More Efficient Induction of HLA-A*0201-Restricted and Carcinoembryonic Antigen (CEA)–Specific CTL Response by Immunization with Exosomes Prepared from Heat-Stressed CEA-Positive Tumor Cells

Carcinoembryonic antigen (CEA) is a heavily glycosylated oncofetal antigen that is overexpressed in human adenocarcinomas, especially in colon, pancreas, breast, and lung (1–3). The tissue expression pattern makes CEA a potential target for tumor-specific immunotherapy. Recent studies have revealed that CEA-specific CTLs can be induced after the administration of CEA protein mixed with adjuvants, altered peptide of CEA, anti-idiotype antibodies of CEA, and CEA-based recombinant poxvirus vaccines (4–9). However, as a self-protein and due to immune tolerance, CEA is poorly immunogenic; thus, more efforts should be put into exploring a successful strategy to improve the efficiency of CEA-based tumor immunotherapy.

Exosomes are membrane vesicles, with a diameter of 30 to 90 nm, derived from the fusion of small internal compartments with the plasma membrane in many types of cells, including tumor cells (10, 11). Recently, studies have suggested that exosomes can serve as a new kind of vaccine with promising therapeutic effects in cancer immunotherapy. Exosomes derived from tumor antigen peptide-pulsed dendritic cells elicit potent tumor-specific immune responses (12). Exosomes derived from tumor cells are also a source of shared tumor rejection antigens for CTL cross-priming in animal model (13). However, the antitumor activity of the pure exosomes applied are not strong and thus need to be enhanced. In combination with proper adjuvants, exosome-based cancer vaccines can enhance the host immune responses against tumors. For example, recently, it is reported that exosomes admixed with CpG oligonucleotides were efficient in prophylactic and therapeutic settings of melanoma in HLA-A2 transgenic mice (14).

Accumulating evidence has shown that hyperthermia is a promising approach in cancer therapy; its possible mechanism of antitumor immunity is attributed to the high expression of heat shock protein (HSP) 70 and MHC-I on heat-stressed tumor cells (15–18). HSP70 prepared from tumor cells or virus-infected cells can elicit potent antigen-specific CD8⁺ CTL response and therapeutic effects. It is an effective molecular adjuvant for the induction of a Th1 immune response. Evidence has revealed that by priming antigen-presenting cells, especially dendritic cells, HSP70 exhibits potent adjuvant functions in stimulating the host immune response (19–23) and has a potent antitumor effect in animal model (21–23). As such, clinical trials using HSP70 for tumor therapy are being carried out (24).

In this study, we investigated whether exosomes from the heat-stressed CEA⁺ human tumor cells (CEA⁺/HS-Exo) are capable of inducing CEA-specific antitumor immune response more efficiently than the exosomes conventionally prepared from the same tumor cells (CEA⁺/Exo). Our results show that CEA⁺/HS-Exo contain more HSP70 and MHC-I molecules and exhibit superior immunogenicity than CEA⁺/Exo. Furthermore, CEA⁺/HS-Exo induce a potent CEA-specific CTL response both in vitro and in vivo, suggesting that exosomes from the heat-stressed tumor cells could be potential vaccines for cancer immunotherapy.

Materials. Peridinin chlorophyll protein–labeled anti-HLA-DR, FITC-conjugated anti-CD80, phycoerythrin (PE)–conjugated anti-CD86, PE-conjugated anti-CD40, and FITC-labeled anti-CD8 were purchased from BD Biosciences PharMingen (Chicago, IL). FITC-conjugated anti-HLA-A2 monoclonal antibody (clone BB7.2) was from Serotec Ltd. (Oxford, United Kingdom). Anti-human CEA (CE05) was from Neomarker (Fremont, CA). Anti-HSP27 (F-4), anti-HSP40 (C-20), anti-HSP60 (H-1), anti-HSP70 (K-20), anti-Hsc70 (B6), anti–intercellular adhesion molecule-1 (C-19), anti-B7.1 (N-20), anti-B7.2 (C-19), anti-transferrin receptor (K-20), and anti–lysosome-associated membrane glycoprotein-3 were from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). Anti-MHC-I (HC10) was from Roswell Park Cancer Institute (Buffalo, NY). Horseradish peroxidase–coupled secondary antibodies were from Jackson ImmunoResearch (West Grove, PA). Recombinant mouse interleukin (IL)-2 and recombinant human IL-2 were commercially obtained from Sigma (St. Louis, MO). Recombinant human IL-7 and IL-10 were from Peprotech, Inc. (Rocky Hill, NJ). Micro SSp DNA Typing for confirmation of HLA-A*0201 expression was purchased from One Lambda, Inc. (CA). The protein assay kit was purchased from Bio-Rad, Inc. (Hercules, CA). The Limulus amebocyte lysate assay was from BioWhittaker (Walkersville, MD). Aldehyde/sulfate latex beads (4 μm in diameter) were from Interfacial Dynamics (Portland, OR). Chemiluminescence detection kits were from Pierce (Rockford, IL). HLA-A*0201/CAP-1-tetramers^PE were from ProImmune (Oxford BioBusiness Centre, Littlemore Park, Oxford, United Kingdom). HLA-A*0201/SSp-1-tetramers^PE were constructed in our laboratory. Ficoll-Paque, [³H]thymidine, and ⁵¹Cr sodium chromate were from Amersham Pharmacia Biotech (Arlington Heights, IL). Fetal bovine serum was from Hyclone (Logan, UT). Mouse and human IFN-γ, human IL-12p70, and tumor necrosis factor-α ELISA kits were from R&D Systems (Minneapolis, MN). Anti-CD3 and anti-CD4 magnetic beads were from Miltenyi Biotech (Auburn, CA).

Peptides. The CEA peptide CAP-1 (YLSGANLNL) and the coronavirus SARS-CoV spike protein peptide SSp-1 (RLNEVAKNL) were synthesized at GL Biochem Ltd. (Shanghai, China) with purity of >95%.

Animals and cell lines. HLA-A2.1/K^b transgenic mice (6-8 weeks of age) were purchased from The Jackson Laboratory (Bar Harbor, ME). Four-week-old athymic nude mice were obtained from SIPPR-BK Experimental Animal Co. (Shanghai, China). The mice were housed in a pathogen-free facility for all experiments.

Human colon carcinoma cell lines LS-174T (HLA-A2⁺CEA⁺), SW480 (HLA-A2⁺CEA⁺), LoVo (HLA-A2⁻CEA⁺), human lung carcinoma cell line A549 (HLA-A2⁺CEA⁻), and human TAP-deficient T2 cell lines were obtained from the American Type Culture Collection (Manassas, VA) and maintained in culture medium according to the supplier's specifications.

Patients and healthy donors. Patients and HLA-A*0201⁺ healthy donors were required to give informed consent before participation in the study. Entry criteria for patients with colon cancer were as follows: histologic confirmation of CEA expression (as defined by immunohistochemical analysis), confirmation of HLA-A*0201 expression by PCR-based DNA typing, high level of CEA in serum, no chemotherapy or radiotherapy 4 weeks before the study, and no organic dysfunction of the liver and kidneys.

Exosome purification. Exosomes were purified from 48-hour supernatants of 80% confluent tumor cells cultured in medium deprived of bovine exosomes by overnight centrifugation of bovine serum at 100,000 × g (25). To prepare heat-stressed tumor cell–derived exosomes (HS-Exo), tumor cells were first cultured at 37°C for 43 hours, sequentially incubated in a 43°C water bath for 1 hour, and recovered in a 37°C incubator for 4 hours. Exosomes derived from both non-heat-stressed (Exo) and heat-stressed tumor cells were purified as described previously (26). Exosome pellets obtained after 100,000 × g centrifugation for 1 hour were washed once in a large volume of PBS and then centrifuged at 100,000 × g for 1 hour, and the pellets were resuspended in 50 to 100 μL PBS. For further purification, the exosome suspensions were layered onto a 30% to 45% (w/v) discontinuous density sucrose cushion (density 1.13-1.19 g/mL) and centrifuged at 100,000 × g for 2 hours; then, exosome interface was collected and washed with PBS. The protein concentrations of exosomes were quantified by Bradford assay. All exosomes are free of endotoxin as confirmed by Limulus amebocyte lysate assay.

Electron microscopy. The purified exosomes were fixed for 1 hour in 4% paraformaldehyde and washed once with PBS. Then, the pellets were fixed in 2.5% glutaraldehyde, loaded on Formwar/carbon–coated EM grids, postfixed in 1% glutaraldehyde, and contrasted successively in 2% methycellulose/0.4% uranyl acetate (pH 4.0). Observations were made with a Philips EM410 electron microscopy (Eindhoven, the Netherlands).

Fluorescence-activated cell sorting analysis of HLA-A2 onexosomes. Exosomes were analyzed by fluorescence-activated cell sorting (FACS) analysis as described previously with minor modification (27). Briefly, 20 μg exosomes or 20 μg FCS proteins (negative control) were incubated with 5 μL of 4-μm-diameter aldehyde/sulfate latex beads for 15 minutes at room temperature in a 20-μL final volume followed by gentle shaking for 1 hour in 1 mL PBS and then centrifuged. The pellet was blocked by incubation with 20 μL FCS for 30 minutes. Exosomes or FCS-coated beads were washed thrice in PBS and resuspended in 50 μL PBS. In parallel, heat-stressed or non-heat-stressed LS-174T cells were washed twice in PBS. Cells (10⁵) or 20 μL coated beads were incubated for 45 minutes with FITC-conjugated anti-HLA-A2 monoclonal antibody, washed, and analyzed on a FACSCalibur (Becton Dickinson, Mountain View, CA).

Western blot analysis. The same amount of exosomes or cell lysate proteins was separated on SDS-PAGE and transferred to Bioblot-NC membranes (Costar, Nepean, Ontario, Canada). The membranes were blocked for 1 hour in TBST containing 5% nonfat milk and then incubated with primary antibodies as indicated at the supplier's recommended dilutions followed by horseradish peroxidase–coupled secondary antibodies and chemiluminescence detection.

Dendritic cell maturation. Human peripheral blood monocyte–derived dendritic cells were generated as described by us (28). On day 5, monocyte-derived dendritic cells were stimulated with 5 μg/mL exosomes or 1 μg/mL lipopolysaccharide (LPS) or PBS for 48 hours, harvested, and washed. For phenotypic analysis, dendritic cells were incubated with the indicated labeled antibodies for 45 minutes at 4°C before flow cytometry analysis. To quantify cytokine secretion, day 5 dendritic cells (5 × 10⁵/mL) were stimulated with 5 μg/mL exosomes or 1 μg/mL LPS or PBS for 24 hours. IL-12p70 and tumor necrosis factor-α in supernatants were measured by ELISA kits (29).

Mixed lymphocyte reaction. Allogeneic T-cell response was assessed as described previously (30). On day 5, immature dendritic cells were incubated with 5 μg/mL exosomes or medium (negative control) and 1 μg/mL LPS (positive control) for 48 hours and then irradiated as stimulator cells. T cells were enriched from peripheral blood mononuclear cells of a different donor using anti-CD3 magnetic beads as responder cells and incubated with the irradiated dendritic cells at different responder/stimulator ratios as indicated. [³H]Thymidine (0.5 μCi/well) was added 72 hours later, and cells were incubated for further 18 hours. [³H]Thymidine incorporation was assayed by liquid scintillation counting.

Cytotoxicity assays. Cytotoxicity assays were done using a standard 4-hour ⁵¹Cr release assay as described (31). T2 cells are pulsed with CAP-1 or SSp-1. Peptide-pulsed T2 cells and SW480 and LoVo cells were labeled with ⁵¹Cr sodium chromate (100 μCi/10⁶ cells) in triplicate for 90 minutes at 37°C and washed thrice as target cells. Target cells (10⁴ per well) and effector cells were plated in a final volume of 200 μL in 96-well round-bottomed plate. After 4 hours, each supernatant (100 μL) was harvested and then measured for the release of ⁵¹Cr (gamma counter). The specific cytotoxicity was determined according to the following formula: [(counts/min of experimental ⁵¹Cr release − counts/min of the spontaneous ⁵¹Cr release) / (counts/min of the maximal ⁵¹Cr release − counts/min of the spontaneous ⁵¹Cr release)] × 100%. The spontaneous ⁵¹Cr release was determined by incubating the target cells alone in the absence of effectors, and the maximal ⁵¹Cr release was obtained by incubating the targets with 5% Triton X-100. The spontaneous release was always <15% of maximum release.

CTL induction in HLA-A2.1/K^b transgenic mice. HLA-A2.1/K^b transgenic mice were s.c. immunized with 5 μg exosomes on the posterior right of the back weekly for three times. The control groups were injected with A549 (HLA2⁺CEA⁻)–derived exosomes or PBS. For the dosage escalation experiments, 2.5, 5, 10, and 15 μg exosomes were used for each vaccination per mouse. The mice were sacrificed 7 days after the third immunization, and murine bone marrow–derived dendritic cells were prepared as described previously (32). On day 7, dendritic cells were harvested and pulsed with 10 μg/mL peptides. Splenocytes (2 × 10⁶) from immunized transgenic mice were restimulated with CAP-1-loaded irradiated syngeneic dendritic cells (2 × 10⁵) for 6 days. The cytotoxicity was assayed as described above.

For the IFN-γ release assay, splenocytes (2 × 10⁵) were restimulated with 10 μg/mL CAP-1, SSp-1, or 5 μg/mL concanavalin A for 72 hours in 96-well round-bottomed microculture plates, the supernatants were collected for IFN-γ measurement. For the tetramer staining assay, the lymphocytes from injected side draining lymph node of immunized transgenic mice were tested according to the manufacturer's protocol.

Adoptive transfer. Four-week-old athymic nude mice were inoculated s.c. on the right side, behind the anterior forelimb with SW480 tumor cells (1 × 10⁷) in 0.2 mL serum-free culture medium, as described previously (33). On day 7, when tumors reached 4 to 5 mm in diameter, the mice were randomly divided into several groups (n = 5). On days 7 and 14, splenocytes (2 × 10⁸) from exosome-immunized transgenic mice were injected i.p. in SW480-bearing nude mice. From days 7 to 21, recombinant IL-2 (10,000 units) was given i.p. Control mice received recombinant IL-2 alone or PBS alone. Tumor growth was measured every other day in two dimensions using a digital caliper. Tumor volume was calculated as (length × width²) / 2 and presented as mean ± SD (mm³). The survival rate was observed.

In vitro CTL generation from HLA-A*0201⁺ healthy donors and HLA-A*0201⁺CEA⁺ cancer patients. CTL were generated as described previously by us (31). Briefly, autologous dendritic cells (2 × 10⁵ per well) were pulsed with 10 μg exosomes in a 24-well plate for 2 hours in 200 μL culture medium. Lymphocytes were then added at a ratio of 10 lymphocytes to 1 dendritic cell in a final volume of 2 mL/well in the presence of 10 ng/mL IL-7. One day later, recombinant human IL-10 was supplemented to the culture to a final concentration of 10 ng/mL. After 7 days, lymphocytes were restimulated with exosome-pulsed autologous dendritic cells in medium containing IL-7 (10 ng/mL) and IL-10 (10 ng/mL) and then supplemented with 20 units/mL recombinant hIL-2 a day later. Individual wells were restimulated separately every 7 days up to four cycles with dendritic cells pulsed with exosomes. On day 28, stimulated lymphocytes were harvested and tested by IFN-γ release assay, cytotoxicity capacity, and tetramer staining. For IFN-γ release assay and cytotoxicity assays, CD8⁺ T lymphocytes were purified by CD4⁺ cell–negative depletion using human CD4 microbeads. IFN-γ release assay was done as described by Kim et al. (34), with minor modification. Peptide-pulsed T2 cells were used as stimulator cells. Effector cells (5 × 10⁴) and stimulator cells (1 × 10⁴) were cocultured in 96-well microplates. After 24 hours of incubation, supernatants were collected and measured for IFN-γ production by ELISA.

Statistical analysis. The differences of the tumor diameters were compared using the Mann-Whitney U test. The differences of the survival were done using Kaplan-Meier test. All other statistical analyses were done using a two-tailed t test. Ps ≤ 0.05 were considered significant.

Heat stress enriches heat shock protein 70 and MHC-I proteins in exosomes. The pellets obtained from ultracentrifugations of culture supernatants of LS-174T cell lines were first analyzed morphologically by electron microscopy. The pellets consisted of 30- to 90-nm-diameter bilayer membrane vesicles (Fig. 1A) and showed typical exosome morphology described by others (10–14). The components in Exo and HS-Exo were analyzed by Western blot and by FACS analysis of exosome-coated beads. The native tumor antigen CEA was detected in all exosomes derived from CEA⁺ tumor cell lines (LS-174T and LoVo cell lines) but not in exosomes derived from CEA⁻ tumor cell line (A549 cell line), and it was not altered after heat stress (Fig. 1B). Inducible and constitutive HSPs, including HSP70 and Hsc70 but not HSP60, HSP40, and HSP27, were selectively detected in Exo and HS-Exo, but HSP70 was greatly enriched in HS-Exo (Fig. 1B). High levels of MHC-I molecules were found in Exo and accumulated in HS-Exo (Fig. 1B), consistent with previous reports showing that exosomes derived from dendritic cells, B cells, or tumor cells bear MHC-I (12, 13, 35, 36). Moreover, transferrin receptor and lysosome-associated membrane glycoprotein-3 (a tetraspanin that is a specific marker of late endosomes) were detectable in both Exo and HS-Exo, suggesting that exosomes were originated from a late endosomal compartment (Fig. 1B). However, adhesion molecules (intercellular adhesion molecule-1) and costimulatory molecules (B7.1 and B7.2) are undetectable in our study (Fig. 1B).

To elucidate whether exosomes from HLA-A2⁺ LS-174T cells (LS-Exo or LS-HS-Exo) bear HLA-A2 and whether the expression of HLA-A2 was different after tumor cells were subjected to heat stress, exosome-coated beads were analyzed by FACS. Both LS-Exo and LS-HS-Exo carried HLA-A2, but the mean fluorescence value increased in LS-HS-Exo (Fig. 1C), in line with the data that heat stress enhanced the surface expression of HLA-A2 in LS-174T cells (Fig. 1D). As a negative control, HLA-A2 molecules were not found in LoVo-derived exosomes (LoVo-Exo and LoVo-HS-Exo; Fig. 1C). These results suggest that heat stress causes the accumulation of MHC-I and HLA-A2 in HS-Exo of HLA-A2⁺ tumor cells.

Therefore, up-regulating HSP70/MHC-I and causing HSP70/MHC-I enrichment in exosomes through heat stress possibly offer great potential as a new approach to enhance immunogenicity.

Exosomes from heat-stressed tumor cells induce dendritic cell maturation more efficiently. Based on the report that mast cell–derived exosomes induce dendritic cell maturation (29), we analyzed the phenotypic and functional alterations of dendritic cells on exposure to different types of exosomes. The results showed that LS-HS-Exo strongly up-regulated the expression of HLA-DR, CD86, and CD40 in dendritic cells compared with LS-Exo (Fig. 2A). IL-12p70 and tumor necrosis factor-α production in the culture supernatants was greatly enhanced by LS-HS-Exo, to the extent comparable with that of LPS but significantly higher than that of LS-Exo (Fig. 2B and C). In mixed lymphocyte reaction, LS-HS-Exo-, LS-Exo-, and LPS-treated dendritic cells all induced T-cell proliferation. However, dendritic cells loaded with LS-HS-Exo were more effective in stimulating a proliferative response than that with LS-Exo (Fig. 2D). These data show that LS-HS-Exo are more effective in inducing phenotypic and functional maturation of dendritic cells than LS-Exo, suggesting that the function of enriched HSP70 in HS-Exo may be enrolled in induction of dendritic cell maturation.

More efficient induction of carcinoembryonic antigen–specific CTL response in HLA-A2.1/K^b transgenic mice by CEA⁺/HS-Exo. To assess the capacity of tumor cell–derived exosomes to induce an effective T-lymphocyte-mediated immune response in vivo, we immunized HLA-A2.1/K^b transgenic mice with 2.5, 5, 10, or 15 μg exosomes. Cytotoxicity assay showed that LS-HS-Exo or LoVo-HS-Exo (5 μg) was efficient in the generation of substantial immunity in HLA-A2.1/K^b transgenic mice, whereas the same amounts of LS-Exo or LoVo-Exo were insufficient (Fig. 3A). CEA-specific HLA-A2.1-restrcited CTL could be more readily induced in HLA-A2.1 transgenic mice by immunization with 5 μg LS-HS-Exo or LoVo-HS-Exo than by immunization with LS-Exo or LoVo-Exo as judged by the lysis of SW480 (HLA-A2⁺CEA⁺) and CAP-1 pulsed T2 cells (data not shown) but not LoVo (HLA-A2⁻CEA⁻; Fig. 3A), SSp-1 pulsed T2 cells and native T2 cells (data not shown) and by an IFN-γ release (Fig. 3B). CEA-A549-derived exosomes (A549-Exo or A549-HS-Exo) could not induce CEA-specific CTL response (Fig. 3A and B).

CAP-1-specific CD8⁺ T cells from the draining lymph node of injection sites in LoVo-HS-Exo- or LS-HS-Exo-injected transgenic mice made up 2.73% and 4.21% of the total number of CD8⁺ T cells by HLA-A*0201/CAP-1-tetramer staining, respectively, but tetramers of control epitope HLA-A*0201/SSp-1-tetramer were not detected (0.03% and 0.02%, respectively; Fig. 3C).

The results suggest that CEA⁺/HS-Exo (either LS-HS-Exo or LoVo-HS-Exo) are effective immunogen and possess a high efficiency of antigen transfer to lead to the induction of CEA-specific CTL response.

Adoptive transfer of splenocytes from CEA⁺/HS-Exoimmunized HLA-A2.1/K^b transgenic mice to SW480-bearing nude mice induced antitumor immunity more efficiently. To observe the antitumor effects of splenocytes from CEA⁺/HS-Exo-immunized HLA-A2/K^b transgenic mice, we further investigated whether those cells could delay or abolish the outgrowth of HLA-A2⁺CEA⁺ SW480 tumor cells in nude mice after the adoptive splenocyte transfer. As shown in Fig. 4A and B, the adoptive transfer of bulk splenocytes from 5 μg LS-HS-Exo- or LoVo-HS-Exo-immunized transgenic mice, combined with IL-2 administration, markedly inhibited the growth of SW480 tumor in nude mice and improved the survival rate of these mice compared with the splenocyte-transferred mice groups marked as “LS-Exo” and “LoVo-Exo.” No effect was observed following the injection of bulk splenocytes from transgenic mice immunized with A549-derived exosomes (heat stressed or not) and injection of PBS or IL-2.

These results indicate substantially that the HLA-A2.1/K^b transgenic mice can be primed in vivo with low dosage of CEA⁺/HS-Exo leading to the generation of CEA-specific HLA-A2.1/K^b-restricted CTL response and that the splenocytes from CEA⁺/HS-Exo-immunized HLA-A2/K^b transgenic mice may significantly mediate the inhibition of tumor outgrowth and survival prolongation in adoptive transfer nude mice model.

Human CTL induction by CEA⁺/HS-Exo-pulsed dendritic cells. Because CEA⁺/HS-Exo show more efficiency to induce CEA-specific CTL response in HLA-A2.1/K^b transgenic mice and these CTLs inhibit the growth of SW480 tumor in nude mice, we further investigated whether CEA⁺/HS-Exo-pulsed dendritic cells could induce CEA-specific CTL in HLA-A*0201⁺ healthy donors and especially in CEA self-tolerance cancer patients in vitro. We found that incubation of lymphocytes from HLA-A*0201⁺ healthy donors and HLA-A*0201⁺CEA⁺ patients with autologous dendritic cells pulsed with LS-HS-Exo or LoVo-HS-Exo did induce a HLA-A0201-restricted and CEA-specific CTL response but not with A549-HS-Exo. CEA-specific CTLs were generated in three of six HLA-A*0201⁺ healthy donors (Fig. 5A) and three of five HLA-A*0201⁺ CEA⁺ colon cancer patients (Fig. 5C) as judged by the lysis of SW480 cells but not LoVo cells. LS-174T-derived lysates (heat stressed or not) could not induce CEA-specific CTL production even at high dosages (data not shown).

Because CTLs are known to produce IFN-γ in an antigen-specific manner, which has been regarded as a reliable indicator for a Th1 response, we measured this cytokine in the culture medium of CTL from healthy donors (Fig. 5B) and cancer patients (Fig. 5D) after T2/CAP-1 restimulation. We found that the in vitro–generated CTL induced by CEA⁺/HS-Exo produced a significant level of IFN-γ after T2/CAP-1 stimulation, in contrast to that stimulated with T2/SSp-1 or native T2. We further examined the specificity of CTL by using peptide and MHC tetramer technology. Although CAP-1-tetramer⁺ lymphocytes are generally not detectable in peripheral blood mononuclear cells before restimulation, CAP-1-tetramer⁺ CD8⁺ T cells could be detected in CTL of both healthy donors and patients after four cycles of in vitro restimulation (Fig. 5E; Table 1).

These results showed that CEA⁺/HS-Exo were capable of eliciting CEA-specific CTL in a HLA-A*0201-restricted fashion in HLA-A*0201⁺ healthy donors and CEA⁺ cancer patients in vitro, suggesting that tolerance to CEA in cancer patients can be broken by CEA⁺/HS-Exo.

Although tumor immunotherapy is a promising treatment for the elimination of cancer cells, many technical challenges hinder its effective use clinically. For example, tumor antigens are often poor immunogen. Tumor cells frequently escape immune response by evolving into variants. CEA is a tumor-associated antigen with weak immunogenicity. However, CEA seems to be an attractive target for immunotherapy as it is overexpressed in ∼50% of all human cancers. Exosomes as potential therapeutic agents for cancers have attracted much attention from oncologists (10, 11). However, the efficacy of conventionally prepared exosomes in antitumor induction remains to be improved. In this investigation, we have used a new method to prepare a type of exosomes from heat-stressed CEA⁺ tumor cells (CEA⁺/HS-Exo) and analyzed their structure and function. We show that both CEA⁺/HS-Exo and CEA⁺/Exo contain CEA, HSP70, Hsc70, and MHC-I, but the former seem to have higher amount of HSP70 and MHC-I molecules than the latter. Correlating with this discovery, CEA⁺/HS-Exo also exhibit stronger immunogenicity than CEA⁺/Exo both in vitro and in vivo.

HSPs are chaperones, which assist in the folding and the translocation of new synthesized proteins. Both inducible HSP70 and constitutive Hsc70 are HSP70 family members. HSP70 has immunotherapeutic potential as an adjuvant, HSP70-chaperoned antigenic peptides can be taken by antigen-presenting cells through a receptor-mediated manner to not only help the antigenic peptide presentation on MHC-I but also promote the function of antigen-presenting cells, especially dendritic cells, to generate antigen peptide-specific CD8⁺ T cells (22, 37–39). Besides, HSP70 can facilitate the loading of MHC-I molecules with peptides (38, 40), and Hsc70 can facilitate MHC-II antigen processing and presentation (40, 41). Therefore, enriched HSPs in exosomes may play important roles in antigen processing and presentation. This postulation might be true because we found that the immunostimulatory capacity of CEA⁺/HS-Exo to dendritic cells was more potent than non-heat-stressed exosomes (CEA⁺/Exo).

We have further assessed the antitumor immunity elicited by exosomes derived from the heat-stressed human tumor cells. On immunization with CEA⁺/HS-Exo (LS-HS-Exo or LoVo-HS-Exo), HLA-A2.1/K^b transgenic mice developed a strong HLA-A2.1/K^b-restricted CTL response specific for CAP-1. It is notable that the adoptive transfer of CEA-specific CTL from HLA-A2.1/K^b transgenic mice immunized by CEA⁺/HS-Exo results in effective inhibition of tumor growth and prolonged survival among tumor-bearing nude mice. In addition, CEA⁺/HS-Exo-pulsed autologous dendritic cells elicit a potent HLA-A*0201-restricted and CEA-specific CTL activity in lymphocytes from HLA-A*0201⁺ healthy donors or CEA⁺ patients. The results suggest that tolerance to CEA in cancer patients may be possibly overcome by CEA⁺/HS-Exo immunization.

There are so many reports showing that tumor-derived exosomes are immunogenic (10, 11, 13, 29, 35); however, there are other reports indicating that tumor-derived exosomes contain immunosuppressive molecules, such as latent membrane protein-1, NKG2D ligand, Fas ligand, and HLA-G, and have been proposed to be immunosuppressive, representing one of mechanisms of tumor evasion from the immune surveillance (42–47). This discrepancy may be due to differences in the types of exosomes produced. Therefore, identification of in-depth proteomic contents of different tumor-derived exosomes or incorporating immunologic molecules into exosomes by some strategies is a crucial aspect for aiding the understanding of their biological functions and especially targeting for use in future therapy of cancer.

In conclusion, the work described here, for the first time, has created a new way to prepare HSP70/CEA/MHC-I–enriched exosomes, and the enhanced expression of HSP70 and MHC-I in exosomes may be beneficial to promote immunogenicity of this tumor-derived membrane vesicles, especially without genetic modification or incorporation of potentially toxic adjuvants. HSP70/CEA/MHC-I–enriched exosomes released by heat-stressed tumor cells may indirectly explain the enhanced antitumor immunity by hyperthermia. Most importantly, HSP70/CEA/MHC-I–enriched exosomes are easy to obtain from CEA⁺ tumor cell lines and do not have limitations requiring surgical tissues. Therefore, we provide a novel strategy to improve the efficacy of exosome-based tumor vaccines, and this approach is simple and applicable for the future clinical trails of CEA⁺ cancer patients.

We thank Dr. Guisheng Li for obtaining the blood samples of cancer patients, Prof. You-Wen He (Duke University) for critical reading of this article, and Jun Li, Linhong Sun, and Xiaohui Huang.