Alarmin-painted exosomes elicit persistent antitumor immunity in large established tumors in mice

Treating large established tumors is challenging for dendritic cell (DC)-based immunotherapy. DC activation with tumor cell-derived exosomes (TEXs) carrying multiple tumor-associated antigen can enhance tumor recognition. Adding a potent adjuvant, high mobility group nucleosome-binding protein 1 (HMGN1), boosts DCs’ ability to activate T cells and improves vaccine efficiency. Here, we demonstrate that TEXs painted with the functional domain of HMGN1 (TEX-N1ND) via an exosomal anchor peptide potentiates DC immunogenicity. TEX-N1ND pulsed DCs (DCTEX-N1ND) elicit long-lasting antitumor immunity and tumor suppression in different syngeneic mouse models with large tumor burdens, most notably large, poorly immunogenic orthotopic hepatocellular carcinoma (HCC). DCTEX-N1ND show increased homing to lymphoid tissues and contribute to augmented memory T cells. Importantly, N1ND-painted serum exosomes from cancer patients also promote DC activation. Our study demonstrates the potency of TEX-N1ND to strengthen DC immunogenicity and to suppress large established tumors, and thus provides an avenue to improve DC-based immunotherapy. The use of tumour exosome-activated dendritic cell (DC) immunotherapy shows promise for the treatment of large established tumours. Here, the authors generate alarmin HMGN1-attached tumour exosomes which significantly improve therapy efficacy by boosting DC activation in several preclinical mouse models.

D endritic cell (DC)-based cancer immunotherapy relies on activation of DCs as antigen-presenting cells by tumor antigens or lysates to stimulate an adaptive immunological response against tumors. This has shown promise for cancer patients with sipuleucel-T, which was approved for advanced prostate cancer 1,2 , as a typical example. However, DC vaccines currently being investigated in trials are somewhat underwhelming with the overall objective response rate rarely exceeding 15% 3 . However, DC vaccines have good safety profiles and the potential to induce innate and adoptive antitumor immunity for long-term efficacy, thus one of the emerging trends in DC vaccine research is to improve the immunogenicity of next-generation DC vaccines 4 according to registered DC clinical trials. Inducers that enhance the immunogenicity of exogenous DCs through ex vivo activation of these cells prior to immunization are under intensive scrutiny 5 , and the implementation of next-generation DC vaccines with improved activities might hold the key to fulfilling the therapeutic potential of DC-based immunotherapy.
Maturation cocktails and other immunoadjuvants can be used to promote the immunostimulatory activity of exogenously activated DCs [6][7][8] . The nature of immunoadjuvants determines the type, the magnitude, the breadth, and the quality of antitumor immune response 9 . DC vaccines that induce a bias towards a Th1 immune response appear to elicit stronger activation of tumorspecific cytotoxic T lymphocytes and natural killer (NK) cells and only tumor-associated antigen (TAA)-specific Th1 immune responses are protective against tumors 10,11 , thus adjuvants that promote this bias are preferred. High-mobility group nucleosome-binding protein 1 (HMGN1) improves DC maturation and activation in response to exogenous antigens and consistently induces Th1 immune responses 12 . Hence HMGN1 has been harnessed to promote generation of antitumor immune responses as DNA vaccines or in combination with chemotherapeutic agents and activators/checkpoint inhibitors in different tumor models [13][14][15] . However, only limited effects were achieved with DNA vaccines and combinatorial therapy since HMGN1 required intratumoral administration, which is not clinically applicable to most tumors. Therefore, efficient systemic delivery of HMGN1 to cells is needed to unlock the full therapeutic potential of HMGN1 as an immunoadjuvant.
Recently, tumor-derived exosomes (TEXs), which harbor multiple TAAs to provide a broad spectrum of antigen presentation, have drawn much interest as antigens for DC-based immunotherapy [16][17][18] . TEXs, particularly patient-derived TEXs, have good biocompatibility, low immunogenicity, and high penetration of biological barriers, thus TEXs are good activators of DC vaccines. It was previously shown that covalent linkage of an alarmin to a target TAA could better promote TAA-specific immune responses and immunoprotection 13,14 . Ideally, a method that combines delivery of molecular adjuvants with the TAAenriched exosomes would improve the immunogenicity of DC immunotherapy and consequently, potential therapeutic outcome.
Here, we covalently link the functional N-terminal domain of HMGN1 (N1ND) to CP05, a recently identified exosomal anchor peptide which allows direct and efficient cargo loading on exosomes 12,19 , and paint N1ND on TEXs (TEX-N1ND). DCs pulsed with TEX-N1ND (DC TEX-N1ND ) elicit durable tumor retardation in multiple syngeneic subcutaneous tumor models bearing large, established tumor burdens, and also in poorly immunogenic orthotopic hepatocellular carcinoma (HCC) mice with large tumor loads. Importantly, TEX-N1ND also augments human DC immunogenicity against different patient-derived cancer cells in vitro. Altogether TEX-N1ND heightens the immunostimulatory activities of DCs and migration to lymph nodes, consequently accelerating the generation of long-lived protective T cell memory response, which is responsible for the persistent antitumor immunity of activated DCs.

Results
N1ND-painted TEX promotes N1ND uptake and DCs' immunity. Motivated by HMGN1's efficacy at inducing Th1 antitumor immune response and our recently discovered strategy for intracellular delivery of cargoes 12,13,19 , we explored the feasibility of painting the surface of TEX with the functional Nterminus of HMGN1 (N1ND) using CD63-binding peptide (CP05) (TEX-N1ND). Incubation of DCs with TEX-N1ND can deliver N1ND into cells to enhance TEX-mediated DC activation and immunogenicity in different tumor models. To determine whether preexisting HMGN1 levels in TEXs would have any impact on the effect of additional N1ND on DC activation, we assayed TEXs, with a buoyant density of 1.13-1.23 g ml −1 (Supplementary Fig. 1a), from murine pancreatic cancer (Panc02), Lewis lung cancer (LLC1), lymphoma (EL4), breast cancer (4T1), HCC (Hepa1-6) cells, and DCs for HMGN1 (Fig. 1a). The levels of HMGN1 in TEXs generally correlated with their parental cells ( Supplementary Fig. 1b). TEXs from Hepa1-6, 4T1, and Panc02 were selected as representative TEXs containing no, medium, or high levels of HMGN1 respectively for subsequent studies. AF680-labeled N1ND-CP05 efficiently bound to TEXs, with the highest binding efficiency to TEX Hepa (88.1%) (Fig. 1b), correlating with the percentage of CD63-positive TEXs (Supplementary Fig. 1c) without altering the morphology of TEXs ( Supplementary Fig. 1d). In contrast, only trace binding was detected in the mixture of TEX Hepa with N1ND without CP05 (Fig. 1b, TEX Hepa /N1ND), demonstrating that N1ND binds TEXs due to CP05. Dramatically increased uptake of AF680-labeled N1ND-CP05 was observed in bone marrow-derived DCs (BMDC) after incubation with TEX (TEX-N1ND) (40 μg ml −1 ) (BMDC TEX-N1ND ) when compared with the mixture of TEXs with N1ND without CP05 (BMDC TEX/ N1ND ) or N1ND-CP05 alone (BMDC N1ND ) ( Fig. 1c and Supplementary Fig. 1e), suggesting that binding to TEXs enabled efficient delivery of N1ND-CP05 to cells. Major histocompatibility complex class (MHC) I/II, costimulatory molecules (Fig. 1d, e) and tumor necrosis factor-α (TNF-α) (Fig. 1f), a proinflammatory cytokine induced by HMGN1, were significantly upregulated in DCs treated with TEX-N1ND compared with the controls. Importantly, TEX-N1ND induced comparable DC activation in different TEXs independent of endogenous HMGN1 levels. Secreted levels of interferon-γ (IFN-γ) and interleukin-2 (IL-2) consistently increased in splenic naive T cells stimulated with BMDC TEX-N1ND (Fig. 1g), resulting in significantly higher cytolysis rates against parental tumor cells, from which TEXs were derived, compared with other controls (Fig. 1h), indicating that N1ND is capable of activating DCs in vitro and increasing the immunogenicity of the DCs.
N1ND enhances TEX-induced DC immunity on large tumors. Treating large established tumors, which closely mimic the typical clinical presentation of patients, remains a challenge for initial DC-based immunotherapy, thus we evaluated BMDC TEX-N1ND in three different subcutaneous syngeneic tumor models: Hepa1-6 HCC, Panc02 pancreatic cancer, and 4T1 breast cancer. Tumor cells (2 × 10 6 ) were injected subcutaneously and allowed to grow for 14 days to~40-60 mm 2 depending on tumor type before initiation of therapy. BMDC TEX-N1ND , BMDC TEX/N1ND , or BMDC TEX (2 × 10 6 ) were administered intravenously once per week for 3 weeks, a protocol identical to our previous study 18 . Robust and persistent tumor growth inhibition with very similar residual and significantly smaller tumor sizes than other groups ( Fig. 2a) was achieved in HCC tumor-bearing mice treated with BMDC TEX-N1ND 42 days after tumor challenge; whereas BMDC TEX/N1ND or BMDC TEX elicited weaker responses with higher variability. Increased CD8 + and reduced CD4 + CD25 + regulatory T cells (Tregs) were consistently detected in blood and tumor tissues from mice treated with BMDC TEX-N1ND compared with other groups (Fig. 2b, c and Supplementary Fig. 2a, b). In contrast, no significant change was found for B cells across different treatment groups (Fig. 2b, c). Similarly, durable tumor retardation was observed for Panc02 and 4T1 subcutaneous tumor-bearing mice treated with BMDC TEX-N1ND compared with other groups on day 42 (Fig. 2d, e) with similar shifts in circulatory and tumor CD8 + T cells and Tregs (Fig. 2f, g), suggesting that the strategy is broadly applicable across tumor types.     Notably, consistent with the previous study 18 , a cross-protective effect was observed in mice bearing subcutaneous HCC treated with DC TEX(Panc)-N1ND ( Supplementary Fig. 2c, d), and the presence of N1ND did not influence the antigen-specific immune response elicited by TEXs as demonstrated by significant increases in alpha-fetoprotein (AFP)-specific CD8 + T cells ( Supplementary Fig. 2e) and in levels of IFN-γ and IL-2 upon stimulation with AFP212 ( Supplementary Fig. 2f), an identified murine AFP epitope 20 , in culture supernatants of splenic T lymphocytes from mice bearing subcutaneous HCC treated with DC TEX(Hepa)-N1ND but not with DC TEX(Panc) -or DC TEX(4T1)-N1ND . Importantly, a hierarchy of efficacy was observed with different BMDC TEX-N1ND in three tumor models and was inversely correlated with the endogenous expression of HMGN1 in TEXs, with TEXs derived from HCC cells showing the greatest efficacy, implying that DC TEX-N1ND is likely more effective for tumors without or with lower expression of HMGN1.
DC TEX-N1ND induce persistent retardation of orthotopic HCC. To understand the cellular and molecular mechanisms underlying the observed therapeutic effect of DC TEX-N1ND , we analyzed responses to DC TEX-N1ND in an immunosuppressive, poorly immunogenic Hepa1-6 HCC model. Subcutaneous tumors do not faithfully recapitulate the histology and complexity of HCC, thus we employed two different orthotopic HCC mouse models bearing large tumor loads. A day-7 HCC model was generated by transplanting more tumor tissues to accelerate the formation of large tumors (0.68 ± 0.02 in longitudinal diameter) within 7 days, and another with tumor loads (0.55 ± 0.04 in longitudinal diameter) formed in 21 days (day-21) by transplanting fewer tumor tissues to allow the formation of immunosuppressive microenvironment ( Supplementary Fig. 3a). Tumors grew progressively in day-7 orthotopic HCC mice resulting in 100% lethality in untreated mice by day 35, replacing most of the livers (Supplementary Fig. 3b-e). Three weekly intravenous doses of DC TEX-N1ND (2 × 10 6 ) significantly inhibited tumor growth and extended the lifespan of tumor-bearing mice as tumor volume and mass were significantly reduced and survival rate increased compared with other groups (Supplementary Fig. 3c-e). Persistent tumor retardation was observed in mice treated with DC TEX-N1ND compared with other groups on day 32 and 46 ( Supplementary Fig. 3c, d), demonstrating that N1ND-CP05 enhanced activated DCs' antitumor immunity. Concordantly, in the day-21 orthotopic HCC mice, DC TEX-N1ND (2 × 10 6 ) promoted robust and sustained inhibition of induced tumors compared with mice treated with DC TEX , with a 60% survival rate at week 9 vs. 0% for DC TEX (Fig. 3a-d). Real-time monitoring of tumor growth via magnetic resonance imaging (MRI) revealed progressive tumor growth and multiple nodular foci formation in DC TEX -treated and untreated mice, whereas no evident tumor progression was found in mice treated with DC TEX-N1ND (Fig. 3c, d), demonstrating that DC TEX-N1ND results in sustainable tumor inhibition. Strikingly, no pulmonary metastasis was found in tumor-bearing mice treated with DC TEX-N1ND at week 9; in contrast, prominent pulmonary metastasis was established in DC TEX -treated mice (Fig. 3e). These findings demonstrate that TEX-N1ND activates DCs to induce greater antitumor immunity.
DC TEX-N1ND reshape orthotopic HCC microenvironment in mice. The tumor microenvironment plays a crucial role in tumor progression and prognosis 21 , therefore modulation of the tumor microenvironment may augment the antitumor immune response. To determine the effect of DC TEX-N1ND on the tumor microenvironment, we examined CD8 + T cells in blood and tumor tissues from day-21 orthotopic HCC mice treated with DC TEX-N1ND as DC TEX and HMGN1 were shown to function primarily via stimulation of CD8 + T cells and increased IFN-γ secretion 13 . Significant increases in CD8 + T cells and declines in CD4 + CD25 + Tregs were detected in blood and tumors from mice treated with DC TEX-N1ND compared with other groups (Fig. 4a, b), indicating that DC TEX-N1ND primarily promotes CD8 + T cell priming. Active recruitment of effector CD3 + T lymphocytes to primary tumor sites in orthotopic HCC mice was observed in tumor cross-sections from DC TEX-N1ND -treated mice compared with DC TEX at different time-points, and fewer effector T cells were detected in PBS-treated mice on week 7 (Fig. 4c). Sporadic and significantly fewer FoxP3 + Tregs were found in DC TEX-N1NDtreated samples compared with other groups (Fig. 4c, d). Level of IFN-γ significantly rose and TGF-β and IL-10 levels were significantly reduced in tumor tissues and blood from DC TEX-N1NDtreated mice, compared with DC TEX and PBS groups (Fig. 4e, f, and Supplementary Fig. 4a, b). Overall, these data demonstrate that DC TEX-N1ND can induce persistent antitumor immunity and reshape tumor microenvironment in orthotopic HCC mice bearing large established tumor burdens.

TEX-N1ND boosts memory T cells for long-lasting immunity.
Generation of memory CD8 + T cells can greatly enhance protective immunity 22 and thus is an important goal of vaccination. Given the critical role of CD8 + T cells in the therapeutic effect, we first analyzed the effect of DC TEX-N1ND on circulatory memory T cells from treated mice. Strikingly, significantly greater amplification of effector and memory (CD8 + CD44 high ) T cells, particularly CD8 + CD44 high CD127 high long-lived memory T cells [23][24][25] , occurred in mice receiving DC TEX-N1ND treatment on Total protein (20 μg) was loaded for TEXs. Exosomes from Hela (EXO Hela ) and DCs (EXO DC ) were used as positive and negative controls, respectively. CD63 was used as an exosomal marker protein. This experiment was repeated once (two in total). b Quantitative analysis on the binding efficiency of N1ND-CP05 to TEXs. TEX Hepa /N1ND or TEX Hepa -N1ND represents the mixture of N1ND and TEXs from Hepa1-6 cells (TEX Hepa ) or the complex of N1ND-CP05 with TEX Hepa , respectively. AF680-labeled N1ND-CP05 or N1ND peptide (80 μg) was incubated with PKH67-labeled TEXs (40 μg) for 12 h. Free peptide was removed by diafiltration. NC refers to unlabeled TEX Hepa . c Flow cytometric analysis of TEX-N1ND uptake in BMDCs at 24 h after incubation. NC refers to untreated DCs. Levels (d) and quantitative analysis (e) of surface markers and co-stimulatory molecules in BMDC TEX-N1ND , BMDC TEX/N1ND , or BMDC TEX (n = 3; one-way ANOVA post hoc Student-Newman-Keuls test). NC refers to untreated DCs. f Measurement of TNF-α secretion in supernatants of BMDC TEX , BMDC TEX/N1ND , or BMDC TEX-N1ND (n = 4; one-way ANOVA on ranks). g Analysis of IFN-γ and IL-2 in supernatants from splenic T lymphocytes activated by BMDC TEX-N1ND , BMDC TEX/N1ND , or BMDC TEX (n = 4; one-way ANOVA post hoc Student-Newman-Keuls test). h Cytolysis rate against different murine tumor cells with T cells activated by BMDC TEX-N1ND , BMDC TEX/N1ND or BMDC TEX , respectively. A LDH-releasing cytotoxic assay was performed to measure the cytolysis efficiency of effector T cells activated by BMDC TEX-N1ND , BMDC TEX/N1ND , or BMDC TEX , respectively in Hepa1-6, Panc02 (n = 5; one-way ANOVA post hoc Student-Newman-Keuls test) and 4T1 cells (one-way ANOVA on ranks). N refers to the number of individual biological replicate unless otherwise specified. Data are presented as means ± s.e.m. (*p < 0.05; **p < 0.001; n.s not significant). Source data are provided as a Source Data file. ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15569-2 week 7 and 9, compared with corresponding DC TEX and PBS treatment groups (Fig. 5a). DC TEX-N1ND treatment also elicited significantly more circulatory CD8 + effector memory T cells (T EM ; CD44 high CD62L low ) on week 7 and 9 24 , compared with DC TEX and PBS treatment (Fig. 5b), suggesting that increased T EM cells contributed to the long-lasting protective immunity. Consistently, the proportion of CD8 + CD44 high and CD8 + CD44 high CD127 high memory T cells, particularly T EM and CD8 + central memory T cells (T CM ; CD44 high CD62L high ), significantly increased in the secondary lymphoid tissue (spleen) from DC TEX-N1ND -treated mice 24 compared with other groups (Fig. 5c, d and Supplementary Fig. 5a). To ascertain if memory T cells induced by repeated DC TEX-N1ND vaccination accounted for the prolonged antitumor immunity observed, we immunized C57BL/6 mice with DC TEX-N1ND (2 × 10 6 ) intravenously once per week for 3 weeks. As expected, circulatory effector and memory  Fig. 5e) significantly rose. To further confirm the direct involvement of memory T cells in the long-lasting antitumor immunity triggered by DC TEX-N1ND , we isolated T EM and T CM from C57BL/6 mice immunized with DC TEX-N1ND under identical conditions as described above and intravenously administered T EM or T CM (5 × 10 6 ) into day-7 orthotopic HCC mice for single injection. Strikingly, tumor growth was significantly inhibited in T EM -and T CMtreated HCC mice compared with untreated controls (Fig. 5j), strengthening the notion that memory T cell induction mediated protective immunity against the tumor. These findings support the conclusion that memory T cells boosted by DC TEX-N1ND contribute to long-lasting protective immune response.
DC activation and lymph node homing boosts immune memory. To determine if DC TEX-N1ND has a direct impact on memory T cells, we immunized C57BL/6 mice with DC TEX (2 × 10 6 ) weekly for 3 weeks and harvested splenic T EM and T CM cells 4 weeks after the last immunization and stimulated 3 × 10 5 of each with DC TEX-N1ND or DC TEX (3 × 10 4 ) for 72 h. Stimulation of purified T CM by DC TEX-N1ND resulted in a dramatic increase in T CM and T EM numbers (Fig. 6a) and secreted IL-2 and IFN-γ ( Fig. 6b) compared with DC TEX treatment of controls, suggesting that DC TEX-N1ND enables efficient T CM proliferation and differentiation to T EM in response to stimulation. In the purified T EM culture, T EM numbers and IFN-γ significantly rose after DC TEX-N1ND treatment compared with other controls (Fig. 6c), indicating N1ND also triggered robust T EM cell proliferation. To assess the direct impact of DC TEX-N1ND on naive T cells, we incubated DC TEX-N1ND or DC TEX with naive T cells for 3 or 6 days. Effector T cells and IFN-γ were significantly elevated in DC TEX-N1ND -and DC TEX -treated naive T cells, though to a lesser extent with DC TEX , compared with untreated T cell controls (Fig. 6d). Examination of memory T cells on day 3 and 6 revealed significantly increased memory T cells in DC TEX-N1ND , compared with the controls (Fig. 6e), indicating that DC TEX-N1ND promotes accelerated generation of effector T cells early on which then promotes the differentiation of memory T cells. These findings demonstrate that DC TEX-N1ND augments generation of effector and memory T cells and results in amplification of immunological memory. Migration of DC to regional lymph nodes, a function of DC activation, is a critical factor influencing therapeutic outcomes of DC vaccines 26,27 . To examine TEX-N1ND's effect on exogenous DC migration to lymph nodes, we tracked the tissue distribution of DiI-labeled DCs (5 × 10 6 ) in tumor-bearing mice 48 h after single intravenous injection. IVIS analysis (Fig. 6f) and flow cytometry of extracted tissues ( Fig. 6g and Supplementary Fig. 6) indicated more DCs migrated to the tumor and secondary lymphoid tissues including spleen, mesenteric, and inguinal lymph nodes in DC TEX-N1ND -treated mice compared with DC TEX and controls. These findings strengthen the conclusion that TEX-N1ND endows DCs with stronger immunogenicity and homing capacity to lymph nodes and results in boosted generation of effector and memory T cells, which accounts for the long-lasting antitumor immunity.
N1ND-painted patient serum exosomes augment DC immunity. It was reported that HMGN1 expression levels are inversely correlated to clinical stage of tumor in cancer patients such as breast and non-small cell lung carcinoma 28 . HCC patient tumors expressed less HMGN1 than para-tumoral tissues, which was inversely correlated to the expression of HCC-specific antigen AFP ( Supplementary Fig. 8a). Consistently, HCC patient serum exosomes contained significantly less HMGN1 than exosomes from normal volunteers' serum ( Fig. 7a, b), whereas AFP was up to eightfold higher (Fig. 7b). To determine the applicability of TEX-N1ND in the human setting, we loaded N1NDmodified exosomes, derived from HCC patients' serum (TEX-N1ND), into human DCs derived from healthy volunteers' PBMCs (PMDC TEX-N1ND ). Allogeneic human lymphocytes Fig. 2 Effect of BMDC TEX-N1ND on large subcutaneous tumors. a Measurement of tumor volume in subcutaneous C57BL/6 HCC mice treated with BMDC TEX-N1ND , BMDC TEX/N1ND , or BMDC TEX (2 × 10 6 cells once per week for 3 weeks) at day 26 (n = 5; one-way ANOVA on ranks), day 29 (n = 5; oneway ANOVA post hoc Student-Newman-Keuls test), day 32 and day 35 (n = 5; one-way ANOVA on ranks), day 38 (n = 5; one-way ANOVA post hoc Student-Newman-Keuls test) and day 42 (n = 5; one-way ANOVA on ranks). TEXs were derived from murine Hepa1-6 cells. Flow cytometric analysis of different immune cells in blood (b) or tumor tissues (c) from HCC mice treated with BMDC TEX-N1ND , BMDC TEX/N1ND , or BMDC TEX , respectively (n = 4; one-way ANOVA post hoc Student-Newman-Keuls test; n represents the number of animals used for each group). Measurement of tumor volume in syngeneic subcutaneous C57BL/6 pancreatic cancer mice (d) or BALB/C breast cancer mice (e) treated with BMDC TEX-N1ND , BMDC TEX/N1ND , or BMDC TEX (2 × 10 6 cells once per week for 3 weeks) day 26 (n = 5; one-way ANOVA post hoc Student-Newman-Keuls test), day 29 (n = 5; one-way ANOVA post hoc Student-Newman-Keuls test), day 32, 35, and 42 (n = 5; one-way ANOVA on ranks). TEXs were derived from murine pancreatic or breast cancer cells. Flow cytometric analysis of different immune cells in blood (f) or tumor tissues (g) from subcutaneous tumor mice treated with BMDC TEX-N1ND , BMDC TEX/N1ND , or BMDC TEX (n = 4; one-way ANOVA post hoc Student-Newman-Keuls test). Data are presented as means ± s.e.m. (*p < 0.05; **p < 0.001; n.s not significant). Source data are provided as a Source Data file.    Supplementary Fig. 8b). Likewise, HMGN1 expression was reduced in exosomes from pancreatic and breast cancer patients' serum compared with normal volunteers (Supplementary Fig. 8c, d). To explore the general applicability of N1ND as an immunoadjuvant in the clinic, we painted exosomes from pancreatic and breast cancer patients' serum and loaded them into human PMDCs (PMDC TEX-N1ND ). Strikingly, significantly higher cytolysis rates against pancreatic and breast cancer patient cells were detected in lymphocytes stimulated by PMDC TEX-N1ND compared with corresponding PMDC TEX or unprimed lymphocytes (Fig. 7d). The data demonstrate that N1ND can augment human DC immunogenicity in vitro in the presence of tumor-specific antigen and thus provide an avenue for improving DC vaccines. a Flow cytometric analysis of CD8 + T lymphocytes in blood and tumor tissues from orthotopic C57BL/6 HCC mice treated with PBS, DC TEX , or DC TEX-N1ND on week 7 (one-way ANOVA post hoc Student-Newman-Keuls test) and 9 (twotailed t test) (n = 5). b Flow cytometric analysis of CD4 + CD25 + Tregs in blood and tumor tissues from orthotopic C57BL/6 HCC mice treated with PBS, DC TEX , or DC TEX-N1ND on week 7 (one-way ANOVA on ranks) and 9 (two-tailed t test) (n = 5). c Immunohistochemistry of CD3 + T cells and Foxp3 + Tregs in tumor sections from different treatment samples to determine the extent of T cell infiltration (scale bar = 100 or 50 μm in magnified images). Arrowheads point to CD3 + or Foxp3 + T cells. This experiment was repeated once (two in total). d Quantification of Foxp3 + T cells in tumor sections from different treatment groups on week 7 (one-way ANOVA on ranks) and 9 (two-tailed Mann-Whitney U test) (n = 3; n represents the number of animals used for each group). e Measurement of IFN-γ in tumor tissues from treated mice with ELISA on week 3 (n = 3), week 7 (n = 5; one-way ANOVA post hoc Student-Newman-Keuls test), and 9 (n = 5; two-tailed t test). f Measurement of immunosuppressive cytokines including TGF-β on week 3 (n = 3), week 7 (one-way ANOVA on ranks), and 9 (two-tailed Mann-Whitney U test) (n = 5) and IL-10 on week 3 (n = 3), 7 (n = 5; one-way ANOVA post hoc Student-Newman-Keuls test), and 9 (n = 5; two-tailed Mann-Whitney U test). N represents the number of animals used for each group. The comparison was conducted between DC TEX-N1ND and DC TEX or PBS groups at the same time-point. Data are presented as means ± s.e.m. (*p < 0.05; **p < 0.001; n.s not significant). Source data are provided as a Source Data file.

Discussion
It remains challenging to induce robust, durable antitumor immune response capable of markedly prolonging patient survival against large, poorly immunogenic tumors. Although DC vaccines show promises in the clinic with great safety profiles, limited clinical success has been achieved in treating large established tumors in either mice or human. Immunoadjuvants, particularly endogenous ones such as HMGN1, likely present fewer nonspecific immunological side effects than exogenous adjuvants such as bacterial lipopolysaccharides. Therefore, HMGN1 has been extensively investigated in combination with other chemotherapeutic agents and activators for other TLR   pathways or checkpoint inhibitors [13][14][15] , but the effective delivery of HMGN1 to tumors has confounded researchers. Here we described a simple and effective approach to load the potent N1ND immunoadjuvant on TEXs via an exosomal anchor peptide (TEX-N1ND), resulting in efficient delivery of N1ND to DCs together with TAAs. TEX-N1ND dramatically enhances DCs' antitumor immunity by improving DC activation, homing capacity to lymph nodes and the generation and amplification of long-lasting memory T cells. Memory T cells consequently contribute to robust and persistent tumor regression in mice bearing large established tumors. Strikingly, the amplified immunogenicity of DCs triggered by TEX-N1ND yielded prolonged tumor regression and persistent antitumor immune response, and also reshaped the immune and tumor microenvironment in mice with poorly immunogenic orthotopic large HCCs. This is particularly important for HCC as it is notorious for highly immunosuppressive tumor microenvironment which facilitates tumor evasion of the immune system 29 . Notably, TEX-N1ND could trigger a greater antitumor immunity of DCs if only CD63-positive TEXs were used. However, given the cost and efforts required for doing so, it might not be a worthwhile trade-off. Furthermore, TEX-N1ND enabled the augmentation of human DCs' immunogenicity irrespective of the endogenous expression level of HMGN1 in or origins of TEXs, demonstrating the clinical and general applicability of N1ND as an immunoadjuvant in DC-based immunotherapy. This is the first study, to our knowledge, to demonstrate the potency of N1ND as an immunoadjuvant and a simple approach to load TEXs with various cargoes for delivery and thus our study provides a platform to augment antitumor immunity induced by DC vaccines. In our current study, although tumor growth was significantly inhibited with DC TEX-N1ND in different mouse models bearing large tumor loads, all the mice still succumbed to tumor, underlining the tenacity of large established tumors at overcoming current therapeutics. Nevertheless, long-lasting antitumor immunity was established in both subcutaneous and orthotopic tumor mouse models, which remarkably prevented tumor progression and lung metastasis, two crucial outcomes that can lead to prolonged quality-adjusted survival in patients with a large tumor burden. The reduction in tumor size may also allow surgical intervention that would have been previously impossible. Further optimization including the delivery route and dosing regimens and long-term efficacy will be investigated in our future studies. Notably, both DC cell line (DC2.4) and the more clinically acceptable bone marrow (BM)-derived primary DCs were tested in our study and similar cellular uptake efficiency and immunogenicity were observed when DCs were pulsed with TEX-N1ND under an identical protocol, indicating the source of DCs has no impact on the potency of N1ND and TEXs. However, considering that the low yield of BMDC limits its scale-up, particularly for in vivo studies, we used DC cell line for subsequent orthotopic studies. Perhaps as a positive, we believe that we are moderately successful at shifting the bottleneck of DC immunotherapy from the lack of immunostimulation to production of sufficient numbers of DC for therapy, which can be considered a significant progress in this field.
HMGN1 was chosen in our study as it can recruit and activate not only DCs but also NK cells, neutrophils, and monocytes/ macrophages 12,13,30 . The truncated N1ND fragment painted on TEX as expected was effective at DC activation, which consequently boosted CD8 + T cells. Also significantly increased numbers of memory T cells were induced by DC TEX-N1ND , which largely contribute to the long-lasting antitumor immunity.

Notably, the increase in memory T cells induced by DC TEX-N1ND over DC TEX in blood was mainly in the number of effector memory T cells rather than central memory T cells as evidenced by significantly elevated numbers of effector memory T cells and only a marginal increase in central memory T cells from DC TEX-
N1ND -treated group compared with DC TEX . As expected, NK cells significantly increased in blood and tumor tissues from mice with large established tumors in three different tumor mouse models ( Supplementary Fig. 9a), suggesting that there could be other non-T cell innate immune effects related to the long-lasting antitumor immunity elicited by TEX-N1ND 12 . Nevertheless, T cell-specific immune responses are primarily responsible for the observed antitumor effect as there was no significant tumor inhibition detected in subcutaneous HCC-bearing thymus-deficient but NK-intact nude mice when treated with BMDC TEX-N1ND ( Supplementary Fig. 9b, c).
Interestingly, HMGN1 was reported to decline in advanced stages of breast and non-small cell lung cancer patients 28 , and is also likely to be the case for HCC and pancreatic cancer. We have indeed observed that HMGN1 expression is lowered in HCC tumor than para-tumoral tissues. In this study, we show that TEX-N1ND is effective irrespective of the endogenous HMGN1 expression in tumors and it seems that it works better in tumors with low HMGN1 expression, which is loosely correlated to staging of cancer. Furthermore, this DC vaccine can also be used in large poorly immunogenic tumors. Thus, we are hopeful that the patients who are going to benefit the most from this treatment modality are those with advanced cancers that do not respond to other therapies. Worthy of mentioning, a uniqueness of our study is that we demonstrated the ability to use cancer exosomes to deliver endogenous tumor antigens and exogenous cargoes (such as N1ND) by painting with CP05, and thus simplify the process of developing improved DC vaccines.
A difficulty to test our system in autochthonous HCC mouse models bearing large tumor burdens is that it is hard to control Flow cytometric analysis of long-lived memory T cells (a) or T EM cells (b) in blood from day-21 orthotopic HCC mice treated with DC TEX-N1ND , DC TEX or PBS on week 7 (one-way ANOVA on ranks) and 9 (two-tailed t test). CD44 hi or CD127 hi means CD44 high or CD127 high , respectively (n = 3). MFI mean fluorescence intensity. Quantitative analysis of long-lived memory T cells (c) or T EM and T CM cells (d) in the spleen from day-21 orthotopic HCC mice treated with DC TEX-N1ND , DC TEX , or PBS on week 7 (one-way ANOVA post hoc Student-Newman-Keuls test) and 9 (two-tailed t test), respectively (n = 5). CD62L hi represents CD62L high . Flow cytometric analysis of effector and memory T cells (e) or T EM cells (f) in blood from immunized wild-type C57BL/6 mice before tumor challenge (one-way ANOVA post hoc Student-Newman-Keuls test). These mice were immunized intravenously with DC TEX-N1ND , DC TEX (2 × 10 6 cells once per week for 3 weeks) or PBS, respectively and blood was collected 4 weeks after last immunization before challenge (n = 4). Measurement of tumor growth (g) or tumor volume/ weight (h) of immunized C57BL/6 mice at 4 weeks after challenge (n = 4; one-way ANOVA post hoc Student-Newman-Keuls test). i Flow cytometric analysis of effector and memory T cells in tumor tissues from C57BL/6 mice immunized with DC TEX or DC TEX-N1ND 4 weeks after tumor challenge (n = 3; one-way ANOVA on ranks). TILs refer to tumor-infiltrating lymphocytes. j Measurement of tumor growth of day-7 established orthotopic HCC mice treated with memory T cells isolated from DC TEX-N1ND -immunized C57BL/6 mice (n = 5; one-way ANOVA post hoc Student-Newman-Keuls test). DC TEX-N1ND was used as a positive control. Data are presented as means ± s.e.m. (*p < 0.05; **p < 0.001; n.s not significant). Source data are provided as a Source Data file.
tumor size as multinodular, dispersed, and nonuniform tumors formed in carcinogen-induced autochthonous HCC mice 31 . In order to closely evaluate the immunosuppressive nature of local tumor immune microenvironment manifested in late-stage HCC patients, we transplanted fewer tumor tissues to allow the establishment of immunosuppressive microenvironment in orthotopic HCC mice as demonstrated by significantly increased numbers of immunosuppressive Tregs and elevated levels of TGF-β in blood and tumor tissues from day-21 orthotopic HCC mice compared with normal controls (Supplementary Fig. 3a).  Therefore, as a proof-of-concept study, orthotopic HCC transplantation models were used. However, further long-term investigation of DC TEX-N1ND will be undertaken in diethylnitrosamineinduced autochthonous HCC models after optimization of treatment doses and kinetics.
Taken together, our study demonstrates a simple and direct approach to efficiently load cargoes for TEX, as well as the potency of N1ND in augmenting DC immunogenicity by promoting DC activation, homing capacity to lymph nodes and thus accelerating the generation of long-lasting memory T cells. This in turn contributes to the robust and persistent antitumor immunity in large established tumors in mice and human cell in vitro. Our study provides a platform for amplifying the antitumor immunity of DC vaccines and thus accelerates the development of DC-based immunotherapy.

Methods
Mice. C57BL/6 wild-type (H-2 b ), BALB/C mice (H-2 d ), and thymus-deficient BALB/C nude mice (6-8-week old) were used in all experiments (five mice were used in each group for ectopic or orthotopic studies unless otherwise specified). Mice were housed under specific pathogen-free conditions in a temperature-and humidity-controlled room. All the animal experiments were carried out in the animal unit, Tianjin Medical University (Tianjin, China) according to procedures authorized and specifically approved by the institutional ethical committee (Permit Number: SYXK 2009-0001). Mice were sacrificed by CO 2 inhalation at desired time-points, and tissues were either snap-frozen in liquid nitrogen-cooled isopentane or fixed with Bouin's solution (Sigma, USA) and embedded with paraffin for histological study.   6 Effect of DC TEX-N1ND on T cells from DC TEX-N1ND -immunized mice. C57BL/6 mice were immunized with DC TEX-N1ND (2 × 10 6 cells once per week for 3 weeks) and splenic memory or naive T cells were isolated from immunized mice 4 weeks after last immunization. a Quantification of T CM and T EM cells in T CM cell populations after the co-incubation of T CM cells with DC TEX-N1ND or DC TEX (3 × 10 4 ) for 3 days (n = 3; one-way ANOVA post hoc Student-Newman-Keuls test; n represents the number of animals immunized with DC TEX-N1ND ). b Levels of IFN-γ and IL-2 in the culture medium after the co-incubation of T CM cells with DC TEX-N1ND or DC TEX (3 × 10 4 ) for 3 days (n = 3; one-way ANOVA post hoc Student-Newman-Keuls test; n represents the number of animals immunized with DC TEX-N1ND ). c Analysis of T EM cells and IFN-γ in T EM cell populations stimulated with DC TEX-N1ND or DC TEX (3 × 10 4 ), respectively (n = 3; one-way ANOVA post hoc Student-Newman-Keuls test; n represents the number of animals immunized with DC TEX-N1ND ). d Analysis of effector T cells (one-way ANOVA on ranks) and IFN-γ (one-way ANOVA post hoc Student-Newman-Keuls test) in naive T cell populations stimulated with DC TEX-N1ND or DC TEX (3 × 10 4 ) for 3 or 6 days, respectively (n = 4; n represents the number of animals used for isolation of naive T cells). e Quantification of memory T cells in naive T cell populations after the co-incubation of naive T cells with DC TEX-N1ND or DC TEX (3 × 10 4 ) on day 3 and 6, respectively (n = 3; one-way ANOVA post hoc Student-Newman-Keuls test; n represents the number of animals used for isolation of naive T cells). f Tissue distribution and quantification of DiI-labeled DC TEX-N1ND or DC TEX (5 × 10 6 ) in orthotopic HCC mice 48 h after single intravenous injection (n = 6; one-way ANOVA on ranks). The ratio is relative to untreated DCs. LN lymph node. g Flow cytometric analysis of DiI-labeled DC TEX-N1ND or DC TEX in orthotopic HCC mice 48 h after single intravenous injection (n = 8; one-way ANOVA post hoc Student-Newman-Keuls test). These experiments were repeated once (two in total). Data are presented as means ± s.e.m. (*p < 0.05; **p < 0.001; n.s not significant). Source data are provided as a Source Data file.  Fig. 7 TEX-N1ND augmented human DC immunity. a Western blot analysis for detecting the levels of HMGN1 and AFP expression in serum exosomes derived from normal healthy volunteers and HCC patients. Total protein (20 μg) was loaded and CD63 was used as an exosomal marker protein. This experiment was repeated once (two in total). b Quantitative analysis of levels of HMGN1 (two-tailed t test) and AFP (two-tailed Mann-Whitney U test) in serum exosomes derived from normal healthy volunteers (n = 7) and HCC patients (n = 8). c Cytolysis rate against different human HCC cells with PMDC TEX-N1ND and PMDC TEX . TEXs were derived from HCC patients' serum. PMDC means peripheral mononuclear-derived dendritic cell. A LDH-releasing cytotoxic assay was performed to measure the cytolysis efficiency of effector lymphocytes activated by PMDC TEX-N1ND or PMDC TEX , respectively (n = 5; one-way ANOVA post hoc Student-Newman-Keuls test except for Hep3B in which one-way ANOVA on ranks was used). Control refers to human PBMCderived lymphocytes. d Cytolysis rate against human pancreatic and breast cancer cells with PMDC TEX-N1ND and PMDC TEX . TEXs were derived from pancreatic or breast cancer patients' serum, respectively (n = 10; one-way ANOVA on ranks). These experiments were repeated once (two in total). Data are presented as means ± s.e.m. (*p < 0.05; **p < 0.001; n.s not significant). Source data are provided as a Source Data file.
centrifugation at 100,000 g for 1 h to remove possible FBS-containing exosomes. Murine HCC cell line Hepa1-6 (H-2 b ) was purchased from Boster Biological Technology Ltd (Wuhan, China) and cultured in DMEM with 2-mM Gln and 10% FBS as per manufacturer's instructions. Murine pancreatic cancer cell line (Panc02, H-2 b ), Lewis lung cancer cell line (LLC1, H-2 b ), breast cancer cell line (4T1, H-2 d ), cervical cancer cell (Hela), and lymphoma cell (EL4) were kept in house and cultured as previously described 33,34 . Human HCC cell lines including HepG2 and Hep3B (purchased from ATCC biobank) and MHCC-97H (purchased from Shanghai Institute for Biological Sciences, Chinese Academy of Sciences) and LM3 (purchased from BeNa Culture Collection, Beijing, China) and cultured as per manufacturer's instructions 35 . Human pancreatic cancer cell line (Panc1) and breast cancer cell line (MCF7) were purchased from ATCC biobank and cultured as per manufacturer's instructions. Briefly, cells were grown at 37°C in 5% CO 2 in DMEM supplemented with 10% FBS obtained by centrifugation at 100,000 g for 1 h to remove possible FBS-containing exosomes, and 1% penicillin and streptomycin.
Peptides. N1ND (HMGN1 1-52 ) 12  Preparation of exosomes. Cell culture medium was sequentially centrifuged at 1000 g for 10 min, followed by 10,000 g for 30 min. The supernatant was collected and filtered with a 0.22-μm filter (Millex, Germany), followed by ultracentrifugation at 100,000 g for 1 h to pellet exosomes. Exosome pellets were washed in a large volume of PBS and recovered by centrifugation at 100,000 g for 1 h. Characterization of exosomes. Exosomal morphology was visualized using a high-resolution transmission electron microscope (TEM, Hitachi HT7700, Tokyo, Japan) 19 . Briefly, the resuspended exosomes were diluted into PBS (1 μg μl −1 ) and mixed with an equal volume of 4% paraformaldehyde. Exosomes were adsorbed onto a glow-discharged, carbon-coated formvar film attached to a metal specimen grid. Excess solution was blotted off and the grid was immersed with a small drop (50 μl) of 1% glutaraldehyde for 5 min followed by washing with 100 μl distilled water for eight times (2 min each time). Subsequently, the grid was transferred to 50 μl uranyl-oxalate solution (pH7.0) for 5 min and then 50 μl methyl celluloseuranyl acetate (100 μl 4% uranyl acetate and 900 μl 2% methyl cellulose) for 10 min on ice. The excess solution was blotted off and the sample was dried and examined with TEM.
Antigen-specific assay. Subcutaneous During the examination, mice were anesthetized with pentobarbital sodium and fixed to minimize body motion, respiration rate was monitored, and body temperature was maintained to be at 37°C using warm airflow. The tumor size was calculated based on MRI images for 3, 5, and 7 weeks in orthotopic HCC mice bearing large established tumor. Each mouse was marked and imaged with MRI at different time-points.
Western blot. Protein extraction and western blot were carried out as previously described 18 . Various amounts of protein prepared from cell lysates and TEXs were loaded onto SDS-polyacrylamide gel (10% , and TRAF6 (1:5000; Abcam, UK), respectively. The bound primary antibodies were detected by peroxidase-conjugated goat anti-mouse, rabbit anti-mouse, or goat anti-rabbit IgGs (1:5000; Sigma, USA), respectively, and the ECL western blot analysis system (Millipore, Billerica, MA) was used. Uncropped images of all western blots can be found in Supplementary Fig. 10.
Flow cytometry and intracellular staining. For the detection of murine DC or BMDC surface markers, DCs or BMDCs were washed in FACS buffer (PBS containing 0.5% BSA and 0.05% NaN3) and stained with FITC-anti-mouse IgG2b Isotype ( Cytokine release and cytolysis assay. Murine DCs or BMDCs were pulsed with N1ND-modified Hepa1-6, Panc02, and 4T1 TEXs or exosomes from HCC, pancreatic cancer, and breast cancer patients' serum and incubated for 48 h at the final concentration of 40 μg ml −1 , respectively. Mixed splenic T lymphocytes were harvested from unimmunized C57BL/6 or BALB/C mice and incubated with (BM) DC TEX-N1ND or (BM)DC TEX for 3 days. Levels of IFN-γ and IL-2 in supernatant were detected in activated T lymphocytes per the manufacturer's instructions. For mouse in vitro study, IFN-γ and IL-2 ELISA kits were purchased from R&D systems and eBioscience (USA). For in vivo cytokine assay, mouse blood was harvested from orthotopic tumor-bearing mice 2 days after the last immunization and centrifuged at 3000 g for 30 min at room temperature, followed by analysis of IFN-γ (R&D systems, USA), IL-10, and TGF-β (eBioscience, USA), respectively. For human cytolysis assay, mixed population of lymphocytes derived from human PBMC were co-cultured with human DC activated by TEX-N1ND for 3 days. Subsequently, activated human lymphocytes were tested again different human cancer cells. The cytotoxicity detection kit (R&D systems, USA) was used to measure the cytolysis rate elicited by effector T cells against different tumor cells. For the cytolysis assay, 1 × 10 5 effector T cells (E) were used to lyse against 1 × 10 4 target cells (T) with an E:T ratio of 10:1. For other E:T ratios, the number of target cells remained the same while the number of effector T cells changed accordingly. The amount of lactate dehydrogenase (LDH) released from lysed target cells was used as an indicator for cytolysis. Cytolysis rate (%) was calculated based on the equation: cytotoxicity (%) = (((effector/target cell mix − effector cell control) − low control)/(high control − low control)) × 100 as per manufacturer's instructions (Roche, CA, USA).
Isolation of leukocytes from mouse blood and tissues. Blood from ectopic and orthotopic tumor-bearing mice treated with (BM)DC TEX-N1ND , (BM)DC TEX , (BM) DC, or PBS was collected with 1% heparin, followed by addition of equal volume of Lymphoprep™ (Stemcell Technologies, Canada) and lysis with the ammonium chloride-potassium (ACK) lysis buffer for 5 min at room temperature to generate lymphocyte suspensions. Isolation of T lymphocytes from mouse tumor tissues was adopted 37 . Briefly, tumor tissues were minced into small pieces with surgical scissors, gently forced through a 200-μm-gauge stainless steel mesh with a sterile syringe plunger and digested in collagenase type IV suspension (0.05 mg ml −1 , Worthington Biochem. Corp., USA) for 40 min at 37°C. The resulting suspension was filtered through the 70-μm cell strainer. The extract was centrifuged at 528 g for 10 min and the supernatant was removed. The mixture was resuspended with 10 ml 40% percoll (Pharmacia, Sweden) followed by centrifugation at 850 g for 30 min at 22°C to remove the supernatant. Cell pellets were resuspended in ACK lysis buffer to remove red blood cells and the rest cells were stained with different fluorescence-labeled antibodies as described above.
Immunohistochemistry and histological staining. To examine the presence of CD3 + T and Treg lymphocytes in tumor sections from orthotopic HCC mice treated with DC TEX-N1ND or control groups, mouse tumor tissues were fixed in Bouin's solution (Sigma, USA) and embedded with paraffin. CD3 + or Foxp3 + T lymphocytes were stained with rabbit anti-mouse polyclonal antibodies CD3 (1:2000; Novus, USA) or Foxp3 (1:2000; Abcam, UK), respectively, and detected by goat anti-rabbit secondary antibody (1:5000; Sigma, USA). Routine Hematoxylin and eosin (H&E) staining was used to examine the pulmonary metastasis of lung tissues from orthotopic HCC mouse models. Briefly, tissues were fixed in Bouin's solution and embedded with paraffin followed by staining with hematoxylin and eosin. The slides were incubated in filtered 0.1% Mayer's haematoxylin for 10 min for nucleus and ribosome staining. Afterwards, sections were rinsed in coolrunning distilled H 2 O to remove excessive haematoxylin, and treated with 0.5% hydrochloric acid, and 1% sodium bicarbonate solution, respectively. After that, sections were incubated in 0.5% eosin (0.5 g dissolved in 100 ml of 95% EtOH) for 5 min for cytoplasm staining, and washed by distilled H 2 O. The sections were dehydrated in ingredients of 70, 80, 95, 100, 100% EtOH for 5 min each. Finally, sections were treated in xylene for 10 min and coverslips were mounted with neutral gum, followed by visualization with an inverted microscope (OLYMPUS, Tokyo, Japan).
In vitro assay for T cell activation and proliferation. To examine the effect of DC TEX-N1ND on memory and naive T cells, wild-type C57BL/6 mice were immunized with DC TEX (2 × 10 6 ) intravenously for once per week for 3 weeks and splenocytes were harvested 4 weeks post vaccination. T cell subsets including CD8 + CD44 high CD62L high/low memory and CD8 + CD44 low CD62L high naive T cells were isolated with FACS sorting with BD FACSAriall (BD Biosciences, USA). Memory or naive T cells (3 × 10 5 ) were then incubated with DC TEX or DC TEX-N1ND (3 × 10 4 ) for 3 or 6 days in 96-well plates. On day 3 or 6 after co-incubation, each CD8 + T cell subset in treated samples was analyzed by flow cytometry.
T cell adoptive transfer. C57BL/6 mice were immunized with DC TEX-N1ND (2 × 10 6 ) once per week for 3 weeks. CD8 + CD44 high CD62L high/low memory T cell subsets were harvested from spleen and lymph nodes of immunized mice 4 weeks after last immunization by flow cytometry sorting with BD FACSAriall (BD Biosciences, USA). CD8 + CD44 high CD62L high/low memory T cells (5 × 10 6 ) were injected intravenously into day-7-established orthotopic HCC mice for once. DC TEX-N1ND was administered as a positive control with the same dosing regimen as used in the immunization protocol. The tumor volume was measure 2 weeks after treatment as described above.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The source data underlying Figs. 1a, e-h, 2a-g, 3b, d, e, 4a, b, d-f, 5a-j, 6a-g, 7a-d and Supplementary Figs. 1a, b, 2d-f, 3a, d, e, 4a, b, 5b-e, 7a, b, d, e, 8a, c, d, 9a and c are provided as a Source Data file. All the other data supporting the findings of this study are available within the article and its supplementary information files and from the corresponding author upon reasonable request.