Biosciences). Followed by staining for CD8a (53-6.7), B220 (RA3-6B2), TCR-b (H56-597) (all from BioLegend) and CD45 (30-F11, BD Biosciences) and anti-H2Kb SIINFEKL pentamer (MP/5399-03; Proimmune) and analyzed by FACS Canto II (BD Bioscience).

Surface levels of OVA on the EVs were determined by ELISA by coating 10 μg of EVs to ELISA plates over night at 4°C, followed by incubation 2 h at RT with mouse anti-OVA Ab (Nordic Biosite) and detected by the secondary Ab anti-mouse IgG HRP (Southern Biotech) by incubation for 1 h at RT. TMB peroxidase substrate was used for detection according to the manufacturer’s protocol (BioLegend) and subsequently stopped by adding 1 M H2SO4 100 μl/well. Total IgG was determined by binding heavy L chain Ab (Southern Biotech) over night at 4°C, followed by 2 h RT serum incubation. For total IgG and IgG subclasses a standard control of the desired Ab subclass was added in a 1:3 dilution series and detected by alkaline phosphatase (AP)-conjugated isotype-specific Abs (Southern Biotech). OVA-specific IgG Ab levels were determined by coating the plates with 10 μg/ml OVA protein overnight at 4°C, followed by incubation for 2 h RT with serum and detected by isotype-specific IgG-AP Abs (Southern Biotech), followed by the addition of AP-substrate buffer. To determine the avidity of the anti-OVA IgG Abs in serum, the ELISA was performed in duplicates as described above with a minor modification. After the serum incubation the control plates were incubated with 100 μl of PBS, whereas the treated plates were treated with 100 μl of 6 M urea for 10 min. Avidity of the Abs was calculated by dividing absorbance of the treated samples by the absorbance of the untreated samples and plotted as percentages. All plates were read at 405 or 450 nm using an ELISA reader (Enspire 2300 Multilabel reader; Perkin Elmer).

IFN-γ ELISPOT was performed according to manufacturer’s instructions (Mabtech). A total of 200,000 splenocytes per well were plated out on PVDF plates (MilliporeSigma) in RPMI 1640 containing 10% FCS, 1% L-Glutamin, and 1% penicillin-streptomycin (Life Technologies). Cells were incubated for 19 h at 37°C with the following stimuli: 2 μg/ml Con A (Sigma), 5 μg/ml CD8 peptide SIINFEKL, 2 μg/ml CD4 peptide OVA323–339 (Innovagen), 1 μg/ml OVA, 1 μg/ml αGC, or 100 μl of medium as control. For detection of OVA-specific, Ab-forming cells the PVDF plates were coated with polyclonal goat anti-mouse IgG (Mabtech). Splenocytes were seeded in a 3-fold dilution starting at 1 × 10⁶ million cells per well. The cells were incubated in RPMI 1640 containing 10% FCS, 1% l-glutamin, 1% penicillin-streptomycin, and 500 μM 2- ME (Sigma) for 20 h at 37°C. After washing, 150 ng/well of biotinylated OVA (Thermo Fisher Scientific) was added for 2 h, followed by incubation with 100 μl of streptavidin-ALP in 0.5% FCS/PBS (Mabtech). Lastly, 100 μl of BCIP substrate (Mabtech) was used to develop the reaction. For detection of OVA-specific Ab-producing cells (i.e., memory B cells), the splenocytes were cultured for 5 d in complete medium containing 2 μg/ml LPS. ELISPOT was performed as described above with the addition of total IgG detection with a biotinylated goat anti-mouse IgG (Mabtech). OVA-specific Ab-producing cells were calculated by dividing the OVA-specific spot by the calculated total IgG spots. All ELISPOT plates were read using an AID iSpot FluoroSpot Reader System and analyzed by AID ELISPOT software (Autoimmun Diagnostika).

IL-6 and TNF-α dual color FluoroSpot was performed according to manufacturer’s instructions (Mabtech). A total of 25,000 splenocytes per well were plated out on low fluorescent PVDF membrane plates (Merck Millipore) in RPMI 1640 containing 10% FCS, 1% L-Glutamine, and 1% penicillin-streptomycin (Life Technologies). Cells were incubated for 24–48 h at 37°C with EVs 1or 5 μg/well or with cell culture medium (negative control). For detection of TNF-α and IL-6, tag-conjugated Abs were used and fluorescently labeled anti-tag Abs for multiplexing. The IL-6 biotinylated detection Ab was visualized using streptavidin labeled with a 550 fluorophore (Cy3 equivalent), and TNF-α WASP-conjugated detection Ab was detected using anti-WASP Ab labeled with 640 fluorophore (Cy5 equivalent). The plates were read using the IRIS reader (Mabtech) and analyzed by Apex software using the RAWspot technology for spot analysis (Mabtech) using LED550 and LED640 light sources.

The luminex was performed according to manufactures instructions (Bio-Plex Pro Mouse Cytokine Th1/Th2 Assay; Bio-Rad). The plasma samples were added in a 1:4 dilution and incubated for 1 h at RT with the beads. After developing, the plates were read using a Bio-Plex200 reader (Bio-Rad).

We prepared and characterized BMDC-derived EVs from C57BL/6 (B6) (syngeneic) and BALB/c (allogeneic) mice mainly according to the Minimal Information for Studies of Extracellular Vesicles guidelines (20). EVs were exposed to EM, which showed that in both vesicle preparations EVs could be detected with expected small EV characteristics (i.e., 30–150-nm–sized round vesicles (Fig. 1A). Particle size was compared by using NTA, presenting a mean size of 165.9 and 136.6 nm (mode size of 124 and 99.4 nm) for B6 and BALB/c EVs, respectively, thus indicating a similar size distribution between vesicles of the two sources (Fig. 1B). EVs were bound to Ab-coated latex beads and phenotyped for surface markers by flow cytometry. The mean fluorescence intensity (MFI) indicated expression levels above background (isotype control) for markers enriched in EVs, such as the tetraspanins CD9 and CD81. In addition, the costimulatory molecules CD40, CD80, CD86, and MHCII (IA/IE) were all present and at similar levels on the surface of EVs from the two sources (Fig. 1C). The phenotype of the BMDCs was also analyzed using flow cytometry, and no phenotypic differences on a cellular level were observed between B6 or BALB/c cells (Supplemental Fig. 1A). In addition to the detection of tetraspanin molecules, heat shock protein (Hsp) 70 levels were examined by Western blot to serve as an intravesicular EV marker, and both B6 and BALB/c EVs presented similar Hsp 70 levels (Fig. 1D). Considering the equal amount of EV protein loaded, the amount of OVA protein was detected at comparable levels by Western blot (Fig. 1D) in both syngeneic and allogeneic EV preparations. Moreover, OVA on EVs was detected at similar levels by ELISA for both B6 and BALB/c EVs (Fig. 1E). Next, the stimulatory effect of both EV preparations was determined in a Fluorospot assay by stimulating naive B6 splenocytes for 48 h with either B6 or BALB/C EVs. Both B6- and BALB/c–derived EVs induced similar levels of IL-6 or TNF-α secretion (Supplemental Fig. 1B). These data show that B6 and BALB/c EVs are phenotypically similar defined by surface marker expression, size distribution, Ag content, and in vitro innate activation capacity, thus suggesting that they can be compared for in vivo activating capacity.

Next, we compared the capacity of syngeneic and allogeneic EVs to stimulate immune responses in vivo. To do this, B6 recipient mice were injected twice with BMDC-derived EVs from B6 or BALB/c mice loaded with OVA according to the immunization schedule (Fig. 2A). Flow cytometric analysis of splenocytes on day 14 showed similar levels of total CD8⁺ T cells in all treatment groups (Fig. 2B), whereas only allogeneic EVs significantly increased the OVA-specific CD8⁺ T cells (Fig. 2C). Splenocyte ex vivo restimulation showed no IFN-γ response in any of the treatment groups toward the CD4 peptide (OVA_323–339) (Fig. 2D), but splenocytes from both syngeneic and allogeneic treatment groups produced IFN-γ upon MHCI-restricted CD8 peptide (SIINFEKL) stimulation (Fig. 2E). Restimulation with whole OVA induced a significant IFN-γ response in splenocytes derived from mice injected with BALB/c EVs compared with those injected with B6 EVs (Fig. 2F). Importantly, BALB/c–injected mice also produced higher levels of OVA-specific total IgG Abs (Fig. 2G), suggesting that allogeneic EVs are strong immune activators and may function as adjuvants in this setting.

We have previously shown that αGC on EVs leads to IFN-γ production by iNKTs and enhances Ag-specific responses. We therefore examined whether allogenicity could further boost the effect of EVs loaded with both OVA and αGC when injected twice according to the immunization schedule in Fig. 2A. Flow cytometry analyses of splenocytes showed that OVA-specific CD8⁺ T cells were significantly upregulated in both EV treated groups compared with control (Fig. 3A). In addition, an increase in GC B cells was observed in mice injected with either B6 or BALB/c EVs compared with controls (Fig. 3B). Furthermore, we observed a significant increase in the proportion of Tfh cells only after injection of BALB/c EVs (Fig. 3C). All treatment groups were nonresponsive to ex vivo restimulation of splenocytes with CD4 peptide (OVA_323–339) (Fig. 3D), whereas both EV-injected groups upregulated the number of IFN-γ–secreting cells upon CD8 peptide (SIINFEKL) stimulation (Fig. 3E). Of note, in response to whole OVA protein only, BALB/c EV-injected mouse splenocytes showed increased IFN-γ production, significantly higher than both PBS control and B6 EV-injected mice (Fig. 3F). Only low levels of anti-OVA IgG were measured in serum day 7 after one immunization (Supplemental Fig. 2A). On day 14 after immunization, the levels of Ag-specific Abs, anti-OVA IgG (Fig. 3G), anti-OVA IgG1 (Fig. 3H), and anti-OVA IgG2c (Fig. 3I) were significantly higher in the serum from mice injected with BALB/c EVs compared with B6 EVs. Furthermore, higher Ag-specific Ab levels at day 14 were also detected in BALB/c EV-injected mice compared with B6 EV-injected mice for anti-OVA IgG2b (Supplemental Fig. 2B) and anti-OVA IgG3 (Supplemental Fig. 2C). Although no effects were seen in any of the groups for total Ab levels (total IgG) (Supplemental Fig. 2D), only BALB/c EVs induced a significant increase in IgG1 (Supplemental Fig. 2E) and IgG2c (Supplemental Fig. 2F). To investigate if the mismatched EVs induced Abs toward allogeneic MHC molecules, we incubated the day 14 serums with thymocytes derived from naive B6 or BALB/c mice. After incubation with serum from B6 mice injected with B6 or BALB/c OVA/αGC–loaded EVs, thymocytes were stained with an anti-total IgG Ab to detect alloreactive Abs according to the experimental setup (Fig. 3J). We detected an increase in MFI above background in BALB/c thymocytes incubated with serum from B6 mice injected with BALB/c but not B6 EVs (Fig. 3K), indicating that the serum contained Abs toward the allogeneic MHCI complex and thus induced an MHC mismatch response. Taken together, these results suggest that allogeneic EVs induce an allogeneic immune response that leads to bystander activation and augmented Ag-specific B cell responses.

As humoral adaptive responses were shown enhanced to a higher extent by allogenicity than cellular responses, we wanted to investigate the long-term effect of syngeneic and allogeneic EVs, respectively. Mice were injected during a 21-d interval as shown in the experimental layout (Fig. 4A), and Ab levels were followed every second week during the experiment (Fig. 4B). Individual days are shown to display statistical significance (Fig. 4C–E). Notably, mice receiving two injections of either syngeneic or allogeneic EVs produced more Ag-specific Abs (anti-OVA IgG) after 28 d compared with one injection (Fig. 4C) and these levels stayed elevated over time (Fig. 4B). On day 103, mice were injected i.v. with free OVA, and this Ag rechallenge boosted anti-OVA IgG production in all groups, except PBS control, independently of syngeneic or allogeneic EV injection and regardless of one or two injections. Ag-specific Ab levels were measured after the OVA boosting on day 107 (Fig. 4D) and on the end day of the experiment (Fig. 4E). Moreover, at the experimental endpoint, day 117, the number of Ab-forming cells (i.e., plasma cells that produced OVA-specific Abs) were significantly elevated in mice injected twice as well as those injected once with allogeneic EVs (Fig. 4F). Importantly, the OVA-specific IgG Abs also had the highest avidity after two injections with EVs (Fig. 4G). This indicates that a booster injection not only increases the Ab levels but also improves Ab quality, and long-term, allogenicity is less important than administering two injections. Also in this study, splenocytes restimulated ex vivo with CD4 peptide (OVA_323–339) did not upregulate IFN-γ production (Fig. 4H). In response to the CD8 peptide (SIINFEKL), all EV treatment groups significantly upregulated IFN-γ production compared with the PBS control (Fig. 4I). In addition, the two times–injected groups significantly upregulated the number of IFN-γ–producing cells when restimulated with whole OVA protein (Fig. 4J). No differences in memory T cell populations (defined as CD4⁺ or CD8⁺, CD62L^low, CD44^high, Supplemental Fig. 3A) were observed in the spleen memory CD4⁺ T cells (Supplemental Fig. 3B) or memory CD8⁺ T cells by FACS analysis (Supplemental Fig. 3C). In the long-term experiments, both OVA-specific IgG1 (Supplemental Fig. 3D) and IgG2c (Supplemental Fig. 3E) were the highest after two injections, but after one injection only mice injected with allogeneic EVs showed a significant response (Supplemental Fig. 3D). To determine whether allogeneic EVs induce long-term inflammation, the end-day sera were analyzed for the presence of inflammatory cytokines. The absence of these cytokines in the sera indicated that both B6 and BALB/c EVs are safe to administer (Supplemental Fig. 3F). Overall, we conclude that two injections of allogeneic EVs are superior for mounting a good adaptive immune response for immunotherapeutic use.

With the positive effects of two injections of allogeneic EVs on immune responses, we aimed to compare the syngeneic and allogeneic EVs in the B16mOVA melanoma mouse model. Mice were injected with 50,000 tumor cells s.c. in the right flank followed by two EV injections with a 7-d interval (day 5 and day 12). Mice were sacrificed when the tumor size reached 1000 mm³, summarized in the experimental layout (Fig. 5A). Both syngeneic and allogeneic EV-injected mice showed a significantly increased overall survival compared with the control; mean survival for the control mice was 21 d compared with the EV-treated groups with a mean survival of 26 d (Fig. 5B). A prominent difference in tumor volume between control and EV treatment groups was observed on day 19 (Fig. 5C). In addition, mice injected with EVs had more Ag-specific CD8⁺ T cell infiltration in the tumor compared with the control; however, no differences were observed between the syngeneic and allogeneic EVs (Fig. 5D). Furthermore, both EV-injected groups also displayed higher OVA-specific Ab titers compared with the PBS control group (Fig. 5E). We conclude that two injections of both syngeneic and allogeneic EVs efficiently delay tumor progression.

In the recent years, EVs have gained interest as therapeutic delivery tools, as they are able to induce immune responses. In cancer therapy, DC-derived EVs have been explored in clinical trials (10–12). These studies demonstrated that EVs were safe to administer and that they induced NK cell activation. However, they failed to induce significant activation of CTL or increase the overall survival in the patients. Our previous findings suggest that whole protein–loaded rather than peptide-loaded EVs are more potent in providing an immune stimulatory effect in vivo (15). We also showed that the MHC molecules on the EVs were dispensable to elicit an immune response, suggesting the possibility to use allogeneic EVs (17). We have previously seen that one injection of Ag-loaded allogeneic EVs is equally good at inhibiting tumor progression as a single injection of syngeneic EVs (17). In the current study, we show that two injections of allogeneic EVs induce a potent Ag-specific immune response and that this induces significantly higher IgG Ab production compared with syngeneic EVs.

Upon repeated injections of OVA-loaded allogeneic EVs, Ag-specific CD8⁺ T cells were significantly upregulated compared with the control. Furthermore, restimulation of splenocytes with whole OVA protein resulted in a significantly higher IFN-γ production in mice injected with allogeneic EVs. Moreover, OVA-specific IgG Ab levels were significantly higher in the sera of allogeneic EV-injected mice. This enhanced effect of allogeneic EVs could be caused by the foreign MHC complex that acts as an alloantigen that enhances the effect of the EVs via bystander activation. Also, the alloantibodies formed after the first injection might serve as opsonizing agents, which could lead to allogeneic EVs that have a different biodistribution or more favorable uptake by recipient cells and are therefore processed more efficiently. Indeed, the characterization of the EVs did not show any differences in immune-relevant markers or OVA between the two species, but we could demonstrate that allogeneic EVs induced low amounts of allo-MHC–specific Abs. These alloantibodies were not detected in the serum collected in the long-term experiment, which suggests that allogeneic EVs are well tolerated in this model. The presence of alloantibodies may explain the humoral-biased effect of allogeneic EVs, as they are able to form immune complexes with EVs, which leads to an extended display of the EVs and Ags on follicular DCs in the spleen and thereby increasing B cell activation (21).

To further improve the immunogenicity of BMDC-derived EVs, αGC was loaded to the BMDCs (13, 17). A double injection of OVA- and αGC-loaded EVs resulted in more Ag-specific CD8⁺ T cells than OVA-loaded EVs, although there was no difference between the allogeneic and syngeneic groups. We have previously shown that repeated injections of αGC-EVs do not induce anergy of iNKTs as αGC alone does (13, 22, 23). We observed a significant increase in Tfh cells in the spleen upon (OVA αGC)–loaded allogeneic EVs, further supporting that allogeneic EVs boost the humoral immune response.

Because repeated injections of both syngeneic and allogeneic EVs indicated that EVs could be a potential agent for preventive vaccination (e.g., infectious diseases), we tested whether EVs also induced a long-term memory response. Our data demonstrate that two injections provide a more efficient immune response with persistent high levels of Ag-specific IgG titers after the immunization. Also, the increased number of Ab-forming cells and higher avidity of the serum Abs demonstrates that the two-times injection strategy improves the quality of long-term memory.

EVs have been used in cancer therapy. Therefore, we tested whether allogeneic EVs could be used to treat melanoma. Our study supports our previous findings that allogeneic EVs are equally efficient as syngeneic EVs in slowing down tumor growth in the B16 melanoma, suggesting that allogeneic EVs can be used in human cancer immunotherapy (13). Furthermore, our current study shows that a two-time injection of allogeneic and syngeneic EVs induce higher levels of tumor-infiltrating CD8⁺ T cells, a feature that has been associated with better prognosis of multiple cancers (24). Possibly, supported by the immunogenicity data, certain cancer types might even benefit from allogeneic EVs compared with autologous ones. Importantly, the use of allogenic EVs in clinical immunotherapies facilitates the use of cell lines for production of EVs and thereby overcoming batch-to-batch variations, fluctuating cell vaccine qualities, and providing a more time- and cost-beneficial approach (25).

In conclusion, the current study demonstrates that allogeneic EVs induce potent Ag-specific immunity. We demonstrate for the first time, to our knowledge, that EVs provide long-term immunity, indicating their potential in different vaccination strategies. Importantly, allogeneic and syngeneic EVs were both capable of efficiently inhibiting tumor progression in our model. Taken together, our findings highlight the possibility to use allogeneic EVs in impersonalized immunotherapeutic approaches in the future.

We thank Kjell Hultenby and Karolinska Institutet for excellent EM expertise, Sebastian Ols (Karolinska Institutet) for scientific advice with setting up the B cell ELISPOT, Kajsa Prokopek (Mabtech) for valuable scientific advice regarding B cell memory experiments, and the animal technicians at the Astrid Fagraeus Laboratory, Karolinska Institutet for assistance during the animal experiments.

S.G. has a patent on B cell–derived EVs in immune therapy and is a scientific advisor for Anjarium Biosciences and Neurotrauma Sciences.