Adoptively transferred tumor-specific IL-9-producing cytotoxic CD8⁺ T cells activate host CD4⁺ T cells to control tumors with antigen loss

Abstract

Host effector CD4⁺ T cells emerge as critical mediators for tumor regression but whether they can be activated by adoptively transferred CD8⁺ T cells remains unknown. We previously reported that adoptive transfer of interleukin 9 (IL-9)-producing cytotoxic CD8⁺ T (Tc9) cells achieved long-term control of tumor growth. Here, we demonstrate that murine tumor-specific Tc9 cells control the outgrowth of antigen-loss relapsed tumors by recruiting and activating host effector CD4⁺ T cells. Tc9 cells secreted IL-24 and recruited CCR7-expressing conventional type 2 dendritic cells (cDC2 cells) into tumor-draining lymph nodes to prime host CD4⁺ T cells against relapsed tumors. Host CD4⁺ T cell or cDC2 deficiency impaired the ability of Tc9 cells to control relapsed tumor outgrowth. Additionally, intratumoral IL24 expression correlates with cDC2 and CD4⁺ T cell gene signatures in human cancers and their expression is associated with better patient survival. This study reports a mechanism for activation of tumor-specific CD4⁺ T cells in vivo.

Cancer immunotherapies relying on targeted destruction of cancer cells by potent antitumor T cells have achieved unprecedented success in recent years^1,2. As a main form of cancer immunotherapies, adoptive T cell therapy (ACT) has shown a durable response in certain persons with cancers^2–4. However, this response is often short-lived and tumor relapse occurs because of the outgrowth of antigen-loss variant (ALV) tumors and poor antitumor immune response^5–7. To prevent tumor relapses, several strategies have been developed, such as the activation of the host immunity following ACT. As the tumor progresses, the host spontaneously initiates an immune response by recognizing the tumor^8,9. Growing evidence suggests a positive correlation between host lymphocyte infiltration and improved prognosis in various cancers, including melanoma, breast cancer and colon cancer^10,11. However, most tumor cells are poorly immunogenic and host immunity is hindered by the immunosuppressive tumor microenvironment (TME)^12,13. Therefore, activation of the host immunity is considered an attractive option for cancer immunotherapy^10,11,13,14. Regarding the activation of the host immunity, most studies have focused on CD8⁺ T cells^8,13,15,16 but recent studies indicate that CD4⁺ T cells also recognize higher percentages of neoantigens than previously appreciated and elicit potent antitumor ability^17,18. It has been reported that host effector CD4⁺ T cells can be induced to elicit strong antitumor effects in response to cytotoxic T lymphocyte-associated protein 4 (CTLA4) therapy or vaccine treatment^18–20. However, whether adoptively transferred CD8⁺ T cells can eradicate tumors by inducing a host CD4⁺ T cell response has not been described.

In addition to traditionally cytotoxic CD8⁺ T cells (CTL/Tc1), emerging studies have demonstrated that naive CD8⁺ T cells can differentiate into various subsets^21,22. Interleukin 9 (IL-9)-secreting cytotoxic T (Tc9) cells, a recently identified subset of CD8⁺ T cells, are generated in vitro from naive CD8⁺ T cells under IL-9-producing T helper (Th9)-polarizing conditions²³ and have been detected in various mouse and human tissues^24,25. Here, we report that tumor-specific Tc9 cells inhibit the growth of relapsed tumors, which have lost the expression of primary tumor antigens, by inducing host CD4⁺ T cell responses targeting the ALV tumors. A mechanistic study revealed that tumor-infiltrating Tc9 cells secreted high levels of IL-24, which attracted dendritic cells (DCs), especially conventional type 2 DCs (cDC2), to activate host effector CD4⁺ T cells and control the growth of ALV tumors in murine tumor models. Similarly, we identified an IL24–cDC2–effector CD4⁺ T cell axis in human melanoma and breast cancer and observed positive correlations between the expression of their gene signatures and the survival of persons with cancer.

Results

Tumor-specific Tc9 cells control the outgrowth of ALVs

In murine melanoma B16^OVA (B16 cells overexpressing ovalbumin) tumor-bearing mice, injection of OVA-specific OT-I Tc9 cells mediated a superior antitumor effect and long-term tumor control, while tumors in OT-I Tc1-treated mice relapsed rapidly (Fig. 1a and Extended Data Fig. 1a–c). Similar results were observed with Pmel-1 Tc9 and Tc1 cells recognizing gp100 antigen in the weakly immunogenic B16 cells and MC38^gp100 colon adenocarcinoma cells (MC38 cells overexpressing gp100) (Fig. 1b and Extended Data Fig. 1d–g). The superior antitumor effect of Tc9 cells is consistent with our previous publications^22,23. As it has been reported that relapsed tumors after immunotherapies often become mutated and lose the expression of the primary tumor antigens under ACT pressure^5,26,27, we hypothesized that the long-term tumor control ability of Tc9 cells may be because of suppression of the outgrowth of ALVs. To test this hypothesis, we determined the expression of OVA antigen in B16^OVA tumor cells on days 10 and 40 after tumor inoculation in OT-I Tc9-treated or Tc1-treated mice. Indeed, the percentage of OVA-expressing B16^OVA cells on day 40 when tumors relapsed in Tc1-treated mice dramatically decreased compared to those on day 10 in mice treated with either OT-I Tc1 or Tc9 cells (Fig. 1c and Extended Data Fig. 1h). Similarly, the percentage of gp100-expressing B16 and MC38^gp100 tumor cells also decreased in Pmel-1 Tc1-treated or Tc9-treated mice on day 40 after tumor cell inoculation (Fig. 1d and Extended Data Fig. 1i). Considering that Tc9 cells exerted long-term control of tumor growth and ACT-treated tumors lost OVA or gp100 expression, these findings suggest that Tc9 cells may possess the capacity to not only kill the primary tumor cells but also suppress the outgrowth of relapsed ALVs in vivo.

Fig. 1 |. Tumor-specific Tc9 cells control the growth of ALVs.

a,c, PBS, CD45.1⁺ OT-I Tc1 or Tc9 cells were transferred into CD45.2⁺ B6 mice bearing B16^OVA tumors with the adjuvant treatment (CTX, DC and rhIL-2). Tumor growth curve (a; n = 5 mice per group, one of two independent experiments) and percentages of OVA^negative and OVA^positive CD45⁻ tumor cells on day 10 (n = 3 mice) and day 40 (n = 4 mice) after tumor inoculation (c; one of two independent experiments). b,d, PBS (n = 5 mice), Thy1.1⁺ Pmel-1 Tc1 (n = 7 mice) or Tc9 (n = 7 mice) cells were transferred into Thy1.2⁺ B6 mice bearing WT B16 tumors with the adjuvant treatment. Tumor growth curve (b; one of two independent experiments) and percentages of gp100^negative and gp100^positive CD45⁻ tumor cells on day 10 (n = 4 mice) and day 40 (n = 6 mice) after tumor inoculation (d; one of two independent experiments). e, Tumor size of CD45.2⁺ B6 mice inoculated with chimeric tumor cells (containing 4 × 10⁵ B16^OVA and 2 × 10⁵ B16 WT cells) followed by injection of PBS, CD45.1⁺ OT-I Tc1 or Tc9 with the adjuvant treatment or inoculated with only 2 × 10⁵ B16 WT cells with the adjuvant treatment (n = 5 mice per group, one of two independent experiments). f, Tumor size of Thy1.2⁺ B6 mice inoculated with chimeric tumor cells (containing 4 × 10⁵ B16 and 2 × 10⁵ B16^gp100-kd cells) followed by injection of PBS, Thy1.1⁺ Pmel-1 Tc1 or Tc9 with the adjuvant treatment or inoculated with only 2 × 10⁵ B16^gp100-kd cells with the adjuvant treatment (n = 5 mice per group, one of three independent experiments). g, Tumor size of Thy1.2⁺ B6 mice inoculated with chimeric tumor cells (containing 8 × 10⁵ MC38^gp100 and 2 × 10⁵ MC38 cells) followed by injection of PBS, Thy1.1⁺ Pmel-1 Tc1 or Tc9 with the adjuvant treatment (n = 5 mice per group, one of two independent experiments). Data are presented as the mean ± s.e.m. In c,d, a two-way ANOVA was used to compare antigen percentages. In a,b,e–g, multiple t-tests (one per row) were used to compare tumor growth on day 55 (a), day 40 (b), day 45 (e), day 35 (f) and day 40 (g).

To verify this, we constructed chimeric tumors containing a 2:1 ratio of B16^OVA cells (0.4 million) and B16 tumor cells (0.2 million) and inoculated them together into CD45.2⁺ B6 mice, followed by OT-I Tc1 or Tc9 cell transfer. Mice injected with 0.2 million B16 tumor cells were used as negative controls to monitor the growth of OVA-lost B16 tumor cells (Extended Data Fig. 2a). Again, OT-I Tc9 cells mediated a sustained suppression of the chimeric tumor growth whereas OT-I Tc1 cells exhibited temporary tumor control followed by aggressive tumor recurrence (Fig. 1e). In another chimeric tumor model that contained B16 and B16^gp100-kd (knockdown) chimeric tumor cells (Extended Data Fig. 2b,c), Pmel-1 Tc9 cells rather than Tc1 cells exerted long-term control of the chimeric tumors (Fig. 1f) and their tumor control ability was stronger than central memory and stem memory CD8⁺ T cells (Extended Data Fig. 2d,e). Similar results were found in B16–B16^gp100-ko (knockout) and MC38^gp100–MC38 chimeric tumor models (Fig. 1g and Extended Data Fig. 2f–i). These findings indicate that tumor-specific Tc9 cells indeed possess the ability to control the growth of tumor antigen-expressing primary and relapsed ALV tumors in vivo.

Tc9 cells suppress the growth of rechallenged antigen-negative tumors

To confirm that ACT with tumor-specific Tc9 cells is able to control the outgrowth of ALV tumors in vivo, we rechallenged OT-I Tc1-treated or Tc9-treated B16^OVA tumor-bearing mice with B16 tumor cells on the contralateral flank of mice on day 11 after T cell transfer, when the primary B16^OVA tumors largely regressed. As a negative control, B6 mice treated with adjuvant therapy alone were injected with B16 tumor cells (Fig. 2a). B16 cells rapidly developed tumors in adjuvant-treated naive mice, whereas OT-I Tc1-treated mice developed much larger tumors and displayed shorter survival than OT-I Tc9-treated mice upon tumor rechallenge (Fig. 2b,c). Similarly, rechallenging Pmel-1 Tc1-treated or Tc9-treated mice with B16^gp100-kd or MC38 tumor cells on the contralateral flank resulted in significantly larger tumors and shorter survival in Pmel-1 Tc1-treated mice compared to Tc9-treated mice (Fig. 2d–h). It should be noted that, because only one injection of Tc9 or Tc1 cells was given to tumor-bearing mice on day 9 after primary tumor cell inoculation, Tc9 cells could not completely control the growth of rechallenged tumors. As the rechallenged tumor cells injected on the contralateral flanks in both models did not express the primary tumor antigens, these results suggest that tumor-specific Tc9 cells may control the outgrowth of relapsed ALV tumors by eliciting a host immune response against ALV tumors in vivo.

Fig. 2 |. Tc9 cells inhibit growth of rechallenged antigen^neg tumors.

a–c, CD45.1⁺ OT-I Tc1 or Tc9 cells were transferred into CD45.2⁺ B6 mice bearing 6 × 10⁵ B16^OVA tumors with the adjuvant treatment. On day 20, OT-I Tc1-pretreated or Tc9-pretreated mice were rechallenged with 2 × 10⁵ WT B16 cells on the contralateral flank of the primary B16^OVA tumor. Naive B6 mice with adjuvant therapy were inoculated with 2 × 10⁵ WT B16 cells. A schema of the experiment (a), tumor growth curves (b; n = 5 mice per group, one of two independent experiments) and survival plots (c; n = 5 mice per group, one of two independent experiments) of treated mice are shown. d–f, Thy1.1⁺ Pmel-1 Tc1 (n = 5 mice) or Tc9 (n = 5 mice) cells were transferred into Thy1.2⁺ B6 mice bearing 6 × 10⁵ WT B16 tumors with the adjuvant treatment. On day 20, Pmel-1 Tc1-pretreated or Tc9-pretreated mice were rechallenged with 2 × 10⁵ B16^gp100-kd cells on the contralateral flank of the primary WT B16 tumor. Naive B6 mice with adjuvant therapy were inoculated with 2 × 10⁵ B16^gp100-kd (n = 4 mice). A schema of the experiment (d), tumor growth curves (e; one of two independent experiments) and survival plots (f; n = 5 mice per group, one of two independent experiments) of treated mice are shown. g,h, Thy1.1⁺ Pmel-1 Tc1 or Tc9 cells were transferred into Thy1.2⁺ B6 mice bearing 1 × 10⁶ MC38^gp100 tumor cells with the adjuvant treatment. On day 20, Pmel-1 Tc1-pretreated or Tc9-pretreated mice were rechallenged with 2 × 10⁵ MC38 cells on the contralateral flank of the primary MC38^gp100 tumor. A schema of the experiment (g) and tumor growth curves (h; n = 5 mice per group, one of two independent experiments) of treated mice are shown. Data are presented as the mean ± s.e.m. In b,e,h, multiple t-tests (one per row) were used to compare tumor growth on day 50 (b), day 40 (e) and day 55 (h). In c,f, a log-rank (Mantel–Cox) test was performed to compare survival curves.

Host CD4⁺ T cells are recruited into chimeric tumors

Because Tc9 cells may elicit host immunity against ALV tumors, we investigated host immune responses against ALVs in Tc1-treated or Tc9-treated mice. First, we examined immune cells within the TME using cytometry by time of flight (CyTOF) (Fig. 3a and Extended Data Fig. 3a). Interestingly, we observed that the percentage of host CD4⁺ but not CD8⁺ (Thy1.2⁺ CD8⁺) T cells was significantly higher in Pmel-1 Tc9-treated ALV-containing chimeric tumors compared to Tc1-treated mice (Extended Data Fig. 3b–d). Consistent with our previous studies^22,23,28, the persistence of adoptively transferred Tc9 cells was significantly greater than that of Tc1 cells (Extended Data Fig. 3e). Notably, host CD4⁺ T cells in Pmel-1 Tc9-treated mice expressed higher levels of interferon-γ (IFNγ) and granzyme B (GZMB) compared to those in Tc1-treated mice, while the Foxp3⁺ ratio in CD4⁺ T cells was similar (Extended Data Fig. 3f,g), suggesting that these host CD4⁺ T cells may possess cytolytic ability and control tumor growth^19,29,30. Furthermore, the numbers of host CD4⁺ T cells, especially IFNγ⁺ and GZMB⁺ CD4⁺ T cells, were significantly higher in Tc9-treated B16 and MC38 chimeric tumors compared to Tc1-treated tumors, as measured by flow cytometry (Fig. 3b,c and Extended Data Fig. 3h,i), whereas the numbers of IFNγ⁺ and GZMB⁺ host CD8⁺ T cells were comparable between Tc1-treated and Tc9-treated groups (Extended Data Fig. 3j, k). Additionally, when we transferred Tc9 cells into wild-type (WT) or CD8^−/− B6 mice bearing B16 and B16^gp100-kd chimeric tumors, tumor growth rates in WT and CD8^−/− mice were similar (Extended Data Fig. 3l). Consistent with the CyTOF results, the number of adoptively transferred CD8⁺ T cells was significantly greater in Tc9-treated tumors than Tc1-treated ones (Extended Data Fig. 3m,n). CD44 and CD69 expression indicates a T cell activation status and inducible T cell costimulator (ICOS)⁺ programmed cell death protein 1 (PD1)^low CD4⁺ T cells are T helper 1(Th1)-like effector CD4⁺ T cells that are critical for the antitumor response^19,29. Accordingly, we observed the upregulated expression of CD44 and CD69, the downregulation of PD1 and an increased ICOS⁺ PD1^low ratio in tumor-infiltrating CD4⁺ T cells from Tc9-treated mice (Fig. 3d,e and Extended Data Fig. 3o,p).

Fig. 3 |. Host CD4⁺ T cells are recruited into chimeric tumors.

a, Thy1.1⁺ Pmel-1 Tc1 or Tc9 cells were transferred into Thy1.2⁺ B6 mice bearing chimeric tumors with the adjuvant treatment. MDSCs, myeloid-derived suppressor cells. Tumor tissues were isolated and CD45⁺ cells were isolated for CyTOF analysis and a t-distributed stochastic neighbor embedding (t-SNE) display of cell populations is shown. b–e, Tumor tissues in mice treated as in a were isolated and analyzed by flow cytometry. The number of CD4⁺ T cells (b; Tc1 (n = 10 mice) and Tc9 (n = 8 mice), one of two independent experiments). The number of IFNγ⁺ or GZMB⁺ CD4⁺ T cells (c; n = 8 mice per group, one of two independent experiments). MFI of CD44 (Tc1, n = 10 mice; Tc9, n = 8 mice), CD69 (n = 8 mice per group) and PD1 (n = 8 mice per group) on CD4⁺ T cells (d; one of two independent experiments). The percentage of ICOS⁺ PD1^low CD4⁺ T cells (e; n = 8 mice per group, one of two independent experiments). Data are presented as the mean ± s.e.m. In b–e, two-tailed unpaired Student’s t-tests were used.

Next, we conducted single-cell RNA sequencing (scRNA-seq) and paired T cell receptor sequencing (scTCR-seq) analyses on tumor-infiltrating CD4⁺ T cells, identifying six distinct clusters. Among them, Th1-like cells were the dominant subset in Tc9-treated tumors (Fig. 4a). The Th1-like cells displayed heightened T cell activation and reduced exhaustion features, along with enrichment of pathways related to T helper cell differentiation and immune response in Tc9-treated tumors (Fig. 4b,c). Combining scRNA-seq and scTCR-seq analysis showed that Th1-like cells from Tc9-treated mice displayed robust oligoclonal expansion, accounting for 70.6% of total TCRs (Fig. 4d and Extended Data Fig. 4a), suggesting an emergency of host tumor-specific CD4⁺ T cell epitope spreading and immune response^31,32. The reduction in immune repertoire diversity indicates an ongoing immune response³². We then analyzed four estimates of repertoire diversity and observed a significant reduction in clonotype diversity among the Tc9-treated tumor-infiltrating Th1-like cells (Extended Data Fig. 4b). Examination of the complementarity-determining region 3 (CDR3) sequences revealed that Trac and Trbc gene expression was highly increased in the Th1-like cells following Tc9 treatment, which was accompanied by an enrichment of several TRAV–TRAJ and TRBV–TRBJ TCR pairings. The top three expanded Vα and Vβ clonotypes were Vα7D.2, Vα7.4 and Vα 14.3 and Vβ31, Vβ12.2 and Vβ5, respectively (Fig. 4e and Extended Data Fig. 4c,d). To test whether these TCRs respond to tumor antigens, tumor-infiltrating CD4⁺ T cells in Tc1-treated and Tc9-treated mice were sorted out and stimulated with B16 tumor antigens and neoantigens¹⁸. We found that CD4⁺ T cells in Tc9-treated mice showed a stronger response against B16 neoantigens such as Dag1 and Kif18b (Fig. 4f and Extended Data Fig. 4e). Consistent with the sequencing results, the numbers of Vβ5, Vβ12 and Vβ13 clonotypes were increased at the protein level from tumor-infiltrating CD4⁺ T cells of Tc9-treated mice compared to Tc1-treated ones, as measured by flow cytometry (Extended Data Fig. 4f). When we sorted out Vβ5⁺, Vβ12⁺ and Vβ13⁺ clonotype CD4⁺ T cells using fluorescence-activated cell sorting (FACS) and stimulated them with Dag1 or Kif18b, their IFNγ secretion increased (Extended Data Fig. 4g), suggesting that the increased Vβ clonotypes respond to tumor-associated antigens. In addition, these CD4⁺ T cells acquired a memory phenotype because the numbers of central memory (CD45⁺CD4⁺CD62L⁺CD44⁺) and, especially, effector memory (CD45⁺CD4⁺CD62L⁻CD44⁺) CD4⁺ T cells with cytolytic potential^33,34 were significantly increased in Tc9-treated tumors (Extended Data Fig. 4h). We then isolated these tumor-infiltrating CD4⁺ T cells, cocultured them with B16^gp100-kd tumor cells and observed that these CD4⁺ T cells displayed stronger killing effects on the antigen-loss B16 tumor cells but not control 5TGM1 myeloma cells compared to CD4⁺ T cells from Tc1-treated mice (Fig. 4g). Additionally, the cytotoxicity of these tumor-infiltrating CD4⁺ T cells was dependent on major histocompatibility complex class II (MHCII) (Fig. 4h). Moreover, although Pmel-1 Tc9 cells showed a weak killing effect against relapsed cells, they only exhibited cytotoxicity against B16 but not B16^gp100-kd or B16^gp100-ko cells (Extended Data Fig. 5a). Lastly, to assess the role of host CD4⁺ T cells in controlling tumor regression, we treated B16 and B16^gp100-kd chimeric tumor-bearing B6 mice with anti-CD4 monoclonal antibodies or isotype controls on day 11 after Tc9 or Tc1 cell transfer (Extended Data Fig. 5b,c). Depleting host CD4⁺ T cells in Tc9-treated mice led to tumor recurrence and shortened survival but had no impact on tumor growth in Tc1-treated mice (Fig. 4i and Extended Data Fig. 5d). We did not use CD4^−/− mice because of an observed increase in CD8⁺ T cell ratio in the peripheral blood of these mice (Extended Data Fig. 5e). The ability of natural killer (NK) cells to control ALV tumors is also important. However, depleting NK cells exacerbated tumor growth in both Tc1-treated and Tc9-treated mice (Extended Data Fig. 5f,g), indicating that the regulation of ALV tumor growth by NK cells is not specific to Tc9 cells. Thus, these data confirm that host effector CD4⁺ T cells have a crucial role in suppressing the outgrowth of ALV tumors in Tc9-treated mice.

Fig. 4 |. CD4⁺ T cells are required for control of chimeric tumor growth.

a–e, Tumor tissues in mice treated as in Fig. 3a were isolated and CD4⁺ T cells were isolated for scRNA-seq and scTCR-seq analyses. UMAP plot of tumor-infiltrating CD4⁺ clusters and proportion of Th1-like cells (a; n = 1 pooled from three mice), violin plots of T cell activation and exhaustion marker genes in Th1-like cells (b), enrichment analysis of genes preferentially upregulated in the Th1-like cells from Tc9 cell-treated mice (c), UMAP plot of clonal expansion in Th1-like cells (blue indicates low expansion and red indicates high expansion) and the percentage of Th1-like cells with different levels of expansion (d). Pie charts represent the cumulative frequency of TCR Vα and Vβ clonotype enrichment in the Th1-like cells, with the remaining unannotated clonotypes are presented as ‘others’ (e). f, IFNγ ELISpot analysis of CD4⁺ T cells sorted from Tc1-treated or Tc9-treated tumors (n = 3 biologically independent samples per group). g, Cytotoxicity of CD4⁺ T cells sorted from Tc1-treated or Tc9-treated tumors against target cells. Target cells were B16^gp100 cells or negative control 5TGM1 cells (n = 12 biologically independent samples per group, pooled from four independent experiments). h, Cytotoxicity of CD4⁺ T cells sorted from Tc1-treated or Tc9-treated tumors against B16^gp100-kd cells in the presence of IgG isotype or antiMHCII antibodies (n = 12 biologically independent samples per group, pooled from four independent experiments). i, PBS (n = 5 mice) or Thy1.1⁺ Pmel-1 Tc9 cells were transferred into Thy1.2⁺ B6 mice bearing chimeric (B16 and B16^gp100-kd) tumors with the adjuvant treatment and the IgG isotype (n = 6 mice) or anti-CD4 antibodies (n = 5 mice) were injected on day 20 after tumor cell inoculation. The tumor growth and survival of treated mice are shown (one of two independent experiments). Data are presented as the mean ± s.e.m. In b, a Wilcoxon rank-sum test was used. In g, a two-way ANOVA was used to compare specific lysis. In f, two-tailed unpaired Student’s t-tests were used. In h, two-tailed paired Student’s t-tests were used. In i, multiple t-tests (one per row) were used to compare tumor growth on day 45 and a log-rank (Mantel–Cox) test was performed to compare survival curves among groups.

cDC2 cells prime a tumor-specific CD4⁺ T cell response

To investigate the mechanism driving the host CD4⁺ T cell response induced by Tc9 cells in vivo, we dissected Pmel-1 Tc1-treated or Tc9-treated B16 chimeric tumors. Immunofluorescence staining showed that CD4⁺ T cells were surrounded by CD11c⁺MHC II⁺ cDCs in Tc9-treated but not Tc1-treated tumors (Fig. 5a). Flow analysis confirmed that the percentages and numbers of cDCs in Tc9-treated B16 and MC38 chimeric tumors were markedly increased compared to those in Tc1-treated mice (Fig. 5b and Extended Data Fig. 6a,b). cDCs are crucial for CD4⁺ T cell activation and be categorized into cDC1 and cDC2 subsets³⁵. We observed that more cDC2 but not cDC1 cells were detected in the tumors of Tc9-treated mice compared to Tc1-treated mice (Fig. 5c and Extended Data Fig. 6c), which is consistent with previous studies showing that cDC2 cells preferentially prime and initiate CD4⁺ T cell responses^29,35,36. Because CD4⁺ T cells are usually activated by cDCs in the tumor-draining lymph nodes (tdLNs) before infiltrating into tumors and cDCs can be divided into lymph-node-resident or migratory cells^37,38, we examined the different subsets of cDCs in tdLNs and observed that only the migratory cDC subset, especially migratory cDC2, was increased in Tc9-treated mice (Fig. 5d,e and Extended Data Fig. 6d–f), which was accompanied by increased CD4⁺ T cell activation in tdLNs (Fig. 5f). To confirm the capacity of cDC2 cells in the activation of CD4⁺ T cells, we isolated cDC2 and naive CD4⁺ T cells from tdLNs in mice treated with Tc9 cells. Results showed that cDC2 cells supported CD4⁺ T cell expansion and activation (Fig. 5g,h and Extended Data Fig. 6g). In addition, cDC2-activated CD4⁺ T cells had strong ALV tumor-killing activity (Fig. 5i). Taken together, these results indicate that cDC2 cells can activate naive CD4⁺ T cells in tdLN and these activated CD4⁺ T cells can kill ALV tumors.

Fig. 5 |. Tc9 cells recruit cDC2 cells to activate CD4⁺ T cells.

a–f, Thy1.1⁺ Pmel-1 Tc1 or Tc9 cells were transferred into Thy1.2⁺ B6 mice bearing chimeric tumors with the adjuvant treatment and tumors and tdLNs were isolated for analysis. Representative immunofluorescence sections of Tc1-treated and Tc9-treated tumors (a; scale bar, 50 μm). Percentages and numbers of cDCs in tumors (b; Tc1 (n = 9 mice) and Tc9 (n = 8 mice), one of two independent experiments). Percentages of cDC1 or cDC2 cells in cDCs and number of cDC2 cells in tumors tissues (c; Tc1 (n = 9 mice) and Tc9 (n = 8 mice), one of two independent experiments). Percentages of migratory or resident cDCs in tdLNs (d; n = 6 mice per group, one of two independent experiments). CD11b⁺ migratory cDC2 cells in CD45⁺ cells in tdLNs (e; n = 6 mice per group, one of two independent experiments). The number of CD44⁺ and CD69⁺ CD4⁺ T cells in tdLNs (f; n = 6 mice per group, one of two independent experiments). g–i, Thy1.1⁺ Pmel-1 Tc9 cells were transferred into Thy1.2⁺ B6 mice bearing chimeric tumors with the adjuvant treatment and tdLNs were isolated to sort cDC2 and naive CD4⁺ T cells. g, Sorted naive CD4⁺ T cells were labeled with CFSE and percentages of CFSE^medium CD4⁺ T cells cultured alone or cocultured with cDC2 cells are shown (n = 4 biologically independent samples per group, one of two independent experiments). h, The MFI of CD69, CD44 and IFNγ in CD4⁺ T cells cultured alone or cocultured with cDC2 cells (n = 3 biologically independent samples per group, one of two independent experiments). i, Specific lysis of CD4⁺ T cells cultured alone or cocultured with cDC2 cells against B16^gp100-kd cells, T cells alone (n = 4 biologically independent samples) or T cells + cDC2 cells (n = 5 biologically independent samples), pooled from two independent experiments. Data are presented as the mean ± s.e.m. In b–i, two-tailed unpaired Student’s t-tests were used.

To assess the role of cDC2 in the induction of host CD4⁺ T cell responses in vivo, Itgax^creIrf4^flox/flox (referred to as Cre-positive hereafter) mice deficient in cDC2 were used^39,40. We injected B16 and B16^gp100-kd chimeric tumor cells into Itgax Cre-positive or Cre-negative (as a control) Irf4^flox/flox B6 mice, followed by transfer of Pmel-1 Tc9 cells and adjuvant therapy (Extended Data Fig. 6h). Consistent with a previous study³⁹, we observed that Cre-positive mice had fewer CD11b⁺ migratory cDC2 cells and reduced CD4⁺ T cell infiltration in tdLNs compared to Cre-negative mice, whereas the number of cDC1 cells in the tdLN remained unchanged (Fig. 6a,b). As a result, the numbers of total and IFNγ⁺ or GZMB⁺ tumor-infiltrating CD4⁺ T cells were significantly decreased and their expression of CD44 and CD69 was reduced in Tc9-transferred Cre-positive mice compared to Cre-negative mice (Fig. 6c–e). Consequently, Tc9-treated Cre-positive mice displayed poorer survival than Cre-negative mice (Fig. 6f).

Fig. 6 |. cDC2 cells are indispensable for host CD4⁺ T cell activation.

a–f, Pmel-1 Tc9 cells were transferred into Itgax Cre^−/−Irf4^flox/flox or Itgax Cre^+/−Irf4^flox/flox B6 mice bearing chimeric tumors with the adjuvant treatment and tumors and tdLNs were isolated for analysis. Percentages and numbers of CD11b⁺ migratory cDC2 and CD103⁺ migratory cDC1 cells in tdLN (a; n = 5 mice per group, one of two independent experiments). Numbers of CD4⁺ T cells in tdLN (b). The number of total CD4⁺ T cells (c). The number of IFNγ⁺ and GZMB⁺ CD4⁺ T cells (d). MFI of CD44 and CD69 on CD4⁺ T cells in tumors (e). In b–e, Cre^−/− (n = 5 mice) and Cre^+/− (n = 4 mice) were used and data represent one of two independent experiments. Survival curves of mice (f; Cre^−/− PBS mice (n = 5 mice), Cre^+/− PBS (n = 5 mice), Cre^−/− Tc9 (n = 10 mice) and Cre^+/− Tc9 (n = 7 mice), one of two independent experiments. g–m, Pmel-1 Tc9 cells were transferred into WT or Ccr7^−/− B6 mice bearing chimeric tumors with the adjuvant treatment and tumors and tdLNs were isolated for analysis. Percentages and numbers of CD11b⁺ migratory cDC2 cells in tdLNs (g). Percentages and numbers of CD4⁺ T cells in tdLNs (h). The number of total CD4⁺ T cells in tumors (i). IFNγ⁺ and GZMB⁺ CD4⁺ T cells in tumors (j). MFI of CD44 (k) and CD69 (l) in CD4⁺ T cells in tumors. In g–i,k, n = 9 mice were used per group, pooled from two independent experiments. In j,l, n = 6 mice were used per group and data represent one of two independent experiments. m, Tumor growth curves of mice (n = 5 mice per group, one of two independent experiments). Data are presented as the mean ± s.e.m. In a–e,g–l, two-tailed unpaired Student’s t-tests were used. In f, a log-rank (Mantel–Cox) test was performed to compare survival curves among groups. In m, multiple t-tests (one per row) were used to compare tumor growth on day 40.

CCR7 has an important role in DC migration from tumor tissues to the tdLN to activate CD4⁺ T cells²⁹. We found that the tumor and tdLN cDC2 cells expressed CCR7 (Extended Data Fig. 6i,j). To test whether cDC2 migration was required for CD4⁺ T cell infiltration and activation in Tc9-treated ALV-containing chimeric tumors, mice deficient in Ccr7 were used (Extended Data Fig. 6k). Relative to WT mice, the number of migratory cDC2 and CD4⁺ T cells in the tdLN was notably lower in Ccr7^−/− mice (Fig. 6g,h). We observed that the numbers of total, IFNγ⁺ or GZMB⁺ CD4⁺ T cells were decreased and the expression of CD69 and CD44 on CD4⁺ T cells was reduced in tumor tissues of Tc9-treated Ccr7^−/− mice (Fig. 6i–l). Consequently, Tc9 cells, although still highly therapeutic against the inoculated tumors, were unable to suppress the outgrowth of ALV-containing chimeric tumors in Ccr7^−/− mice (Fig. 6m). Collectively, these findings indicate that host CD4⁺ T cell activation in Tc9-treated mice is achieved by CCR7-dependent recruitment of cDCs, especially migratory cDC2 cells in vivo.

IL-24 recruits cDC2 cells to activate host CD4⁺ T cells

As cDC2 recruitment is critical for the accumulation and activation of host effector CD4⁺ T cells in Tc9-treated mice and Tc9 cells are characterized by secretion of IL-9, we first determined whether signaling of IL-9 and its receptor is required for the recruitment of cDC2 cells. Our results showed that host deficiency in the IL-9 receptor (Il9r^−/−) did not abrogate the infiltration of cDC2 cells into tdLNs and tumors (Extended Data Fig. 7a,b). We then investigated whether other Tc9 cell-secreted cytokines and chemokines may have roles in promoting cDC2 migration and priming the CD4⁺ T cell response in vivo. Our microarray analysis identified Il24 as one of the top ten genes preferentially expressed by Tc9 cells relative to Tc1 cells (Fig. 7a). Flow analysis and ELISA confirmed that tumor-infiltrating Tc9 cells exhibited an increased expression of IL-24 compared to Tc1 cells (Fig. 7b and Extended Data Fig. 7c), which resulted in an elevated IL-24 level in the TME (Fig. 7c). IL-24 is a newly identified member of the expanding IL-10 gene family⁴¹ and its function in DC migration is largely unknown. To determine whether IL-24 affects DC migration, we added recombinant IL-24 in transwell cultures and observed a dose-dependent increase in cDC2 migration (Fig. 7d and Extended Data Fig. 7d). Similarly, supernatants from Tc9-treated but not Tc1-treated tumors promoted cDC2 migration (Fig. 7e). To confirm whether Tc9-derived IL-24 was necessary for cDC2 recruitment into tumors to prime CD4⁺ T cells in vivo, we knocked down Il24 in Tc9 cells and injected vector or Il24-KD Tc9 cells into mice bearing B16 and B16^gp100-kd chimeric tumors (Extended Data Fig. 7e–h). Il24 KD in Tc9 cells resulted in markedly reduced cDC2 accumulation in tumors and fewer migratory cDC2 cells in tdLNs (Fig. 7f,g). Consequently, the numbers of CD4⁺ T cells in the tdLNs and tumors were significantly decreased in Il24-KD Tc9-treated mice (Fig. 7h,i). In addition, the numbers of IFNγ⁺ or GZMB⁺ CD4⁺ T cells and the expression of CD44 and CD69 on CD4⁺ T cells decreased in tumor tissues (Fig. 7j,k). Consequently, Il24-KD Tc9 cells had a compromised ability to suppress the outgrowth of ALV-containing chimeric tumors, leading to shortened mouse survival compared to mice treated with vector-control Tc9 cells (Fig. 7l). However, KD of IL-24 in Tc1 cells had no effect on the numbers of cDC2 and CD4⁺ T cells in tdLN and tumors (Extended Data Fig. 7i–l), which might be attributed to relatively low secretion of IL-24 by Tc1 cells and their poor persistence, resulting in only slight changes in IL-24 levels in the TME (Extended Data Fig. 7m,n). Consequently, knocking down IL-24 in Tc1 cells did not impact the therapeutic efficacy of the cells against ALV tumors (Extended Data Fig. 7o). These findings indicate that Tc9 cell-derived IL-24 promotes cDC2 migration into tumors to activate a host CD4⁺ T cell immune response to ALV tumors.

Fig. 7 |. IL-24 promotes cDC2 migration to activate host CD4⁺ T cells.

a, Pmel-1 CD8⁺ T cells were polarized under related conditions. Microarray analysis showing the top ten genes enriched in Tc9 cells compared to Tc1 cells on day 5 of polarization (n = 1 per group). b,c, Thy1.1⁺ Pmel-1 Tc1 or Tc9 cells were transferred into Thy1.2⁺ B6 mice bearing B16 chimeric tumors with the adjuvant treatment. Flow analysis of IL-24 expression in tumor-infiltrating Tc1 or Tc9 cells (b; n = 5 mice per group, one of two independent experiments). ELISA detection of IL-24 in Tc1-treated or Tc9-treated TME (c; n = 12 mice per group, pooled from two independent experiments). d,e, Migration of cDC2 in response to different doses of IL-24 (d; n = 5 biologically independent samples per group, one of two independent experiments) or supernatants of Tc9-treated or Tc1-treated tumors (e; Tc1 (n = 6 biologically independent samples) and Tc9 (n = 5 biologically independent samples), one of two independent experiments). f–l, Thy1.1⁺ Pmel-1 Tc9 cells transduced with vector or Il24 shRNA were transferred into B6 mice bearing chimeric tumors with the adjuvant treatment. Tumors and tdLNs were isolated for analysis. The number of cDC2 cells in tumors (f). The number of CD11b⁺ migratory cDC2 cells in tdLN (g). The number of CD4⁺ T cells in tdLN (h). The number of total CD4⁺ T cells in tumors (i). The number of IFNγ⁺ and GZMB⁺ CD4⁺ T cells in tumors (j). MFI of CD44 and CD69 on CD4⁺ T cells in tumors (k). In f–k, n = 8 mice were used per group, pooled from two independent experiments. l, Tumor growth curves and survival plots (n = 5 mice per group, one of two independent experiments) of treated mice. Data are presented as the mean ± s.e.m. In b,c,e–k, two-tailed unpaired Student’s t-tests were used. In d, a one-way ANOVA was used. In l, multiple t-tests (one per row) were used to compare tumor growth on day 45 and a log-rank (Mantel–Cox) test was performed to compare survival curves.

IL24, cDC2 cells and effector CD4⁺ T cell genes in persons with cancer

To assess the clinical importance of the findings, we determined the expression of IL24, cDC2 and effector CD4⁺ T cells in cancers. By analyzing the published data at Oncomine and The Cancer Genome Atlas (TCGA) database^42,43, we first observed that the expression of IL24 was strongly and positively correlated with patient prognosis. High IL24 expression in tumors was associated with better overall survival compared to low expression in persons with melanoma or breast cancer (Fig. 8a and Extended Data Fig. 8a). Second, we examined whether cDC2 accumulation correlates with the levels of IL24 expression. To determine the gene signature best suited for identifying cDC2 cells, we analyzed a public dataset for DC populations in various human tissues⁴⁴ and identified a human cDC2 gene signature that includes FCRE1A, CD1C, CLEC10A and CD1E, which is consistent with previous studies^29,45. We excluded LGALS2, AREG, CFP, NDRG2, JAML and LTB, which have previously been reported to be expressed in cDC2 cells²⁹ but displayed promiscuous expression in our analysis (Fig. 8b). The cDC2 gene signature displayed a positive correlation with IL24 expression in persons with melanoma or breast cancer (Fig. 8c and Extended Data Fig. 8b). Importantly, the high expression of cDC2 signature genes was significantly positively correlated with improved survival and CD1C, a cDC2 marker, was a reliable prognostic biomarker for patient survival in both human cancers (Fig. 8d–f and Extended Data Fig. 8c–e). In addition, the stratification of persons with cancer by the expression of host effector CD4⁺ T cell-associated genes revealed positive associations between the high expression of effector CD4⁺ T cell signature genes in tumors and improved patient survival (Fig. 8g,h and Extended Data Fig. 8f–h). Lastly, the host effector CD4⁺ T cell gene signature displayed a significant positive correlation with both the cDC2 gene signature and IL24 expression (Fig. 8i and Extended Data Fig. 8i). Altogether, these results indicate a strong positive correlation between IL24 or the cDC2 gene signature and the effector CD4⁺ T cell gene signature in melanoma and breast cancer samples. Persons with higher IL24 expression, cDC2 cells and effector CD4⁺ T cell gene signatures have better prognosis and survival.

Fig. 8 |. IL24, cDC2 cells and effector CD4⁺ T cell genes in human melanoma.

a, Prognostic value of IL24 expression for overall survival of persons with melanoma comparing high versus low quartiles (n = 20 per group). b, Identification of cDC2-specific genes in human DC subsets based on global gene expression data (GSE77671; n = 37). c, Correlation of IL24 and cDC2 gene signature in persons with melanoma (n = 83). d, Heat map (n = 83) of ordered, z-transformed expression values of cDC2-specific genes and expression of cDC2 signature genes for top and bottom quartiles (n = 20) in persons with melanoma. e,f, Prognostic value of cDC2 gene signature (e) or CD1C expression (f) for overall survival in persons with melanoma comparing high versus low quartiles (n = 20 per group). g, Heat map (n = 83) of ordered, z-transformed expression values of effector CD4⁺ T cell-specific genes and expression of effector CD4⁺ T cell-specific signature genes for top and bottom quartiles (n = 20) in persons with melanoma. h, Prognostic value of effector CD4⁺ T cell signature for overall survival in persons with melanoma comparing high versus low quartiles (n = 20 per group). i, Correlation of effector CD4⁺ T cell signature with IL24 or with cDC2 gene signature in persons with melanoma (n = 83). Data are presented as the mean ± s.e.m. In a,e,f,h, a log-rank (Mantel–Cox) test was performed to compare survival curves. In c,i, linear regression was used.

Discussion

ACT is a promising and effective approach for cancer therapy; however, its therapeutic benefits are typically short-lived, primarily because of the inability of transferred tumor-specific T cells to recognize and kill emerged ALV tumors within the suppressive TME. To overcome this challenge, researchers have explored various strategies, including developing new T cell subsets such as Th9 cells to promote monocyte infiltration⁴⁶, identifying and using neoepitope-reactive CD8⁺ T cells^47,48 and engineering chimeric antigen receptor (CAR) T cells by overexpressing cytokines to activate macrophages and DCs^38,49. Although these approaches mitigate ALV outgrowth through different mechanisms, whether transferred T cells can activate host CD4⁺ T cells to suppress ALV tumors is largely unknown. Here, we describe a key role for host effector CD4⁺ T cells in suppressing the growth of ALV-containing tumors in Tc9 cell-treated mice (Extended Data Fig. 9). Tc9 cell-derived IL-24 attracts cDC2 cells to the tumor site, which then migrate into tdLN by CCR7 to prime host CD4⁺ T cells and elicit a strong effector CD4⁺ T cell response in tumors. Depletion of host CD4⁺ T cells or disruption of the Il24–cDC2 axis diminished the antitumor efficacy of Tc9 cells. We also showed that, similar to the murine cancer models, an IL24–cDC2–host effector CD4⁺ T cell axis exists in human melanoma and breast cancers. Moreover, the expression of IL24 and infiltration of cDC2 and effector CD4⁺ T cells were positively correlated with survival. These findings illuminate the superior antitumor ability of Tc9 cells in the scenario of tumor heterogeneity and emphasize the critical role of host effector CD4⁺ T cell activation in suppressing tumor relapses.

Emerging evidence indicates that host effector CD4⁺ T cells are crucial for the antitumor response. In myeloma and melanoma models, naive CD4⁺ T cells are activated in tdLNs, migrate to tumors and become effector T cells to drive the protective antitumor immune response^29,50. Notably, ICOS⁺ PD1^low CD4⁺ effector T cells generated in CTLA4 therapy are critical for an effective antitumor response¹⁹. Moreover, CD4⁺ T cells can specifically recognize tumor neoantigens and secrete IFNγ against mutant tumors^18,51. Lastly, recent single-cell sequencing analysis in human bladder cancer showed that intratumoral CD4⁺ T cells exert antitumor activity and directly eliminate autologous tumors³⁰. Our findings propose a paradigm for the induction of host effector CD4⁺ T cells by transferred tumor-specific Tc9 cells. Emerged effector CD4⁺ T cells then secrete cytotoxic cytokines IFNγ and GZMB and work together with long persistent Tc9 cells to mitigate the outgrowth of ALV tumors. Additionally, scRNA-seq and scTCR-seq analyses displayed that Th1-like cells were the predominant subset among the CD4⁺ T cells infiltrating Tc9-treated tumors. These Th1-like cells displayed a robust clonal expansion and reduced clonotype diversity, suggesting an enhanced CD4⁺ T cell-specific immune response. While we could not confirm whether the top three expanded Vα and Vβ clonotypes specifically respond to tumor-associated antigens and the nature of this response remains to be investigated, our data show that CD4⁺ T cells in Tc9-treated tumors exhibit strong response against certain B16 neoantigens.

Consistent with a previous study²⁹, our results showed that host effector CD4⁺ T cells were primed in tdLNs by cDC2 cells before infiltrating the tumor. cDC2 cells are divided into lymph-node-resident and migratory subsets and CD11b⁺ migratory cDC2 cells can migrate from the tumor into tdLNs through CCR7 to induce T cell activation^29,52. Similarly, others have reported an enhanced cDC2 functionality in the absence of cDC1 cells or in tumor resistance to cytotoxic T cell, with cDC2 migration crucial for activating CD4⁺ T cells to control tumor growth⁵³. In addition, it has been reported that cDC2 cells isolated from tdLNs can activate naive CD4⁺ T cells without additional antigen loading in vitro, indicating that cDC2 cells already capture tumor antigens in the tdLN^29,52,54. However, the importance of cDC2 activation and migration warrants further investigation. In our study, although we observed that cDC2 cells from Tc9-treated tdLNs can activate naive CD4⁺ T cells in vitro and because the frequency of tumor antigen-specific T cells in tdLNs is low, activation of some tumor-nonspecific T cells may have occurred. A comparison between CFSE-labeled T cells from tumor-bearing and non-tumor-bearing mice may be insightful. Furthermore, we confirmed that the selective reduction of cDC2 in Itgax^creIrf4^flox/flox mice or blockade of cDC2 migration in Ccr7^−/− mice impaired CD4⁺ T cell activation and antitumor ability. Although we could not preclude the potential Cre-mediated recombination in other cells in Itgax^creIrf4^flox/flox mice, our data do show a potent activation link between cDC2 cells and CD4⁺ T cells. Taken together with previous studies showing that CCR7 is critical for cDC1 migration and priming antitumor CD8⁺ T cells^55,56, these findings demonstrated that CCR7 is essential for cDC1 and cDC2 migration, which is crucial for tumor suppression. However, the deficiency of CCR7 did not completely abolish the migration of cDC2 cells and tumor control in Tc9-treated mice, suggesting that other mechanisms associated with cDC2 recruitment to tdLNs and Tc9-mediated tumor suppression may exist and require further investigation.

Tc9 cells represent a recently identified T cell subtype that exhibits cytolytic activity like classical cytolytic Tc1 cells after transfer but persists substantially longer than Tc1 cells in vivo^23,57. We previously demonstrate that the extended longevity of Tc9 cells enables them to have greater antitumor responses against antigen-specific tumor cells as more IFNγ⁺ and GZMB⁺ transferred T cells exist in the tumor tissues of Tc9-treated mice than those of Tc1-treated littermates^22,23,28. In this study, we propose another new property for Tc9 cells. We observed that Tc9 cell-derived IL-24 can promote cDC2 migration into tumors, which then uptake tumor antigens and activate host effector CD4⁺ T cells to suppress the progression of antigen-loss tumors. Collectively, the combined suppression of both antigen-specific and antigen-loss tumor cells by Tc9 and induced host CD4⁺ T cells endows Tc9 cells with superior antitumor ability. Interestingly, IL-24-expressing CAR T cells have been reported to exhibit enhanced antitumor ability, supporting the importance of IL-24 (refs. 58,59). However, the specific pathway underlying how IL-24 attracts cDC2 cells remains to be elucidated.

In persons with melanoma and breast cancer, we observed a phenomenon consistent with murine data, where IL24 is positively correlated with cDC2 and effector CD4⁺ T cell gene signatures. Importantly, expression levels of IL24 were strongly and positively correlated with patient prognosis. In line with this finding, higher expression of both cDC2 and effector CD4⁺ T cell signature genes was also significantly associated with better patient survival. Consistent with our results, others also reported that cDC2 abundance can be a biomarker for effector CD4⁺ T cell quality and responsiveness of immune checkpoint blockade therapy in human cancers²⁹. In persons with breast cancer, Stephen et al. observed that cDC2 gene signatures are positively associated with enhanced survival in tumors⁵³. Thus, these observations provide a rationale for exploiting human tumor-specific Tc9 cells to treat human cancers. As our current experiments primarily focused on B16 melanoma and MC38 colon cancer models, further studies will be needed to investigate whether Tc9-derived IL-24 can recruit cDC2 to induce the CD4⁺ T cell response, thereby controlling the outgrowth of ALV in other cancer types and cancers at different organ sites.

Taken together, our work reports a mechanism of Tc9 cells in suppressing ALV-containing tumor growth in the TME by inducing a host effector CD4⁺ T cell response through the IL24–DC circuit. This suggests that activating host effector CD4⁺ T cells could be a promising approach to enhance the antitumor efficacy of ACT for human cancers.

Methods

Ethics statement

All experiments complied with protocols approved by the Institutional Animal Care and Use Committee (IACUC) of the Houston Methodist Research Institute.

Mice

Pmel-1, OT-I, CD45.1, Itgax Cre, Irf4^flox/flox, Ccr7^−/−, Cd4^−/−, Cd8^−/− and B6 mice were purchased from the Jackson Laboratories. Il9r^−/− mice were generated by inserting a neomycin resistance gene at nucleotide position 2826 (ref. 60). CD45.1⁺ OT-I mice were generated by crossing CD45.1 mice with OT-I mice. Itgax Cre^+/− or Cre^−/− Irf4^flox/flox were produced by crossing Itgax Cre^+/− mice with Irf4^flox/flox mice. We used 8–12-week-old male and female mice for animal experiments. Mice were kept in a clean facility with a 12-h light–dark cycle, an ambient temperature of 65–75 °F and a humidity level of 40–60%. All detailed information on mouse strains, reagents, oligonucleotide sequences and antibodies is provided in Supplementary Table 1.

Cell purification and culture

Murine B16 (CRL6475) and MC38 (SCC172) cell lines were purchased from the American Type Culture Collection and Sigma-Aldrich, respectively. B16^OVA and B16^gp100-kd cells were obtained by transducing B16 with pUltra plasmid encoding OVA or pLKO.1 plasmid encoding Pmel short hairpin RNA (shRNA), respectively. B16, B16^OVA and B16^gp100-kd cells were cultured in complete DMEM medium (10% FBS,100 U per ml penicillin–streptomycin and 2 mM l-glutamine). Naive CD8⁺ T cells were isolated using an EasySep isolation kit from STEMCELL Technologies and primed with anti-CD3 (2 μg ml⁻¹) and anti-CD28 (1 μg ml⁻¹) antibodies under Tc1 (10 ng ml⁻¹ IL-2) or Tc9 (10 ng ml⁻¹ IL-4, 1 ng ml⁻¹ transforming growth factor-β and 20 μg ml⁻¹ anti-IFNγ monoclonal antibody) polarizing conditions for 3 days. The cells were then expanded in the polarizing medium with 10 ng ml⁻¹ IL-2 for an additional 2 days. In certain experiments, Pmel-1 splenocytes were activated using Hgp100_25–33 peptide (1 μg ml⁻¹; Genscript) under polarizing medium. Central memory or stem memory T cells were cultured with Hgp100_25–33 in either recombinant human (rh)IL-15 (10 ng ml⁻¹) or rhIL-2 (10 ng ml⁻¹) with TWS119 (7 μM). Then, 5 days after antigenic stimulation, Tc1, Tc9, central memory T or stem memory cells were adoptively transferred into mice bearing chimeric B16 and B16^gp100-kd tumors in combination with adjuvant treatment.

Viral production and T cell transduction

Viruses were packaged in 293T cells using Lipofectamine 3000 (ref. 22). Viral supernatants were isolated 48 h after plasmid transfection and subsequently filtered. For Pmel (gp100) KD, filtered viral supernatants were added to B16 cells with protamine sulfate (10 μg ml⁻¹) for 16 h, followed by selection with puromycin and subsequent culturing for further analysis. For Il24 KD, after an initial activation of 18–24 h, Tc9 cells were transduced with Il24 KD viral supernatant for 16 h, with centrifugation at 500g for 2 h at 32 °C, followed by 3-day culture and subsequent selection of GFP⁺ cells for adoptive transfer experiments. For Pmel KO, two single guide RNAs targeting regions around the Pmel start codon were used. B16 cells were transduced with the pX330 virus supernatants encoding single guide RNAs and Cas9 (ref. 61). After sorting GFP⁺ cells, the positive cells were replated at a low dilution to generate single-cell-derived clones. Western blotting was used to select completely knocked-out single-cell clones.

Construction and detection of B16^OVA and MC38^gp100 cells

For OVA overexpression, complementary DNA (cDNA) encoding OVAL was obtained from GenScript and subsequently subcloned into the pUltra vector. For gp100 overexpression, the complete human Pmel gene was extracted from the WRG-gp100 vector, which was generously provided by N. Restifo, and subsequently inserted into the pLV414G. The constructions were verified by DNA sequencing. To assess OVA–GFP expression in cultured B16^OVA cells, the cells were harvested and GFP mean fluorescence intensity (MFI) was measured. For detection of OVA(SIINFEKL) expression in cultured B16^OVA cells, cells were stained with OVA_257–264(SIINFEKL)-H-2K^b or isotype antibodies. For detection of gp100 expression in cultured B16 or MC38^gp100 cells, cells were intracellularly stained with gp100 or isotype antibodies overnight by BD Cytofix/Cytoperm fixation and permeabilization solution, followed by staining with secondary antibodies at room temperature for 45 min. For detection of OVA–GFP, OVA(SIINFEKL) or gp100 expression in ALV tumors, tumor tissues were collected and mashed in ice-cold FACS buffer. Dissociated tissues were then filtered through 70-μm filter meshes and collected for Fc block. For detection of OVA–GFP or OVA(SIINFEKL), blocked single-cell suspensions were stained with CD45 and OVA_257–264(SIINFEKL)-H-2Kb or isotype antibodies. For detection of gp100, blocked cells were stained with CD45 antibody and then intracellularly stained with gp100 or isotype antibodies overnight, followed by staining with secondary antibodies. Samples were resuspended in FACS buffer for flow cytometry analysis.

Real-time PCR

RNA of cells was extracted using the RNeasy Mini kit and reverse-transcribed into cDNA using a high-capacity cDNA reverse transcription kit. Real-time PCR was performed with SYBR select master mix. The mouse housekeeping gene β-actin was used to normalize gene expression. The primers are listed in the Supplementary Table 1.

Tumor models and adoptive transfer

For subcutaneous primary tumor models, male and female B6 mice were inoculated subcutaneously (s.c.) in the right back flank with 6 × 10⁵ B16 cells, 6 × 10⁵ B16^OVA cells or 1 × 10⁶ MC38^gp100 tumor cells (Fig. 1 and Extended Data Figs. 1 and 2). For the B16^OVA and WT B16 chimeric tumor model in Fig. 1e, mice were inoculated s.c. in the right back flank with 2 × 10⁵ WT B16 cells with or without 4 × 10⁵ B16^OVA tumor cells. For WT B16 and B16^gp100-kd chimeric tumor models in Figs. 1f and 3–7, mice were inoculated s.c. in the right back flank with 2 × 10⁵ B16^gp100-kd cells with or without 4 × 10⁵ WT B16 tumor cells. For the WT MC38 and MC38^gp100 chimeric tumor model in Fig. 1g, mice were inoculated s.c. in the right back flank with 8 × 10⁵ WT MC38 cells and 2 × 10⁵ MC38^gp100 tumor cells. On day 8, one dose of cyclophosphamide (CTX) was given intraperitoneally (i.p.) at 200 mg kg⁻¹ body weight. On day 9, 2 × 10⁶ indicated CD8⁺ T cells, 5 × 10⁵ peptide-pulsed BM-derived DCs and four doses of rhIL-2 (6 × 10⁵ U) were intravenously (i.v.) injected²³. For antigen-loss tumor cell rechallenge models in Fig. 2, primary tumor-bearing mice received T cell adoptive transfer with the adjuvant treatment during days 8–12. On day 20, the same mice were rechallenged with antigen-loss tumor cells on the contralateral flank. Naive B6 mice that were inoculated with antigen-loss tumor cells and received adjuvant therapy without T cell transfer were used as a negative control. All recipient mice of Tc1 and Tc9 cell transfer in the study received the adjuvant treatment (CTX, DC and rhIL-2) and a single injection of either Tc1 or Tc9 cells. Tumor size was calculated as follows: length × width²/2. Mice were killed following protocols approved by the IACUC when the length of the tumors exceeded 2 cm or ulceration exceeded 0.5 cm. Tumors were measured every 2–3 days. In some cases, the limit (tumor length of 2 cm) exceeded the last day of measurement; tumor size was recorded and the mice were immediately killed. Tumor tissues and tdLNs were collected for further detection. The absolute number of T cells or DCs in tumor tissues (100 mg) and tdLNs (per lymph node) was determined using CountBright absolute counting beads.

In vivo CD4⁺ T or NK cell depletion

For depletion of CD4⁺ T cells or NK cells, 250 μg of CD4-depleting antibodies (YTS191.1) or 200 μg of NK-depleting antibodies (PK136) were injected i.p. on day 20 after tumor transplantation and every 3 days thereafter for a total of four injections.

Tumor supernatant preparation

Pmel-1 Thy1.1⁺ Tc1 or Tc9 cells were transferred into B6 mice inoculated with 4 × 10⁵ B16 and 2 × 10⁵ B16^gp100-kd chimeric tumor cells in combination with the adjuvant treatment (CTX, DC and rhIL-2). Then, 30 days after tumor inoculation, tumors were isolated and tumor supernatants were collected⁶². Briefly, tumor tissues were collected, weighed, minced and incubated in DMEM overnight. Conditioned media were collected and filtered for IL-24 detection or DC migration.

ELISA

For measuring IL-24 levels in the TME, tumor supernatants were collected as mentioned above and tested by ELISA using the mouse IL-24 DuoSet ELISA kit according to the manufacturer’s protocol.

To measure the Vβ⁺ CD4⁺ T cell response to indicated peptides, Vβ5⁺, Vβ12⁺ and Vβ13⁺ tumor-infiltrating CD4⁺ T cells were sorted from Tc9 cell-treated mice using a BD FACSAria Fusion and stimulated with splenocytes isolated from naive mice, which were pulsed with 5 μg ml⁻¹ synthetic peptide in a 96-well V-bottom plate. Cells were stimulated overnight and supernatants were collected and tested using an ELISA MAX deluxe set mouse IFNγ kit.

Generation of bone-marrow-derived DCs and DC migration assays

Mouse femurs and tibias from 6–12-week-old mice were cut at the ends and flushed with RPMI medium (2% FBS)⁶³. Flushed bone marrow cells were filtered and collected by centrifugation. The pellets were resuspended in ACK lysis buffer to remove red blood cells. The cells were then seeded in a T75 flask at a density of 0.5 million cells per ml in complete RPMI medium (10% FBS, 100 U per ml of penicillin–streptomycin, 2 mM l-glutamine and 50 μM 2-mercaptoethanol) with 20 ng ml⁻¹ recombinant mouse granulocyte macrophage colony-stimulating factor (rmGM-CSF) and unattached cells were removed by replacing the medium on the next day. Then, 4 days later, half of the medium was refreshed with RPMI containing rmGM-CSF and incubated for another 4 days. After 8 days, immature DCs were pulsed for 8 h with OVA or Hgp100 (1 μg ml⁻¹) in the presence of tumor necrosis factor (10 ng ml⁻¹) and IL-1β (10 ng ml⁻¹) for 2 days to induce DC maturation.

On day 10, mature DCs were transferred to mice as an adjuvant or cDC2 cells were sorted using a BD FACSAria Fusion. Their chemotaxis was then analyzed using a transwell migration assay as follows: 5 × 10⁵ cDC2 cells were suspended in 200 μl of RPMI medium (1% BSA) and seeded into transwell inserts (5-μm pore size) in a 24-well plate containing 500 μl of RPMI medium (1% BSA) ± rmIL-24 (25 or 50 ng ml⁻¹) or tumor supernatant⁶⁴. After a 2-h incubation, cells in the lower compartment were quantified by flow cytometry and normalized by counting beads.

Coculture of cDC2 cells and CD4⁺ T cells

cDC2 cells and naive (CD62L⁺CD44⁻) CD4⁺ T cells from tdLNs of mice treated with Tc9 cells were isolated using a BD FACSAria Fusion. Sorted naive CD4⁺ T cells were labeled with CFSE and were either cultured alone or cocultured with sorted cDC2 cells at a 5:1 ratio in complete RPMI medium in 96-well V-bottom plates. Then, 7 days later, CD4⁺ T cells were harvested to determine the percentage of CFSE^medium cells. In other experiments, sorted CD4⁺ T cells were cultured alone or cocultured with sorted cDC2 cells at a 5:1 ratio for 2 days and activation of CD4⁺ T cells was detected. Alternatively, sorted naive CD4⁺ T cells were cultured alone or cocultured with sorted cDC2 cells at a 5:1 ratio in 48-well plates for 5 days and specific lysis of CD4⁺ T cells was detected.

Immunofluorescence imaging

Tumors were fixed in 2% paraformaldehyde for 2 h, then dehydrated in 30% sucrose overnight for cryoprotection, subsequently embedded in TissueTek OCT freezing medium and stored at −80 °C. Sample sections were processed in the Houston Methodist Research Institute Pathology core. The following antibodies were used, along with DAPI: anti-CD11c, anti-CD4 and anti-I-A/I-E (MHCII). Stained sections were mounted in type F immersion liquid and analyzed using a Leica TCS SPE confocal microscope (Leica).

CyTOF

Tumors from six mice in each group were minced in cold RPMI medium, dissociated and digested using a tumor dissociation kit (Miltenyi Biotec, 130–096-730) on a GentleMACS Octo dissociator. Filtered cells were purified by Percoll gradient centrifugation to obtain leukocytes⁶⁵ and enriched using a CD45 microbead kit (Miltenyi Biotec, 130–052-301). The obtained single-cell suspension was sent to the Houston Methodist Research Institute ImmunoMonitoring Core for blocking, staining, washing and CyTOF detection. After CyTOF acquisition, raw measurement data (FCS files) and all metadata were preprocessed for concatenation, normalization and compensation using CATALYST⁶⁶ in R software⁶⁷. Dead cells were manually removed using FlowJo. Cell clustering was conducted using R packages including ‘FlowSOM’, ‘ConsensusClusterPlus’, ‘CATALYST’ and ‘Rtsne’^68,69. The transformation of marker expression data was performed using the ‘prepData()’ SCE constructor arcsinh(x) function with a cofactor of 5 (ref. 70). The cluster annotation and merging were conducted on the basis of marker expression heat maps in cell populations.

Gene microarray

Pmel-1 CD8⁺ T cells were polarized under the specified conditions and harvested on day 5 of polarization. RNA was extracted using TRIzol and sent to the Gene Expression and Genotyping Facility at Case Western Reserve University for microarray analysis using the Affymetrix WT Plus expression platform.

Flow cytometry

For surface antibody staining, single-cell suspensions were incubated with Fc blocker on ice for 15 min, followed by surface antibody staining on ice for 15–30 min in FACS buffer. For intracellular cytokine staining, cells were stimulated with PMA and ionomycin in the presence of brefeldin A for 4 h. Stimulated cells were then stained with surface antibodies, followed by fixation and permeabilization for intracellular cytokine staining. Results were acquired using BD LSRFortessa X30 systems and FACSDiva software. Data were analyzed with FlowJo software.

Western blot assay

Cells were collected and lysed in lysis buffer with protease inhibitors. The concentrations of protein samples were measured by Bradford protein assay. Ten μg protein samples were loaded to NuPAGE Bis–Tris protein gels and transferred to nitrocellulose membranes⁷¹. Nitrocellulose membranes were then incubated with primary antibody at 4 °C overnight and horseradish-peroxidase-conjugated secondary antibodies at 37 °C for 1 h. Primary antibodies were melanoma anti-gp100 antibody from Abcam, anti-ovalbumin antibody from Abcam, anti-GAPDH antibody from Santa Cruz Biotechnology and anti-β-actin monoclonal antibodies from Cell Signaling Technology.

Cytotoxicity assay

The cytotoxicity of tumor-infiltrating CD4⁺ T cells was measured using the CytoTox 96 nonradioactive cytotoxicity assay according to the manufacturer’s protocol⁷². Briefly, Pmel-1 Tc9 cells or tumor-infiltrating CD4⁺ T cells from Tc1-treated and Tc9-treated mice were isolated using a BD FACSAria Fusion and cocultured with target cells at a 1:10 (tumor:T cell) ratio in RPMI medium (5% FBS) in a 96-well V-bottom plate for 6 h. Supernatants were collected and measured at 492 nm. Specific lysis was calculated using the following formula: (experimental − effector spontaneous − target spontaneous)/(target maximum − target spontaneous) × 100.

ELISpot assay

CD4⁺ T cells from tumor tissues were enriched using Miltenyi mouse CD4⁺ kits according to the manufacturer’s protocols. A total of 5 × 10⁴ tumor-infiltrating T cells were stimulated with 5 × 10⁵ splenocytes isolated from naive mice pulsed with 5 μg ml⁻¹ synthetic peptide. Cells were cultured overnight in an anti-mouse IFNγ-coated ELISpot plate. The plate was processed according to the manufacturer’s introduction. Spots were quantified using a CTL ImmunoSpot S6 Universal machine and data were analyzed using Professional 6.0.0 software.

Single-cell raw data generation and processing

Tumors from six mice in each group were minced in cold RPMI medium, dissociated and digested using a tumor dissociation kit on a GentleMACS Octo dissociator⁶⁵. Tumor-infiltrating CD4⁺ T cells were enriched by Percoll density gradient centrifugation and sorted with a BD FACSAria Fusion. We followed the Chromium Next GEM single-cell 5ʹ reagent kits v2 protocol (CG000331) and Chromium Next GEM single-cell 5ʹ GEM, library and gel bead kit v2 (PN-1000267) for single-cell library construction. In short, we generated beads in emulsion from the cell suspension of 700–1,200 cells per μl and over 85% cell viability. The barcoded cDNA was amplified and fragmented to construct 5′ gene expression and V(D)J libraries. Then, we assessed the quality of libraries using the Agilent high-sensitivity DNA kit (5067–4626) on an Agilent Bioanalyzer 2100 (Agilent Technologies) and sequenced the qualified libraries using an Illumina Novaseq System (Illumina). The raw single-cell data from the paired RNA library and V(D)J library were demultiplexed and aligned to the mm10–2020-A genome and GRCm38-alts-ensembl-7.0.0 reference using Cell Ranger multi software (version 7.0.0). We used the output filtered_feature_bc_matrix and filtered_contig_annotations for downstream analyses.

scRNA-seq data analysis

We used Seurat (version 4.1.1)⁷³ for scRNA-seq data analysis. We removed low-quality cells with <200 genes, >8,000 genes or >10% unique molecular identifiers mapped to mitochondrial genes, followed by normalization and log transformation in a standard way. Subsequently, we selected top 2,000 highly variable genes for downstream analyses. We performed principal component analysis and used Harmony (version 0.1.1)⁷⁴ for data integration of the two samples. The first 30 Harmony embeddings were used for clustering and uniform manifold approximation and projection (UMAP) visualization. To characterize CD4-positive T cells, we removed CD4-negative clusters, including macrophages, CD8-positive T cells and nonimmune cells. We then identified CD4-positive T cell subclusters by performing clustering on the retained CD4-positive T cells and annotated subclusters by examining the expression of T cell lineage marker genes. We compared the proportion of the Th1-like T cell cluster and the gene expression in this cluster between two samples. We used the Wilcoxon rank-sum test to compare gene expression between Tc1 and Tc9 and the Benjamini–Hochberg procedure to adjust P values. The significantly differentially expressed genes (DEGs) were identified on the basis of adjusted P values. For biological function characterization, we used clusterProfiler (version 4.2.2)⁷⁵ and DEGs with a log₂ fold change > 0.5 between two samples to perform the gene ontology pathway enrichment analysis.

scTCR-seq data analysis

We used scRepertoire (version 1.11.0)⁷⁶ for analyzing T cell clonotypes and clonal expansion. The clonotypes of T cells were identified by TCR α-chain and β-chain CDR3 amino acid sequences. We estimated the usage of TCR genes and clonal expansion of Th1-like T cells in two samples. The diversity of Th1-like T cell clonotypes in two samples was measured using multiple metrics, all indicating lower diversity in Tc9-treated tumors. Clonal expansion and most highly expanded TCRs were also visualized using UMAP generated in scRNA-seq data analysis.

Analysis of gene expression and patient survival in public datasets

For analysis of genes expressed by cDC2 cells, the GSE77671 dataset was used. Series matrix files were downloaded and the gene identifiers were converted to gene symbols. The expression profiles of different DC subsets, T cells, B cells and CD14⁺ monocytes were transferred from log₂-transformed values to the original values⁷⁷. For the generation of gene expression signatures, data were log₂-transformed to normalize expression and ranked by the mean expression value. The gene signatures of cDC2 cells (CD1C, CD1E, CLEC10A and FCER1A) and effector CD4⁺ T cells (CD3E TRAC, CD4, CD69 and IFNG) were used for analysis. Overall survival analyses were performed as follows: the top and bottom quartile expression ranked values for selected genes or ranked sum expression of gene signatures were used and plotted for Kaplan–Meier curves using GraphPad Prism⁷⁷.

Statistical analysis

Statistical analyses and normality and lognormality evaluations of the data were conducted using Prism 8.0 (GraphPad). For multiple comparisons, significance was determined using a two-way analysis of variance (ANOVA) or multiple t-tests (one per row). Single comparisons were analyzed with an unpaired or paired two-tailed Student’s t-test, while survival curves were evaluated using the log-rank (Mantel–Cox) test. The mice were randomly grouped before being treated and no data were excluded from the analysis. Data collection and analysis were not performed blind to the conditions of the experiments. No statistical methods were used to predetermine sample size but our sample sizes are similar to those reported in previous publications^{22,23,28,46,78,79}.

Extended Data

Extended Data Fig. 1 |. Tc9 cells control the outgrowth of natural ALV tumors.

a, MFI of OVA-GFP and OVA (SIINFEKL) on WT B16 and B16^OVA cells (n = 3 technical replicates from a single experiment, repeated two times with similar results). b, Western blot detection of OVA in B16 and B16^OVA cells (one of two independent experiments). c, Schema of experiments in Fig. 1a. d, Schema of experiments in Fig. 1b. e, Western blot and flow cytometry (n = 3 technical replicates from a single experiment, repeated two times with similar results) detection of gp100 in MC38 and MC38^gp100 cells. f, Schema of experiments in Extended Data Fig. 1g. g, PBS, Thy1.1⁺ Pmel-1 Tc1 or Tc9 cells were transferred into CD45.2⁺ B6 mice bearing MC38^gp100 tumors with the adjuvant treatment, and tumor growth curves were shown (n = 5 mice per group, one of two independent experiments). h. Percentages of OVA-GFP^negative and OVA-GFP^positive CD45⁻ tumor cells on day 10 (n = 3 mice) and day 40 (n = 4 mice) after B16^OVA tumor inoculation (one of two independent experiments). i. Percentages of gp100^negative and gp100^positive CD45⁻ tumor cells on Day 10 and Day 40 after MC38^gp100 tumor inoculation (n = 5 mice per group). Data are presented as mean ± SEM. Two-way ANOVA was used to compare antigen percentages in h, i. neg, negative; pos, positive.

Extended Data Fig. 2 |. Tc9 cells control the growth of ALV chimeric tumors.

a, Schema of experiments in Fig. 1e. b, Relative Pmel mRNA expression (n = 3 biologically independent samples per group) and gp100 protein level (one of two independent experiments) in WT B16 and B16^gp100-kd cells. c, Schema of experiments in Fig. 1f. d, Schema of experiments in Extended Data Fig. 1e. e, Thy1.1⁺ Pmel-1 Tc1, Tsm, Tcm, or Tc9 cells were transferred into CD45.2⁺ B6 mice bearing B16 chimeric tumors with the adjuvant treatment, and tumor growth curve were shown (n = 6 mice per group). f, Schema of experiments in Fig. 1g. g, Western blot detection of gp100 protein level in WT B16 and B16^gp100-ko cells (one of two independent experiments). h, Schema of experiments in Extended Data Fig. 2i. i, PBS, Thy1.1⁺ Pmel-1 Tc1 or Tc9 cells were transferred into CD45.2⁺ B6 mice bearing B16 chimeric (B16 + B16 ^gp100-ko) tumors with the adjuvant treatment, and tumor growth curve were shown (PBS, n = 3 mice; Tc1, n = 6 mice; Tc9, n = 6 mice). Data are presented as mean ± SEM. Multiple t tests-one per row were used to compare tumor growth in e (day 35), i (day 35); Two-tailed unpaired Student’s t-test was used in b.

Extended Data Fig. 3 |. Host CD4 + T cells are enriched in Tc9-treated chimeric tumor.

a, Schema of experiments in Fig. 3a–e. b–g, Thy1.1⁺ Pmel-1 Tc1 or Tc9 cells were transferred into Thy1.2⁺ B6 mice bearing B16 chimeric tumor with the adjuvant treatment. Tumor-infiltrating CD45⁺ cells were isolated for CyTOF analysis. t-SNE analysis of CD3, CD4, CD8 and Thy1.2 expression (b), percentage of CD4⁺ T cells in CD45⁺ cells (c), percentage of host Thy1.2⁺ CD8⁺ T cells (d), Thy1.2⁻ transferred CD8⁺ T cells (e) in CD45⁺ cells, relative median intensity of IFNγ and GZMB in CD4⁺ T cells (f), percentage of Foxp3⁺ in CD4⁺ T cells (g). n = 2 in b-g, pooled from 6 mice. h, i, Thy1.1⁺ Pmel-1 Tc1 or Tc9 cells were transferred into Thy1.2⁺ mice bearing MC38 chimeric tumor with the adjuvant treatment, and tumors were analyzed. The number of CD4⁺ T cells, and IFNγ⁺ or GZMB⁺ CD4⁺ T cells were shown (n = 6 mice per group, one of two independent experiments). j, k, Tumors in mice treated as Fig. 3a were analyzed. MFI of IFNγ⁺ or GZMB⁺ and number of host Thy1.1⁻ CD8⁺ T cells were shown (n = 9 mice per group, pooled from two independent experiments). l, Thy1.1⁺ Pmel-1 Tc9 cells were transferred into WT or CD8^−/− mice bearing B16 chimeric tumors with the adjuvant treatment, and tumor growth curve were shown (n = 5 mice per group). m, n, Tumors in mice treated as Fig. 3a were analyzed. MFI of IFNγ⁺ or GZMB⁺ and number of transferred Thy1.1⁺ CD8⁺ T cells were shown (n = 9 mice per group, pooled from two independent experiments). o, p, Thy1.1⁺ Pmel-1 Tc1 or Tc9 cells were transferred into Thy1.2⁺ B6 mice bearing MC38 chimeric tumor with the adjuvant treatment, and tumors were analyzed. MFI of CD44, CD69 and PD-1 on CD4⁺ T cells, percentage of ICOS⁺PD-1^low CD4⁺ T cells in tumors (n = 6 mice per group). Data are presented as mean ± SEM. Two-tailed unpaired Student’s t-tests were used in h–k and m–p. Multiple t tests-one per row were used to compare tumor growth in l (day 40).

Extended Data Fig. 4 |. Analysis of host tumor-infiltrating CD4 + T cells.

a, UMAP plot showing the top 3 most highly expanded TCRs in the Th1-like cells from Tc1- and Tc9-treated mice (n = 1pooled from three mice per group). CDR3 amino acid sequences of α/β chains for each TCR are shown. b, Indexes of TCR clonotype diversity for the Th1-like cells from Tc1- and Tc9-treated mice. c, Violin plots of the most differentially regulated TCR genes in the Th1-like cells from Tc1- and Tc9-treated mice. d, Pie charts represent cumulative frequency of TCRα and TCRβ clonotypes. Bar charts illustrate the count of TCR clonotypes with a frequency greater than 1% in the Th1-like cells from mice treated with Tc1 or Tc9 cells. e, Representative IFNγ ELISPOT assay of CD4⁺ T cells isolated from tumors that were ex vivo stimulated with naive splenocytes pulsed with 5 μg/ml indicated individual peptide (n = 3 biologically independent samples per group). f, Flow cytometry analysis of Vβ5⁺(reacts with the Vβ 5.1 and Vβ 5.2 TCR), Vβ12⁺(reacts with the Vβ 12 TCR), Vβ13⁺ (reacts with the Vβ 13 TCR) clonotypes percentages in tumor-infiltrating CD4⁺ T cells (n = 5 mice per group, one of two independent experiments). g, IFNγ ELISA analysis of Vβ5⁺, Vβ12⁺, Vβ13⁺ tumor-infiltrating CD4⁺ T cells reponse to indicated peptide (n = 6 biologically independent samples per group, pooled from two independent experiments). h, Percentages and the numbers of central memory (CM) and effector memory (EM) CD4⁺ T cells. Tc1, n = 10 mice; Tc9, n = 8 mice. Data are presented as mean ± SEM. Wilcoxon rank sum test were used in c. Two-tailed unpaired Student’s t-tests were used in f, h. One-way ANOVA was used in g.

Extended Data Fig. 5 |. Host CD4⁺ T cell is required for Tc9-mediated tumor restraint.

a, Cytotoxicity of Pmel-1 Thy1.1⁺ Tc9 cells against target cells was determined by a 6-hour cytotoxicity assay at 10:1 E:T ratio. Target cells were WT B16 cells, relapsed B16 cells from tumors on day 40, B16^gp100-kd cells and B16^gp100-ko cells (n = 6 biologically independent samples per group, pooled from three independent experiments). b, Schema of experiments in Fig. 4i. c, Schema of experiments in Extended Data Fig. 5d. d, Thy1.1⁺ Pmel-1 Tc1 cells were transferred into Thy1.2⁺ B6 mice bearing chimeric (B16 and B16^gp100-kd) tumors with the adjuvant treatment, and IgG isotype or αCD4 antibodies were injected on day 20 after tumor cell inoculation. Tumor growth curves of treated mice were shown (n = 5 mice per group). e, Representative plots of CD4⁺ and CD8⁺ cell percentages in peripheral blood of WT and Cd4^−/− mice. f, g, on day 0, mice (n = 6 for each group) were inoculated s.c. in the right back flank with 0.4 million WT B16 and 0.2 million B16^gp100-kd chimeric tumor cells. On day 8, one dose of CTX was given intraperitoneally at 200 mg/kg body weight. On day 9 after tumor inoculation, mice were treated with intravenous injection of 2 million Tc1 or Tc9 cells, followed by i.v. injection of 0.5 million peptide-pulsed bone marrow-derived DCs and 4 doses of rhIL-2 (6 × 10⁵ U). On day 20, mice were injected intraperitoneal every three days with 200μg of InVivoMAb anti mouse NK1.1 or IgG. Schema of experiments in f and tumor growth curve in g. Data are presented as mean ± SEM. One-way ANOVA was used in a; Multiple t tests-one per row were used to compare tumor growth in d (day 40).

Extended Data Fig. 6 |. Tc9 ACT increases the infiltration of DCs into tumors.

a, cDC gating strategy in tumors. b, c, Thy1.1⁺ Pmel-1 Tc1 or Tc9 cells were transferred into Thy1.2⁺ B6 mice bearing MC38 chimeric tumors with the adjuvant treatment, and tumor tissues were harvested for analysis. Percentages and numbers of cDCs in tumors tissues (b, n = 6 mice per group, one of two independent experiments), number of cDC2 in tumors tissues (c, n = 6 mice per group, one of two independent experiments). d, cDC gating strategy in tdLN. e, f, Thy1.1⁺ Pmel-1 Tc1 or Tc9 cells were transferred into Thy1.2⁺ B6 mice bearing MC38 chimeric tumors with the adjuvant treatment, and tdLN were harvested for analysis. Percentages of total migratory cDCs (e, n = 8 mice per group, pooled from two independent experiments) and CD11b⁺ migratory cDC2 in CD45⁺ cells in tdLNs (f, n = 8 mice per group, pooled from two independent experiments). g, CD4⁺ T cell gating strategy in Fig. 5g. h, Schema of experiments in Fig. 6a–f. i, j, Thy1.1⁺ Pmel-1 Tc9 cells were transferred into Thy1.2⁺ B6 mice bearing chimeric tumor with the adjuvant treatment, and CCR7 expression on cDC2 cells from tumor tissues (i, n = 4 mice per group, one of two independent experiments) or tdLNs (j, n = 4 mice per group). k, Schema of experiments in Fig. 6g–m. Data are presented as mean ± SEM. Two-tailed unpaired Student’s t-tests were used in b, c, e, f, i, j.

Extended Data Fig. 7 |. IL-24 promotes cDC2 migration to activate host CD4⁺ T cells.

a, b, Tc9 cells were transferred into WT or Il9r^−/− B6 mice bearing chimeric tumor with the adjuvant treatment, and tumors were harvested for analysis. Schema of experiments (a), percentages and numbers of cDC2 (b, n = 5 mice per group). c, Pmel-1 T cells were transferred into B6 mice bearing MC38 chimeric tumor with the adjuvant treatment. IL-24 expression in tumor-infiltrating Tc1 or Tc9 cells (n = 6 mice per group, one of two independent experiments). d, cDC2 gating strategy for cultured DC sorting. e, f, Pmel-1 Tc9 cells transduced with vector or Il24 shRNA, IL-24 protein (e) and mRNA expression (f); n = 3 biologically independent samples per group in e, f. g, Schema of experiments in Fig. 7f–l. h, IL-24 expression in tumor-infiltrating vector or Il24 shRNA transduced Tc9 cells (n = 8 mice per group, pooled from two independent experiments). i, Il24 mRNA expression in Pmel-1 Tc1 cells transduced with vector or Il24 shRNA (n = 3 biologically independent samples per group). j-l, Vector- or Il24 shRNA-transduced Tc1 cells were transferred into mice bearing B16 chimeric tumor with the adjuvant treatment, tumors and tdLNs were harvested for analysis. Schema of experiments (j); Number of CD11b⁺ migratory cDC2 and CD4⁺ T cells in tdLNs (k); Number of cDC2 and CD4⁺ T cells in tumors (l); n = 5 mice per group in j–l. m, n, vector- or Il24 shRNA-transduced Tc1 cells and vector-transduced Tc9 cells were transferred into mice bearing B16 chimeric tumor with the adjuvant treatment, tumors were harvested for analysis. The percentage of Thy1.1⁺ CD8⁺ T cells (m) and IL-24 concentration (n) in tumors; n = 5 mice per group in m, n. o, Vector- or Il24 shRNA-transduced Tc1 cells were transferred into mice bearing B16 chimeric tumor with the adjuvant treatment, tumor growth curves of treated mice were shown (n = 5 mice per group). Data are presented as mean ± SEM. Unpaired Student’s t-test were used in b,c,e,f,h,i,k,l. One-way ANOVA was used in m,n. Multiple t tests-one per row were used to compare tumor growth in o (day 35).

Extended Data Fig. 8 |. IL24, cDC2 and effector CD4 + T cell genes in breast cancer.

a, Prognostic value of IL24 expression for overall survival of breast cancer patients comparing high versus low quartiles (n = 263 per group). b, Correlation of IL24 and cDC2 signature in breast cancer patients (n = 1050). c, d, Heatmap (n = 1050) of ordered, z-transformed expression values of cDC2-specific genes and expression of cDC2 signature genes for top and bottom quartiles (n = 263 per group) in breast cancer patients. e, Prognostic value of cDC2 signature or CD1C expression for overall survival of breast cancer patients comparing high versus low quartiles (n = 263 per group). f, g, Heatmap (n = 1050) of ordered, z-transformed expression values of effector CD4⁺ T cell-specific genes and expression of effector CD4⁺ T cell-specific signature genes for top and bottom quartiles (n = 263 per group) in breast cancer patients. h, Prognostic value of effector CD4⁺ T cell signature expression for overall survival in breast cancer patients comparing high versus low quartiles (n = 263 per group). (i) Correlation of effector CD4⁺ T cell signature with IL24 or with cDC2 signature in breast cancer patients (n = 1050). Data are presented as mean ± SEM. The log-rank (Mantel–Cox) test was performed to compare survival curves in a, e, h; Linear regression were used in b and i.

Extended Data Fig. 9 |. Schematic diagram of the study.

After adoptive T cell transfer, antigen-positive tumor cells lost their primary antigens. Adoptively transferred Tc9 cells not only killed antigen-expressing primary tumors but also secreted high levels of IL-24, which recruited dendritic cells (DCs), particularly cDC2, to the tumor microenvironment. cDC2 then migrated to tumor-draining lymph nodes (tdLNs) in a CCR7-dependent manner, where they activated host Th1-like effector CD4⁺ T cells. These activated CD4⁺ T cells subsequently infiltrated tumors and contributed to controlling ALV tumor growth.

Supplementary Material

Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s43018-025-00935-0.

Data availability

Microarray, scRNA-seq and scTCR-seq data that support the findings of this study were deposited to the Gene Expression Omnibus under accession codes GSE239842 and GSE240762. Mass cytometry data were deposited to ImmPort under accession code SDY2914. Previously published microarray data that were reanalyzed here are available from the Gene Expression Omnibus udner accession codes GSE77671 (ref. 44) and GSE8401 (ref. 42). For the analysis of genes expressed by cDC2 cells, the GSE77671 dataset was used. For the analysis of persons with melanoma, gene expression and clinical parameters were retrieved from Oncomine (www.oncomine.org) and dataset GSE8401. The human breast cancer data were derived from TCGA Research Network: (https://www.cancer.gov/tcga) using UCSC Xena (https://xenabrowser.net/datapages/)⁴³. The remaining data are available within the article and Supplementary Information or from the corresponding authors upon reasonable request. Source data are provided with this paper.