Chicken ovalbumin (OVA) was obtained from SIGMA-Aldrich Canada Ltd, Oakville, Ontario, Canada. Phycoerythrin (PE)-labeled anti-IFN-γ antibody (Ab) was obtained from BD Biosciences (Mississauga, ON, Canada). Biotin-labeled anti-T-bet, anti-phosphor-EIF4E (S209), anti-CD44, anti-PD-1, anti-PD-L1 and anti-LAG3 Abs were obtained from eBioscience (San Diego, CA). PE-labeled H-2K^b/OVA_257-264 tetramer (PE-tetramer) and Fluorescein isothiocyanate (FITC)-labeled anti-CD8 Ab were obtained from Beckman Coulter (Miami, FL). Rabbit anti-CD9 and anti-LAMP-1 Abs were obtained from BD Bioscience. The H-2K^b-restricted OVAI (OVA_257-264, SIINFEKL) peptide and the control Lewis lung carcinoma H-2K^b-restricted MutI (FEQNTAQP) peptide were obtained from Multiple Peptide Systems (San Diego, CA). All cytokines were obtained from Peprotech (Peprotech, Rocky Hill, NJ). Recombinant adenoviral vectors, including transgene OVA-expressing AdV_OVA, transgene Gag-expressing AdV_Gag and the no transgene-expressing AdV_Null were available in our laboratory [18] [32] . Adenoviral vector AdV_IL-21 expressing mouse IL-21 was constructed by the insertion of IL-21 open reading frame of pUNO1-mIL-21 vector (InVivoGen, San Diego, CA) into the pShuttle-CMV vector (Stratagene, La Jolla, CA) by recombinant technology as we previously described [18] [32] . The Pmel-digested pShuttle-CMV-mIL-21 was transformed into BJ5183 Escherichia coli cells containing pAdEasy-1 backbone vector for the homologous recombination [18] [32] . The recombinant AdV vector was linealized by PacI digestion and transfected into HEK-293 cells using Lipofectamine 2000 (Invitrogene, Carlsbad, CA) to generate AdVIL-21 expressing transgene IL-21 [18] [32] . AdV_IL-21 was then amplified in HEK-293 cells and purified by a series of cesium chloride ultracentrifugation [18] [32] . The highly lung metastatic, OVA- and Gag-expressing BL6-10_OVA and BL6-10Gag mouse B16 melanoma cell lines were available in our laboratory [31] [32] [33] . IL-21 enzyme-linked immunosorbent assay (ELISA) kit was obtained from R & D Systems Inc, Minneapolis, MN. Female 6- to 8-week-old C57BL/6 (B6) (18 - 20 g) and OVA-specific T cell receptor (TCR) transgenic OTI mice were obtained from the Jackson Laboratory (Bar Harbor, MA). All experiments were performed according to protocols and guidelines approved by the Animal Research Ethics Board, University of Saskatchewan (Protocol No. 20130020).

Bone marrow-derived dendritic cells (DCs) were obtained by culturing bone marrow cells of wild-type (WT) B6 mice in culture medium containing granulocyte monocyte colony-stimulating factor (GM-CSF) (20 ng/ml) and IL-4 (20 ng/ml) for six days, as previously described [29] . DCs were pulsed with OVA (0.5 mg/ml) for overnight and termed DC_OVA. DC_OVA stained with various Abs were analysed by flow cytometry. DC_OVA-released exosomes (EXO_OVA) were then purified from DC_OVA culture supernatants by differential ultracentrifugation [29] . In addition, DCs were also infected with AdV_Gag (100 pfu/cell) for overnight in a humidified 37˚C, 5% CO₂ incubator, and termed DC_Gag. DC_Gag-released EXOs were purified from DC_Gag culture supernatants by ultracentrifugation and termed EXO_Gag.

Polyclonal naïve CD8⁺ T cells were isolated from WT B6 mouse spleens, enriched by passage through nylon wool columns, and then purified by negative selection using anti-mouse CD4 paramagnetic beads (Life Technologies, Waltham, MA), as previously described [34] [35] . Purified polyclonal CD8⁺ T cells were cultured for 3 days in RPMI 1640 medium, containing IL-2 (20 U/ml) and ConA (1 μg/ml). CD8⁺ T cells were then purified from ConA-activated T cells using MACS anti-CD8 microbeads (Miltenyi Biotech, Auburn, CA) to yield CD8⁺ T cell populations with >95% purity [29] , and termed ConA-T cells.

The OVA-T_EXO and Gag-T_EXO vaccines were generated by the incubation of CD8⁺ ConA-T cells with EXO_OVA or EXO_Gag at 3 × 10⁶ cells/10 µg for 1 hour at 37˚C, followed by the transfection of T cells with the uptake of EXO_OVA or EXO_Gag with AdV_41BBL at 100 pfu/cell for another 2 hours to form the vaccines, as previously described [32] . To prepare the OVA-T_EXO/IL-21 or Gag-T_EXO/IL-21 and the control OVA-T_EXO/Null or Gag-T_EXO/Null vaccines, the above OVA-T_EXO or Gag-T_EXO cells were further transfected with AdV_IL-21 or AdV_Null [100 pfu/cell] for 2 hours to form OVA-T_EXO/IL-21 or Gag-T_EXO/IL-21 and the control OVA-T_EXO/Null or Gag-T_EXO/Null vaccine respectively, as we previously described [32] .

To develop a chronic infection model, B6 mice (4/group) were intravenously (i.v.) injected with anti-CD4 Ab (300 µg/mouse) to deplete CD4⁺ T cells, followed by i.v. infection of mice with AdV_OVA (1 × 10⁶ pfu/mouse) one day after the anti-CD4 Ab treatment performed for CD4⁺ T cell depletion. This experimental procedure was aimed to develop a more stringent mouse model of chronic infection, since AdV_OVA infection of mice was performed under the condition of CD4⁺ T cell depletion such that the ‘helpless’ CD8⁺ T cells activated by AdV_OVA infection became functionally exhausted with severe defects in IFN-γ production and cytotoxicity [36] . These mice were chronically AdV_OVA-infected B6 (AdV_OVA-B6) mice and termed AdV_OVA-B6 mice with chronic infection. To develop an acute infection model, B6 mice (4/group) were i.v. infected with recombinant OVA-expressing rLmOVA bacteria [2000 colony forming unit (cfu)/mouse]. These mice were termed rLmOVA-B6 mice. After infections, OVA-specific CD8⁺ CTL responses in AdV_OVA-B6 and rLmOVA-B6 mice were kinetically analyzed by flow cytometry.

Peripheral blood samples derived from AdV_OVA-B6 and rLmOVA-B6 mice were double stained with FITC-CD8 Ab and PE-tetramer and analyzed by flow cytometry to assess CD8⁺ T cell responses at different days after infection. To increase the amount of OVA-specific CD8⁺ memory T (Tm) cells formed 30 days or over 30 days post infection, such that phenotypes of OVA-specific CD8⁺ Tm cells can be more accurately analyzed, B6 mice were infected with an increased amount of AdV_OVA (1.2 - 1.4 × 10⁶ pfu/mouse) for the development of chronic infection AdV_OVA-B6 mice, while B6 mice with prior i.v. injection of naïve CD8⁺ T cells (1 × 10³/mouse) purified from OTI mouse spelnocytes were infected with rLmOVA bacteria for the development of acute infection rLmOVA-B6 mice [18] . To assess the phenotype of OVA-specific CD8⁺ Tm cells, peripheral blood samples derived from AdV_OVA-B6 and rLmOVA-B6 mice 60 days post the primary infection were triply stained with FITC-CD8 Ab, PE-tetramer and PE-Cy5-Abs for various immune molecules, and then analyzed by flow cytometry. To assess the conversion of CTL exhaustion, AdV_OVA-B6 mice with chronic infection were i.v. immunized with OVA-T_EXO/IL-21 or the control OVA-T_EXO/Null vaccine (1 × 10⁶ cells/mouse), followed by the analysis of OVA-specific CD8⁺ T cell proliferation and phenotypes 4 days post immunization. To stain the intracellular molecules, mouse splenocytes derived from AdV_OVA-B6 and rLmOVA-B6 mice were first incubated with FcR blocking anti-16/32 Ab (eBioscience) for 30 min on ice for eliminating any nonspecific staining. The cells were then stained with FITC-CD8 Ab and PE-tetramer, followed by the fixation and permeabilization with Cytofix/Cytoperm (BD Biosciences) according to the manufacturer’s instruction. The cells were further stained with PE-Cy5-antibodies specific for various molecules such as EIF4E and T-bet, and the expression of these molecules was assessed by flow cytometry. For intracellular staining of IFN-γ, mouse splenocytes were first incubated in culture medium containing OVAI peptide (2 µg/ml) and Golgi-stop (0.7 µg/ml) (BD Biosciences) at 37˚C for 5 hrs, followed by the incubation with FcR blocking anti-16/32 Ab for 30 min on ice. The cells were then stained with FITC-CD8 Ab and PE-tetramer, followed by fixation and permeabilization with Cytofix/Cytoperm (BD Biosciences). Intracellular staining of IFN-γ was conducted using PE-Cy5-anti-IFN-γ Ab, and the expression of IFN-γ was assessed by flow cytometry. Data were acquired by CytoFlex (Beckman Coulter) and analyzed with FlowJo software (TreeStar, San Diego, CA).

EXO (10 µg/well) were resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), and subsequently transferred onto the nitrocellulose membrane (Bio-Rad Laboratories, Hercules, CA). The membrane was blocked with 5% bovine serum albumin (BSA) in PBS, immunoblotted with rabbit anti-CD9 or anti-LAMP-1 Ab, followed by the incubation with anti-rabbit IRDyeR800CW Ab, and then, scanning with the ODYSSEY imager according to manufacturer’s instruction (Li-COR Bioscience, Lincoln, NE).

EXO were fixed in 4% paraformaldehyde. The pellets were loaded onto carbon-coated formvar grids. The exosome sample-loaded grids were stained with saturated aqueous uranyl and then, examined with a JEOL 1200EX electron microscope at 60 kV.

To examine the antitumor immunity conferred by the OVA-T_EXO/IL-21 vaccine, WT B6 mice (6/group) were i.v. immunized with OVA-T_EXO/IL-21 cells or the control OVA-T_EXO/Null (1 × 10⁶ cells/mouse). Mice were i.v. injected with 0.5 × 10⁶ BL6-10_OVA cells six days post immunization. To assess therapeutic immunity of the Gag-T_EXO/IL-21 vaccine, chronic infection AdV_OVA-B6 mice (n = 6) with CTL inflation and functional exhaustion were first injected i.v. with 0.5 × 10⁶ BL6-10_Gag cells. Three days after tumor cell inoculation, AdVova-B6 mice were i.v. immunized with Gag-T_EXO/IL-21 or the control Gag-T_EXO/Null cells (1 × 10⁶ cells/mouse). The mice were sacrificed 3 weeks after tumor cell injection, and the lung metastatic B16 melanoma colonies were counted in a blind fashion. Metastatic B16 melanoma colonies on freshly isolated lungs appeared as black color foci that were easily distinguishable from normal lung tissues. Metastasis was also confirmed by histological examination.

Statistical analyses were performed with the Mann-Whitney U test using Prism software (GraphPad Software, San Diago, CA) to compare variables of different groups in animal studies or with the Student t test to compare variables of different groups in flow cytometric analysis [32] . Unless stated otherwise, data are expressed as mean (with SD). A value of p < 0.05 was considered statistically significant.

We constructed a replication-deficient transgene IL-21-expressing recombinant adenovirus AdV_IL-21 by recombinant DNA technology as we described in MATERIALS & METHODS (Figure 1(a)) [18] [32] . B6 mouse dendritic cells were generated by culturing B6 mouse bone marrow cells in the culture medium containing GM-CSF and IL-4 for six days, pulsed with OVA for overnight and termed DC_OVA [18] [32] . OVA-specific exosomes (EXO_OVA) were purified from DC_OVA culture supernatants by ultracentrifugation, and analyzed by electron microscopy and Western blotting analyses, respectively. We demonstrated that EXO_OVA had exosomal “saucer” or round shape with 50 - 90 nm in diameter (Figure 1(b)) and contained EXO-associated proteins such as CD9 and LAMP-1 (Figure 1(c)) [37] . We infected the OVA-T_EXO cells [32] with AdV_IL-21 to generate a transgene IL-21-engineered T cell-based vaccine, OVA-T_EXO/IL-21. IL-21 secretion in OVA-T_EXO/IL-21 culture supernatants was estimated to be ~200 pg/ml using IL-21 ELISA kit. In contrast, there was no IL-21 detected in OVA-T_EXO culture supernatants.

To examine the immunogenicity, we i.v. immunized B6 mice with OVA-T_EXO/IL-21 or the control OVA-T_EXO/Null vaccine. We demonstrated that both OVA-T_EXO/IL-21

and control OVA-T_EXO/Null vaccines stimulated comparable OVA-specific CTL responses (Figure 2(a)). To assess, whether these CTLs are of functional effect, we challenged immunized mice with OVA-expressing BL6-10_OVA tumor cells and examined its protective immunity against lung BL6-10_OVA tumor metastases. We showed that OVA-T_EXO/IL-21 or the control OVA-T_EXO/Null vaccine induced complete protection from lung tumor metastases in all 6/6 mice (Figure 2(b) & Figure 2(c)), indicating that OVA-T_EXO/IL-21 vaccine stimulates comparable CTL responses compared to the control OVA-T_EXO vaccine in wild-type B6 mice.

We previously found that infection of B6 mice with AdV_OVA, led to establishment of an OVA-specific chronic infection mouse model with OVA-specific CTL inflation and exhaustion [18] . To develop acute and chronic infection models, B6 mice were i.v. infected with OVA-expressing recombinant rLmOVA bacteria and AdV_OVA, respectively, followed by the kinetic analysis of OVA-specific CTL responses [18] . We demonstrated that AdV_OVA infection resulted in OVA-specific memory CD8⁺ T cell inflation, when compared to CTLs developed in rLmOVA-immunized mice (Figure 3(a)), suggesting that rLmOVA induces an acute infection, while AdV_OVA induces a chronic infection with CD8⁺ CTL inflation [18] [19] . To confirm it, the phenotypes of OVA-specific memory CTLs were analyzed by flow cytometry. We demonstrated that memory CTLs expressed cell

surface T-cell memory markers IL-7R and CD44 as well as IL-21R sixty days after infection in both AdV_OVA-B6 and rLmOVA-B6 mice (Figure 3(b)). Importantly, we also found that these CTLs developed in AdV_OVA-immunized mice up-regulated inhibitory molecules, such as PD-1, PD-L1 and LAG-3, when compared to CTLs developed in rLmOVA-immunized mice (Figure 3(b)), indicating that these CTLs may be exhausted. To further confirm it, we assessed the functional effect (expression of the effector cytokine IFN-γ) of these CTLs by flow cytometry. We showed that ~72% OVA-specific CTLs were IFN-γ positive in rLmOVA-B6 mice, compared to only ~10% OVA-specific CTLs were IFN-γ positive in rLmOVA-B6 mice (Figure 3(c)), thus confirming that those OVA-specific CTLs in AdV_OVA-B6 mice are exhausted CTLs [18] [19] .

We examined in our next set of experiments, whether the OVA-T_EXO/IL-21 vaccine converts CTL exhaustion in chronic infection. The AdV_OVA-B6 mice were boosted with OVA-T_EXO/IL-21 and the control OVA-T_EXO/Null vaccine. Four days after the boost, cell proliferation and intracellular IFN-γ expression of OVA-specific CTLs were assessed by flow cytometry. We demonstrated that there were ~6-fold (5.20% vs 0.81%) of CTL increase in OVA-T_EXO/IL-21-boosted mouse peripheral blood, which is more than ~3-fold of CTL increase in the control OVA-T_EXO/Null-boosted mouse peripheral blood at day 4 after the boost (Figure 4(a)), indicating that OVA-T_EXO/IL-21 vaccine can more potently convert CTL exhaustion by significantly stimulating the proliferation of previously exhausted CTLs. We next assessed expression of an effector T cell cytokine, IFN-γ, in exhausted CTLs on a “per-cell” basis by intracellular staining of T cell IFN-γ. We demonstrated that only ~20% of OVA-specific CTLs were IFN-γ positive in AdV_OVA-B6 mice (Figure 4(b)). In OVA-T_EXO/Null- and OVA-T_EXO/IL-21-boosted AdV_OVA-B6 mice, however, we found ~60% and ~75% of IFN-γ-producing CTLs in the total OVA-specific CTLs, respectively (Figure 4(b)), confirming that OVA-T_EXO/IL-21 vaccine more potently converts CTL exhaustion in chronic infection by not only increasing the number of CTLs, but also restoring CTL functional effect.

We previously showed that the OVA-T_EXO vaccine rescued exhausted CTLs in chronic infection via its CD40L signaling, inducing the activation of the mTORC1 pathway [18] . Four days after the boost, cell phenotypes of the OVA-specific CTLs were assessed by flow cytometry to examine, whether the OVA-T_EXO/IL-21 vaccine activates the mTORC1 pathway. We analyzed CTLs for the phosphorylation status of mTORC1-regulated EIF4E (pEIF4E) and for the intracellular expression of the transcription factor T-bet by flow cytometry. We determined that OVA-specific CTLs up-regulated levels of phospho-EIF4E and T-bet in OVA-T_EXO/Null-boosted AdV_OVA-B6 mice (Figure 4(a)), which was consistent with our previous report [18] . Interestingly, the intracellular expression of pEIF4E and T-bet in OVA-specific CTLs was significantly higher in the OVA-T_EXO/IL-21-boosted AdV_OVA-B6 mice than that in the OVA-T_EXO/Null-boosted AdV_OVA-B6 mice (pEIF4E, p = 0.02; T-bet, p = 0.001) (Figure 4(a)), confirming

that OVA-T_EXO/IL-21 vaccine converts CTL exhaustion mostly through the activation of the mTORC1 pathway by both CD40L and IL-21 signaling.

To assess a potential therapeutic immunity of the Gag-T_EXO/IL-21 vaccine in chronic infection, AdV_OVA-B6 mice with chronic infection of AdV_OVA were first i.v. injected with Gag-expressing BL6-10_Gag melanoma cells, and four days post B16 melanoma cell challenge, mice were i.v. immunized with the Gag-T_EXO/IL-21 or the control Gag-T_EXO/Null vaccine (Figure 5(a)). Seventeen days later, mouse lung tissues were collected for histopathological examination (Figure 5(a)). Importantly, the Gag-T_EXO/IL-21 vaccine demonstrated a complete eradication of established BL6-10_Gag lung metastases in 5/6 AdV_OVA-B6 mice, thus stimulating more efficient therapeutic immunity against Gag-expressing BL6-10_Gag melanoma than the control Gag-T_EXO/Null vaccination (Figure 5(b)). Our data indicate that Gag-T_EXO/IL-21 vaccine is capable of inducing potent therapeutic immunity against established Gag-expressing tumors in the presence of chronic infection.