Distinct migratory dendritic cell subsets cooperate to prime tissue-resident memory-like CD8⁺ T cells in mouse tumors

Abstract

Tumor-infiltrating tissue-resident memory (T_RM)-like CD8⁺ T cells correlate with positive prognosis and better responses to cancer immunotherapy, yet the mechanisms underlying their specification remain ill-understood. Here, we use the KP (Kras^G12D, p53^-/-) mouse model of lung adenocarcinoma to investigate how different dendritic cell (DC) subsets shape tumor-specific CD8⁺ T cell responses. Selective ablation of XCR1⁺ DC1s or IRF4-dependent CD11b⁺ DC2s reveal that DC1s, but not DC2s, are required for the generation of tumor-infiltrating effector and exhausted CD8⁺ T cells. By contrast, both migratory DC1s and migratory DC2s contribute to the specification of CD103⁺CD69⁺ T_RM-like CD8⁺ T cells in tumor-draining lymph nodes. Mechanistically, low peptide/MHC-I density on migratory DC2s favors TGFβ-dependent T_RM differentiation, while high peptide/MHC-I density on DC1s promote proliferative expansion of T_RM-like cells; in addition, IL-12 secreted by DC1s enhances CXCR6 expression. Together, our findings uncover complementary functions of migratory DC subsets in shaping anti-tumor T_RM-like CD8⁺ T cell responses, and also implicate targeting multiple DC subsets for optimizing cancer immunotherapy.

Tissue-resident memory (T_RM) CD8⁺ T cells in the tumor microenvironment are relevant for anti-tumor immunity and immune checkpoint blockade response. Here, the authors conditionally ablate type 1 and type 2 dendritic cells in a lung cancer mouse model, finding both to be critical for the differentiation and expansion of T_RM-like T cells.

Introduction

Tumor-infiltrating CD8⁺ T cells (TIL) are critical mediators of anti-tumor immunity. CD8⁺ TILs are characterized by the pervasive expression of exhaustion-associated features initially characterized in chronic infection models, including reduced proliferative response to TCR triggering and reduced polyfunctional cytokine responses. Multiple studies suggest that the qualitative composition of CD8⁺ TILs in solid tumors predicts clinical outcome and response to immunotherapies^1,2. Accordingly, a subset of CD8⁺ TILs with a tissue-resident memory phenotype (T_RM) has been identified in solid tumors, including lung and breast cancer, which correlates with improved prognosis and better response to immunotherapies^1,3–9.

Bona fide T_RM have first been identified in peripheral tissues after clearance of viral infections¹⁰. Unlike other activated and memory CD8⁺ T cells, T_RM do not recirculate but persist in tissues where they ensure immunosurveillance. T_RM are characterized by a specific transcriptional program including the expression of transcription factors Hobit (encoded by Zfp683), Runx3 and Blimp-1^4,11,12. In some tissues, T_RM express αE integrin/CD103, which associates with integrin β7 to bind E-cadherin, stabilizing contacts with epithelial cells, and low levels of S1P receptors ensuring tissue residency¹³. Also, T_RM stably express the CD69 lectin mediating S1P desensitization, further contributing to tissue residency¹⁴.

Similar to bona fide anti-infectious T_RM, T_RM-like TILs harbor canonical features of residency, including the concomitant expression of CD103, CD69, Hobit, and reduced levels of S1PR1. A recent study in a murine model of breast cancer suggests that the CD69⁺CD49a⁺P2Rx7⁺ T_RM phenotype within CD8⁺ TILs may not indicate actual residency within tumors¹⁵. Still, several studies show that CD8⁺ T_RM-like TILs can display potent anti-tumor function¹⁶ as they secrete cytokines, exhibit cytotoxic activity¹⁷, and can potentially generate a large repertoire of T cell subsets upon antigenic restimulation^18,19 or after immune checkpoint blockade²⁰. T_RM-like TILs express some features of exhausted T cells, but compared to other exhausted TILs, T_RM-like TILs display enhanced cytotoxicity and provide local protection in a murine model of breast cancer²⁰. Consistently, numerous studies have identified a strong correlation between CD8⁺ TILs with a T_RM phenotype and favorable outcomes, and increased response rates to immune checkpoint inhibition in human lung cancer^1,3,5–7,9, triple-negative breast cancer and other solid tumors^16,21. This calls for a better understanding of the mechanisms ensuring the acquisition of the T_RM program during T cell activation in solid tumors.

In cancer, CD8⁺ T cells are initially activated in tumor-draining lymph nodes (tdLN) and terminally differentiate within the tumor^22,23. However, the anatomical location where T_RM-like fate specification occurs in cancer is unclear (i.e., during priming in tdLN or after infiltration in the TME). In a mouse model of skin vaccination, TGF-β activated by steady-state migratory DCs imprints an epigenetic state in naïve T cells, predisposing them to differentiate into T_RM²⁴. Other studies identify T_RM-committed circulating progenitors during vaccination or viral infection, arguing in favor of early T_RM fate decision^25,26. Anti-infectious T_RM can also be generated from circulating effector cells receiving T_RM instruction after tissue infiltration^4,27,28.

Dendritic cells (DC) are the sentinel cells of the immune system, ensuring T cell activation and differentiation. Classical DCs are diverse and composed of two main families expressing and relying on IRF8 or IRF4 transcription factors. There is wide evidence that IRF8⁺, BATF3-dependent XCR1⁺ DC1s control the activation of effector CD8⁺ T cell responses in the context of immunogenic tumors^29–31 and maintain the fitness of the CD8⁺ T cell effector compartment by expanding TCF1⁺ T cells in tdLN³². Efficient phagocytosis of tumor-derived cell debris and efficient cross-presentation of cell-associated antigens by MHCI^33–35and MHCII presentation to helper T cells³⁶ all support the paramount importance of XCR1⁺ DC1s in shaping effector CD8⁺ T cell responses. By contrast, the nature of antigen-presenting DC subtypes controlling T_RM activation during T cell priming and infiltration is less clear and has mostly been investigated in the context of vaccination or infections. Skin XCR1⁺ Batf3-dependent DC1s are essential for the efficient activation of skin T_RM after vaccination³⁷. IRF4-dependent DC2s are essential for the priming of influenza-specific T_RM in lungs³⁸ and the maintenance of CXCR3⁺ T_RM against HSV cervical infection²⁷. Also, human IRF4⁺CD163⁺CD1c⁺ DCs are the most potent antigen-presenting cells for T_RM specification in vitro or in humanized mice^39,40.

Despite numerous evidence of a role for CD8⁺ T_RM-like TILs in non-small cell lung cancer (NSCLC) immunosurveillance^{1,3,5–7,9,21}, little is known on (i) the anatomical location where the T_RM-like phenotype is specified and (ii) if and how migratory DCs subsets participate to the induction of T_RM-like CD8⁺ T cells within tumor-draining lymph nodes.

In this study, we delineate the activation of tumor-infiltrating T_RM-like cells in the KP (Kras^G12D, p53^-/-) model of lung adenocarcinoma. We identify an unexpected contribution of both migratory XCR1⁺ DC1s and CD11b⁺ DC2s to the activation of T_RM-like precursors in tdLNs, which subsequently give rise to T_RM-like TILs within tumors. In addition, we show that the low density of MHCI-peptide complexes on migratory DC2s elicits a mild TCR triggering, thereby promoting specification toward a T_RM-like phenotype. Together, these findings reveal a cooperation between distinct DC subsets and suggest that effective immunotherapies aimed at inducing T_RM-like TIL responses may require the coordinated targeting of multiple DC subsets.

Results

CD8⁺ T cells with a T_RM-like phenotype infiltrate lung tumors

To determine the extent of T_RM-like cells infiltrating the lungs after engraftment of tumor cells, we employed a cell line derived from Kras^G12D p53^-/- mice^32,41,42 (Supplementary Fig. 1A). We found that a large proportion of CD8⁺ TILs stained positive for both CD69 and CD103 expression (27.3%, Fig. 1A), which is a classical gating for the T_RM phenotype. CD8⁺ TILs also stained positive for PD1 and TIM3 (21.7%, Fig. 1B), a feature of exhausted T cells. In line with other studies highlighting the expression of exhaustion markers on T_RM-like TILs^3,20, we found that these two subsets were largely overlapping (Fig. 1C and Supplementary Fig. 1B). This led us to investigate the correspondence between cell surface phenotype and transcriptional profile in unbiased settings using a single-cell RNA-sequencing (scRNA-seq) combined with CITE-Seq approach (Fig. 1D and Supplementary Fig. 1C). scRNA-seq identified 5 clusters of Cd44⁺Sell^- (coding respectively for CD44 and CD62L) cells (clusters 1, 4, 7, 8, and 9) (Fig. 1D–F and Supplementary Fig. 1D). Clusters 4 expressed stemness genes, low level of Pdcd1 (coding for PD1) and scored high for the circulating signature, aligning it to recently extravasated from the blood, effector CD8⁺ T cells. Cluster 7 expressed stemness genes, intermediate levels of Pdcd1, and intermediate score for the exhaustion module, aligning it to a progenitor exhausted phenotype. Clusters 8 was the only cluster expressing high levels of proliferation-associated genes and scoring high for the Proliferation module, corresponding to a proliferative cluster. Clusters 1 and 9 expressed high levels of multiple exhaustion molecules (Pdcd1, Havcr2/Tim3, Lag3, Tigit, Tox) and scored high for an exhaustion signature. Among these clusters, cluster 1 expressed core T_RM-associated genes (such as Itgae/CD103, Cd69, Itga1/CD49a and Cxcr6) and scored high for Residency and T_RM signatures. Using CITE-Seq antibodies, we identified T_RM-like TILs (CD44⁺CD62L^-CD103⁺CD69⁺) and found that they overlapped mostly with cluster 1 (Fig. 1G and Supplementary Fig. 1E). By contrast, cluster 9 was low for Residency and T_RM signatures. We therefore classified cluster 1 as T_RM-like and cluster 9 as terminally exhausted.

Fig. 1. T_RM-like CD8⁺ T cells infiltrate lung tumors.

A Representative flow cytometry plots and quantifications of CD69⁺CD103⁺ CD8⁺ T_RM in the lungs of tumor-free and KP tumor-bearing mice 14 days post-tumor cell injection (n = 4 tumor-free and n = 8 KP mice, two-sided unpaired t-test). Data are presented as box plots with median, 25th–75th percentiles, minimum–maximum whiskers, with individual replicates displayed. B Representative flow cytometry plots and quantifications of exhausted PD1⁺TIM3⁺ CD8⁺ T cells in the lungs of tumor-free and KP tumor-bearing mice 14 days post-tumor cell injection (n = 4 tumor-free and n = 8 KP mice, two-sided unpaired t-test). Data are presented as box plots with median, 25th–75th percentiles, minimum–maximum whiskers, with individual replicates displayed. C Venn diagram showing the overlap between CD69⁺CD103⁺ and PD1⁺TIM3⁺ CD8⁺ T cells in the lungs of KP-tumor-bearing mice 14 days post-tumor cell injection (n = 8 mice). D–G scRNAseq analysis of sorted total lung CD8⁺ T cells from KP-bearing mice 14 days post-tumor cell injection (see also Supplementary Fig. 1C). D UMAP dimensionality reduction embedding the different clusters. E Dot plot showing the RNA expression of core marker genes across clusters. F Violin Plots showing the scoring of indicated gene signatures across Cd44⁺Sell^- clusters^4,11,16. G Distribution of Cd44⁺Sell^- clusters across CITE-Seq gated CD44⁺CD62L^-PD1⁺CD69⁺CD103⁺ T_RM, CD44⁺CD62L^-PD1⁺Not(CD69⁺CD103⁺) T_Exh, and CD44⁺CD62L^-PD1^- T_Eff. H Representative flow cytometry gating strategy used to identify CD8⁺ T cell populations infiltrating lung KP tumors. I Quantification of the proportions of CD62L^-CD44⁺ CD8⁺ T cell populations infiltrating lung KP tumors (n = 10 mice). Data are presented as stacked mean values ± SD. J, K Bulk RNA-sequencing analysis of sorted CD8⁺ T cell populations as defined in (H). J Signature scoring of gene modules related to stemness, circulation, activation-cytotoxicity, exhaustion, and tissue-residency as shown in Supplementary Fig. 1L. K Principal-component analysis (PCA). Top high- and low-loading genes are indicated for PC1 and PC2. L Relative MFI protein expression of indicated marker genes across CD8⁺ T cell populations as determined by flow cytometry in (H). M Hobit-tdTomato MFI expression across lung CD8⁺ T cell populations quantified in Hobit-tdTomato reporter mice¹⁸ bearing KP tumors (n = 5 mice, repeated measures one-way ANOVA with Tukey’s multiple comparisons test). Data are presented as mean values ± SD with individual replicates displayed. N Representative histogram and quantifications of CD45.2-labelled cells following intravenous injection of a CD45.2-fluorescent antibody 5 min before euthanasia (n = 5 mice, repeated measures one-way ANOVA with Tukey’s multiple comparisons test). Data are presented as mean values ± SD with individual replicates displayed. O Representative flow cytometry plots showing the gating of tetramer-OVA⁺ CD8⁺ T cells infiltrating KP-OVA but not KP lung tumors. P Representative flow cytometry gating strategy showing the phenotype of tetramer-OVA⁺ (dark) or tetramer-OVA^- cells (grey) infiltrating KP-OVA lung tumors. Q Quantification of the proportions of Tet-OVA⁺ or Tet-OVA^- CD8⁺ T cell populations infiltrating lung KP-OVA tumors (n = 6 mice, repeated measures two-way ANOVA with Šídák’s multiple comparisons test). Data are presented as stacked mean values ± SD. R Quantification of the absolute numbers of Tet-OVA⁺ or Tet-OVA^- CD8⁺ T cell populations infiltrating lung KP-OVA tumors (n = 6 mice, repeated measures two-way ANOVA with Šídák’s multiple comparisons test). Data are presented as mean values ± SD with individual replicates displayed. KP: (Kras^G12D, p53^-/-) lung adenocarcinoma; T_RM: tissue-resident memory-like CD8⁺ T cell; T_Ex: exhausted-like CD8⁺ T cell; T_Eff: effector-like CD8⁺ T cell; i.v.: intravenous; Tet: tetramer.

To further confirm these results, we designed a gating strategy enabling the prospective isolation of TILs populations identified above. As already shown previously, CD69⁺CD103⁺ T_RM-like TILs (cluster 1) were largely enriched in an exhausted phenotype PD1⁺TIM3⁺ (68.0%, termed here as TIM3⁺ T_RM-like) but also contained a population of PD1⁺TIM3^- (29.2%, termed here as TIM3^- T_RM-like) (Fig. 1H, I). Conversely, we also identified PD1⁺TIM3⁺ (37.0%, termed here as TIM3⁺ T_Ex, cluster 9) and PD1⁺TIM3^- (42.9%, termed here as TIM3^- T_Ex, cluster 7) distinct from CD69⁺CD103⁺. PD1-negative CD8⁺ T cells also comprised CD62L^-CD44⁺ activated effector T cells (20.4%, termed here as PD1^- T_Eff, cluster 4). Similar populations were found in the autochthonous KP cancer model, which recapitulates the human physiopathology from hyperplasia to adenocarcinoma and the onset of T cell dysfunction^32,41–43 (Supplementary Fig. 1F–I).

Bulk RNA-seq profiling further validated the correspondence between the T_RM-like gate and the enrichment in T_RM-associated genes and signatures (e.g., Itgae, Itga1, Rgs1, Cxcr6) and gene ontology pathways “regulation of cell migration,” “TGF-β signalling pathway,” “R-SMAD binding” (Fig. 1J, K, and Supplementary Fig. 1J–M). By contrast, T_RM expressed lower levels of effector-associated genes (e.g., Eomes, Gzmk) and of genes ensuring recirculation (e.g., Klf2 and S1pr1)^4,11. Finally, we confirmed by FACS that T_RM-associated proteins (Hobit in Hobit reporter mice¹⁸ and CD49a encoded by Itga1) were expressed at higher levels in TIM3⁺ T_RM, while they express low levels of EOMES and KLRG1 (Fig. 1L, M). Also, we found that the human T_RM signature was enriched in TIM3⁺ T_RM-like as compared to TIM3⁺ T_Ex and in TIM3^- T_RM-like as compared to TIM3^- T_Ex (Supplementary Fig. 1N), thereby showing the alignment of TILs in mouse lung KP tumors and human cancer. Unlike other T cell subsets, T_RM-like TILs (CD69⁺CD103⁺TIM3^+/-) were not stained after intravenous injection of anti-CD45.2-PE antibodies labeling the vascular compartment. This demonstrates their localization in the lung parenchyma (Fig. 1N).

We next focused our analysis on tumor antigen-specific cells after engraftment of a KP cell line expressing the ovalbumin protein (KP-OVA, Fig. 1O, P). Tumor-antigen-specific tetramer-OVA⁺ cells (Tet-OVA⁺) encompassed similar subsets but displayed a significant enrichment in TIM3-expressing cells, including both TIM3⁺ T_Ex and TIM3⁺ T_RM-like, unlike PD1^- T_Eff that were mostly found in the Tet-OVA^- fraction (Fig. 1P–R and Supplementary Fig. 2D, E). Similar results were found in the AT3-OVA orthotopic model of breast cancer (Supplementary Fig. 2A–C). Functionally, we found that both TIM3⁺ T_RM-like (15.5%) and TIM3⁺ T_Ex (11.9%) were able to produce pro-inflammatory cytokines IFNγ and TNF after ex vivo antigen-specific restimulation with the OVA_257–264 peptide (Supplementary Fig. 2F). Moreover, both T_Ex and T_RM-like exert cytotoxicity and release granzyme B during ex vivo co-culture assays with KP-OVA tumors (Supplementary Fig. 2G). However, T_RM-like-containing cultures demonstrate higher IFNγ release as compared to T_Ex. We next sought to determine if T_RM-like would respond to immune checkpoint blockade (ICB). To this end, we analyzed KP TILs in mice treated with anti-PD-1 and anti-CTLA-4 antibodies alone or in combination (Supplementary Fig. 2H–J). We found that combination treatment induced a relative expansion of TIM3^- T_RM-like as compared to other CD8⁺ TILs. Acute exposure to anti-PD-1 was sufficient to increase the percentage of polyfunctional IFNγ⁺TNF⁺ cells in all TIM3⁺ TILs, resulting in the most efficient cytokine production in TIM3⁺ T_RM-like as compared to TIM3⁺ T_Ex. (Supplementary Fig. 2K–M).

Altogether, we conclude that the transplanted KP model recapitulates some features of human NSCLC in terms of partial dysfunctionality and T_RM-like phenotype^1,3,7,9,44. Our results define T_RM-like TILs as a specific, separable subset within a larger population of CD8⁺ TILs expressing co-inhibitory receptors. Therefore, this model is suitable to investigate how the T_RM-like phenotype is instructed in comparison to other TILs.

Tumor-draining lymph nodes contain a fraction of CD103⁺ activated CD8⁺ T cells poised to generate T_RM-like TILs

The anatomical localization where T_RM specification occurs is still a matter of debate. Therefore, we wondered whether CD8⁺ T_RM-like TILs require egress from tumor-draining mediastinal lymph nodes (tdLN) or differentiate in situ from a T cell pool pre-existing tumor development. To address this, we inhibited lymph node egress using S1PR1 blockade with fingolimod (FTY720) and analysed T_RM-like TILs at day 14 post-tumor engraftment (Fig. 2A). We found that FTY720 administration from day 1 to 14 or from day 7 to 14 prevented the accumulation of activated CD44⁺CD62L^-CD8⁺ T cells in the lungs, including T_RM-like cells (Fig. 2B). We conclude that tdLNs is a crucial site during the development of CD8⁺ T_RM-like TILs.

Fig. 2. Tumor-draining lymph nodes contain a fraction of CD103⁺ activated CD8⁺ T cells poised to generate T_RM-like TILs.

A Experimental design for T cell trapping in the LN. Mice were injected i.v. with KP tumors. Mice were treated with vehicle or FTY720 to block lymphocyte egress from all lymphoid tissues every 2–3 days from day 1 or from day 7 post-KP injection. Created in BioRender. Guermonprez, P. (2026) https://BioRender.com/t3f9mbm. B Quantification of T_RM after FTY720 treatment at day 14 post-KP injection (n = 9 vehicle, n = 4 FTY720 mice day 1–14, and n = 9 FTY720 day 7–14 mice, one-way ANOVA with Tukey’s multiple comparisons test). Data are presented as mean values ± SD with individual replicates displayed. C Representative flow cytometry plots showing CD103⁺ and CD103^- T_Act in the mediastinal LN of tumor-free or KP-tumor-bearing mice. D Quantification of the numbers of CD103⁺ and CD103^- T_Act in the mediastinal LN of tumor-free and KP-tumor-bearing mice (n = 3 naïve and n = 8 KP mice, two-sided unpaired t-test between total T_Act numbers). Data are presented as stacked mean values ± SD. E Representative flow cytometry plots showing the phenotype of tetramer-OVA⁺ (dark) or tetramer-OVA^- cells (grey) in the tdLN of KP-OVA tumor-bearing mice. F Quantifications of the proportions of CD103⁺ and CD103^- T_Act within tetramer-OVA⁺ and tetramer-OVA^- cells in the tdLN of KP-OVA tumor-bearing mice (n = 4 KP-OVA mice, repeated measures two-way ANOVA with Šídák’s multiple comparisons test). Data are presented as stacked mean values ± SD. G Volcano plot representing differentially expressed genes between CD103⁺ T_Act and CD103^- T_Act (two-sided Wald test, p-values adjusted for multiple testing using the Benjamini–Hochberg method). Genes with Log2(fold change, FC) > ± 0.25 and a false discovery rate (FDR)-adjusted p-value of less than 0.05 were considered significant. H GSEA of pairwise comparisons of CD103⁺ T_Act vs CD103^- T_Act using differentially expressed genes between lung TIM3⁺ T_RM and TIM3⁺ T_Ex as defined in Supplementary Fig. 1J. Enrichment significance was assessed by phenotype permutation testing, and false discovery rate (FDR) q-values were used to correct for multiple hypothesis testing. I Flow cytometry quantifications of T_effector core markers in CD103^- T_Act and CD103⁺ T_Act (n = 5 mice for T-bet, Eomes, and CX3CR1; n = 7 mice for KLRG1; two-sided paired t-tests). Median fluorescent intensity was calculated for transcription factors and percentage of positivity for membrane proteins. Data are presented as individual paired replicates. J Flow cytometry quantifications of T_RM core markers in CD103^- T_Act and CD103⁺ T_Act (n = 4 mice for CTLA4 and CXCR6; n = 5 mice for CD69; two-sided paired t-tests). Median fluorescent intensity was calculated for transcription factors and percentage of positivity for membrane proteins. Data are presented as individual paired replicates. K–P Adoptive transfer of total tdLN from CD103^ERcre x ROSA^LSL-tdTomato mice. K Experimental design. Donor CD45.2⁺ CD103^ERcre x ROSA^LSL-tdTomato and host WT CD45.1⁺ mice were injected with KP cells. Eight days later, tomato labeling was induced in donor mice by intraperitoneal injection of tamoxifen. At day 9, donor tdLNs were transferred intravenously in host mice (i.e., each total donor tdLN was transferred into one recipient mice). Lungs from host mice were analyzed 5 days post-transfer to assess the progeny of tdTomato-labelled transferred cells. Created with BioRender. L Representative flow cytometry plots and quantification of the percentage of tdTomato⁺ within donor CD103^- or CD103⁺ T_Act in tdLN pre-transfer (n = 10 mice, two-sided paired t-test). Data are presented as mean values ± SD with individual replicates displayed. M Representative flow cytometry plots and quantification of the percentage of CD103⁺ within donor tdTomato^- or tdTomato⁺ T_Act in tdLN pre-transfer (n = 10 mice, two-sided paired t-test). Data are presented as mean values ± SD with individual replicates displayed. N Representative flow cytometry plots and quantification of the percentage of T_RM-like within donor tdTomato^- or tdTomato⁺ CD44⁺CD62L^-CD8⁺T in recipient lungs post-transfer (n = 10 mice, two-sided paired t-test). Data are presented as mean values ± SD with individual replicates displayed. O Representative flow cytometry plots and quantification of the percentage of tdTomato⁺ within donor T_RM-like or not-T_RM in recipient lungs post-transfer (n = 10 mice, two-sided paired t-test). Data are presented as mean values ± SD with individual replicates displayed. P Quantifications of the expression of PD1, TIM3, and CXCR6 within donor tdTomato^- or tdTomato⁺ in recipient lungs post-transfer (n = 5 mice, two-sided paired t-tests). Data are presented as mean values ± SD with individual replicates displayed. KP: (Kras^G12D, p53^-/-) lung adenocarcinoma; tdLN: tumor-draining lymph node; T_Act: activated CD8⁺ T cell; T_RM: tissue-resident memory-like CD8⁺ T cell; T_Ex: exhausted-like CD8⁺ T cell; T_Effector: effector-like CD8⁺ T cell; i.v.: intravenous; i.p.: intraperitoneal; Tet: tetramer; Tam: tamoxifen.

Having established a role for tdLN in T_RM infiltration, we next sought to characterize tdLN CD8⁺ T cells after KP tumor engraftment. We found that KP tumor development induced the expansion of CD62L^-CD44⁺ activated CD8⁺ T cells (T_Act) (Fig. 2C and Supplementary Fig. 3A–D). Because CD103 expression has been widely used as a marker of T_RM-biased T cells in other settings²⁴, we analysed CD103 expression and found that 41.3% of T_Act expressed CD103 (Fig. 2C and Supplementary Fig. 3A, B). KP tumors induced a large expansion of CD103⁺ and CD103^- T_Act (Fig. 2D). CD103⁺ T_Act population was similarly observed in OVA-specific T_Act after KP-OVA engraftment (15.0%) (Fig. 2E, F, and Supplementary Fig. 3E). Bulk RNA-sequencing analysis showed that CD103⁺ T_Act were enriched in T_RM core genes, such as Cxcr6, Fgl2, Itga1, and Innp4b (Fig. 2G and Supplementary Fig. 3F–H). On the contrary, CD103^- T_act expressed more effector genes (Eomes, Gzmk) and proliferation-associated genes (Mki67). Consistently, GSEA analysis revealed that lymphocyte residency signatures were enriched in CD103⁺ T_Act while signatures of genes associated to circulating lymphocytes were enriched in CD103^- T_Act^4,11 (Supplementary Fig. 3I, J). CD103⁺ T_Act were also enriched for genes associated to KP-bearing lung TIM3⁺ T_RM-like, while CD103^- T_Act were enriched for genes associated to KP-bearing lung TIM3⁺ T_Ex (Fig. 2H). At the protein level, CD103⁺ T_Act expressed lower levels of effector-associated markers (T-BET, EOMES, KLRG1, and CX3CR1) (Fig. 2I) but higher T_RM-enriched markers, such as CD49a, Hobit, CTLA4, and CXCR6 (Fig. 2J and Supplementary Fig. 3K, L). We therefore hypothesized that CD103⁺ T_Act are biased to give rise to lung T_RM-like TILs. To assess the relationship between CD103⁺ T cells in tdLN and lung T_RM-like TILs, we took advantage of the CD103-cre^ERT2-ROSA^LSL-tdTomato reporter mouse⁴⁵ (hereafter referred as CD103-tdTomato). Tamoxifen induction efficiently labelled CD103⁺ T_Act (45.6%) but not CD103^- T_Act (1.5%) in tdLNs from CD103-tdTomato donor mice (Fig. 2L). As a result, 84.7% of tdTomato⁺ cells stained positive for CD103, validating labeling specificity (Fig. 2M). We thus transferred whole CD45.2⁺ tdLN containing genetically labelled CD103⁺ T cells in KP tumor-bearing CD45.1⁺ WT recipient mice and assessed T_RM phenotype in the progeny of tdTomato^- or tdTomato⁺ transferred tdLN CD8⁺ T cells in the lungs (Fig. 2K–P and Supplementary Fig. 3M–O). We found that CD45.2⁺ tdTomato⁺ transferred T cells predominantly gave rise to T_RM-like but barely no other CD44⁺CD62L^- CD8⁺ TILs (67.0%, Fig. 2N–P and Supplementary Fig. 3N, O). By contrast, only a minor fraction of tdTomato^- transferred CD8⁺ T cells differentiated into T_RM-like TILs (7.0% of CD44⁺CD62L^- CD8⁺ TILs).

These results demonstrate that T_RM specification is initiated in tdLNs, where CD103⁺ T_Act cells are pre-committed to differentiate into T_RM-like TILs after seeding tumor-bearing lungs.

CD103⁺ activated T cells in tdLNs contain a tumor antigen-specific CXCR6⁺ population with T_RM characteristics

To better characterize the heterogeneity of tdLN activated CD103⁺CD8⁺T cells giving rise to lung T_RM-like TILs in unbiased settings, we performed single-cell RNA-sequencing of a mixture of 80% of T_Act and 20% of T_Naive and T_CM (Fig. 3A and Supplementary Fig. 4A). We identified 4 clusters of activated Cd44⁺Sell^- CD8⁺ T cells (clusters 0, 3, 5, and 8) (Fig. 3B, C, and Supplementary Fig. 4B). Cluster 5 was highly proliferating (Mki67 ^high). Among the remaining activated clusters, Itgae expression was restricted to clusters 0 and 8. Strikingly, cluster 8 expressed high levels of Cxcr6 and scored high for residency but low for circulating signatures or stemness-associated genes (Slamf6, Tcf7) (Fig. 3B, C). Consistently, this cluster scored the highest for genes associated to KP-bearing lung TIM3⁺ T_RM (Supplementary Fig. 4B). CITE-Seq gating also showed that CD103 was most highly expressed by cluster 8 (Supplementary Fig. 4C). However, the majority of CD103⁺ T_Act cells were found in cluster 0 since cluster 8 contains very few cells (Fig. 3D). Of note, Sell⁺ cells contained a cluster of CD103⁺ T cells (cluster 4) which might represent preconditioned naïve progenitors as described in other studies^24,46. To further compare single-cell RNA-seq clusters obtained in both lung and tdLN locations, we projected lung clusters on the tdLN dataset. We found that T_RM cluster 1 from the lung was almost exclusively projected on Cxcr6⁺ cluster 8 from the tdLN (84.25%, Fig. 3E), thereby validating their transcriptional alignment. Taking advantage of this information, we refined our FACS gating strategy to separate previously gated CD103⁺ T_Act into CD103⁺CXCR6^- and CD103⁺CXCR6⁺ T_Act (Fig. 3F). We found that CD103⁺CXCR6⁺ T_Act expressed low levels of SLAMF6 but higher levels of PD1, CD49a and Hobit than CD103^- or CD103⁺CXCR6^- T_Act (Fig. 3G and Supplementary Fig. 4D). CD103⁺CXCR6⁺ T_Act were found highly enriched within tumor antigen-specific Tet-OVA⁺ T_Act from KP-OVA bearing mice (4.2% of Tet-OVA^- T_Act vs 16.9% of Tet-OVA⁺ T_Act) (Fig. 3H–J and Supplementary Fig. 4E). Phenotyping Tet-OVA⁺ T_Act led to similar results compared to polyclonal T_Act as CD103⁺CXCR6⁺ T_Act expressed high levels of T_RM marker CD49a but low levels of SLAMF6 (Supplementary Fig. 4F, G). Similar results were obtained in the AT3-OVA orthotopic breast tumor model in which tumor-specific CD8⁺ T cells contained a large CD103⁺CXCR6⁺ fraction (39.4% of Tet-OVA⁺ T_Act) (Supplementary Fig. 4I–K).

Fig. 3. CD103⁺ activated T cells in tdLN contain a tumor antigen-specific CXCR6⁺ population with T_RM-like characteristics.

A–E scRNAseq of 80% T_Act and 20% T_CM/T_Naive sorted from the tdLN of 14-day KP-bearing mice (gating strategy in Supplementary Fig. 4A). A UMAP dimensionality reduction embedding the different clusters. B Dot plot showing the expression of core marker genes across clusters. C Violin plots showing the scoring of indicated gene signatures across activated clusters^4,11,16. D Distribution of T_Act clusters across CITE-Seq gated CD103^- T_Act and CD103⁺ T_Act. E Projection of lung Cd44⁺Sell^- scRNA-seq clusters 1, 4, 7, 8, and 9 on tdLN T_Act clusters 0, 3, 5, and 8. F Representative flow cytometry gating strategy and quantification of the proportions of T_Act populations in KP tdLN (n = 12 mice). Data are presented as stacked mean values ± SD. G Representative flow cytometry histograms and quantifications of SLAMF6, CD49a, and Hobit-tdTomato expression across tdLN T_Act subsets (n = 5 Hobit-tdTomato mice, repeated measures one-way ANOVA with Tukey’s multiple comparisons test). Data are presented as mean values ± SD with individual replicates displayed. H Representative flow cytometry gating strategy showing the phenotype of tetramer-OVA⁺ (dark) or tetramer-OVA^- cells (grey) in the tdLN of KP-OVA tumor-bearing mice. I Quantification of the proportions of Tet-OVA⁺ or Tet-OVA^- CD8⁺ T cell populations in the tdLN of KP-OVA tumor-bearing mice (n = 4 mice, repeated measures two-way ANOVA with Šídák’s multiple comparisons test). Data are presented as stacked mean values ± SD. J Quantification of the absolute numbers of Tet-OVA⁺ or Tet-OVA^- CD8⁺ T cell populations in the tdLN of KP-OVA tumor-bearing mice (n = 4 mice, repeated measures two-way ANOVA with Šídák’s multiple comparisons test). Data are presented as mean values ± SD with individual replicates displayed. K Representative flow cytometry plots showing tdLN Tet-OVA⁺ T_Act subsets at day 17 post KP-OVA tumor injection and chronic isotype, anti-PD-1, anti-CTLA-4, or combined treatment as depicted in Supplementary Fig. 2H. L Quantifications of the proportions of tdLN Tet-OVA⁺ T_Act subsets at day 17 post KP-OVA tumor injection and chronic isotype, anti-PD-1, anti-CTLA-4, or combined treatment as depicted in Supplementary Fig. 2K (n = 7 mice per group, one-way ANOVA with Tukey’s multiple comparisons test). Data are presented as mean values ± SD with individual replicates displayed. KP: (Kras^G12D, p53^-/-) lung adenocarcinoma; tdLN: tumor-draining lymph node; T_Act: activated CD8⁺ T cell; T_RM: tissue-resident memory-like CD8⁺ T cell; T_Ex: exhausted-like CD8⁺ T cell; T_Effector: effector-like CD8⁺ T cell.

In order to identify conserved T_RM markers across anatomical locations in the KP model, we combined our different datasets and found that 38 genes overlapped between lung T_RM-like cluster 1, tdLN Cxcr6⁺ cluster 8, and genes associated to lung TIM3⁺ T_RM-like as compared to TIM3⁺ T_Ex (Supplementary Fig. 4L). We validated at the level of protein expression that JAML, DNAM (encoded by Cd226), and CD49a were indeed expressed at higher levels by lung T_RM and tdLN CD103⁺CXCR6⁺ T_Act as compared to other CD8⁺ T cell subsets (Supplementary Fig. 4M).

We next assessed if ICB immunotherapy would impact on the CD103⁺CXCR6⁺ T_Act compartment. We found that chronic anti-PD-1 and anti-CTLA-4 synergized to increase the frequency of both CD103⁺CXCR6^- and CD103⁺CXCR6⁺ within tdLN Tet-OVA⁺ T_Act (Fig. 3K, L). Together, these results demonstrate that the pool of tdLN CD103⁺ T_Act contains a tumor antigen-specific, immunotherapy-responsive CXCR6⁺ subpopulation with T_RM-like characteristics.

Both XCR1⁺ and CD11b⁺ DCs contribute to the differentiation of CD103⁺ T_Act in tdLN

Because DCs are the main tumor antigen-presenting cells in the LN, we assessed their requirements for the generation of CD103⁺ T_Act in tdLN. We found that KP tumor development triggered a massive increase of both resident and migratory XCR1⁺ DC1s and CD11b⁺ DC2/3s compared to the mediastinal LN from naïve mouse (Fig. 4A–C and Supplementary Fig. 5A). To better characterize DC populations in tdLN, we performed scRNA-Seq of CD11c⁺MHCII⁺ cells 14 days after KP tumor engraftment (Supplementary Fig. 5B–D). We identified 2 clusters of resident DC1 (clusters 7, 14) and 3 clusters of resident DC2 (clusters 3, 9, and 17). Within the clusters expressing high levels of maturation markers (e.g., Ccr7, Cd80, Cd86, Cd200, Cd274, and Pdcd1lg2) (clusters 2, 5, and 8), CITE-Seq protein expression of XCR1 and SIRPα, respectively expressed in DC1 and DC2, enabled to discriminate migratory DC1 (cluster 8) from migratory DC2 (cluster 2), although cluster 5 was composed of a mix of migratory DC1 and DC2 as reported earlier⁴⁷ (Supplementary Fig. 5F). Migratory DC2s (cluster 2) expressed Irf4 (Supplementary Fig. 5E). Fate mapping experiments using Zbtb46^crexRosa^LsL-tdTomato mice demonstrated that most migratory DC2/3s belong to the classical DC lineage⁴⁸ (73.2%) (Supplementary Fig. 5G). By contrast, only a minor fraction of migratory DC2/3s stained positive in GMP/DC3s labeling Ms4a3^crexRosa^{LsL-dTomato49} (11.4%) (Supplementary Fig. 5H) or LyzM^crexRosa^{LsL-dTomato50} (33.0%) (Supplementary Fig. 5I). These results identified migratory CD11b⁺ DCs as a mixture of bona fide DC2s and DC3-related phenotypes. For convenience, DC2/3 will be termed as DC2s hereafter. To assess the contribution of DC subsets in T_RM differentiation, we engrafted KP-OVA tumors in Xcr1^crexRosa^LsL-DTA (XCR1^DTA51) or Itgax^crexIrf4^fl/fl (CD11c^IRF452) mice. XCR1^DTA mice were deficient in DC1s, while CD11c^IRF4 mice had reduced numbers of migratory but not resident DC2s (Fig. 4D and Supplementary Fig. 5J, K). By performing scRNA-seq in CD11c^IRF4 mice, we confirmed that migratory DC2 cluster 2 was reduced as compared to WT mice, while most clusters, including resident DC2 clusters 3, 9, and 17, were left unaffected (Supplementary Fig. 5L, M). Consistent with a well-established role of DC1s in activating CD8⁺ T cell responses^29,30,53,54, we found that Tet-OVA⁺ T_Act were severely reduced in XCR1^DTA mice bearing KP-OVA tumors, including CD103⁺ T_Act (Fig. 4E, F, and Supplementary Fig. 5N). Conversely, depletion of IRF4-dependent DCs in CD11c^IRF4 mice did not impact Tet-OVA⁺ CD8⁺ T cell activation but led to a selective decrease in Tet-OVA⁺CD103⁺ T_Act numbers (Fig. 4E, F, and Supplementary Fig. 5N). Similar results were found after adoptive transfer of OT-1, OVA-specific transgenic CD8⁺ T cells, prior to tumor engraftment (Supplementary Fig. 5O). Analysis of OT-1 in tdLN at day 9 revealed impaired CD103 expression in both XCR1^DTA and CD11c^IRF4 mice (Fig. 4G, H). We thus wondered whether CD103⁺CXCR6⁺ T_Act would be selectively affected by lack of DC populations in XCR1^DTA and CD11c^IRF4 mice bearing KP tumors. We found that all T_Act populations, including CD103⁺CXCR6⁺ T_Act were severely reduced in XCR1^DTA mice (Fig. 4I–K). By contrast, total T_Act numbers were similar in CD11c^IRF4 mice, but both CD103⁺CXCR6^- and CD103⁺CXCR6⁺ T_Act were reduced. Similar results were obtained in AT3-OVA breast tumors in which the absence of DC1 in XCR1^DTA mice broadly impact the activation of all CD8⁺ T cells in tdLNs (Supplementary Fig. 6A, B, E–G). Impairment of migratory DC2s in CD11c^IRF4 mice resulted in a decrease in CD103⁺CXCR6^- and CD103⁺CXCR6⁺ T_Act, although total Tet-OVA⁺ T_Act were also impacted (Supplementary Fig. 6C, D, H–J).

Fig. 4. Both XCR1⁺ and CD11b⁺ DCs contribute to the differentiation of CD103⁺ T_Act in tdLN.

A Representative flow cytometry plot for the gating of CD11c^highMHCII^int resident vs CD11c^intMHCII^high migratory dendritic cells in tdLNs of KP tumor-bearing mice. B Representative flow cytometry plots for the gating of resident DC1, resident DC2, migratory DC1, and migratory DC2s in KP tdLN. C Quantification of the numbers of resident DC1, resident DC2, migratory DC1, and migratory DC2s in the mediastinal lymph node of tumor-free or KP tumor-bearing mice (n = 5 tumor-free and n = 5 KP mice, two-sided unpaired t-tests). Data are presented as mean values ± SD with individual replicates displayed. D Quantification of resident DC1, resident DC2, migratory DC1, and migratory DC2s in the tdLN of WT, XCR1^DTA, and CD11c^IRF4 mice bearing KP tumors (n = 13 WT, n = 4 XCR1^DTA, and n = 7 CD11c^IRF4 mice, one-way ANOVA with Tukey’s multiple comparisons test). Data are presented as mean values ± SD with individual replicates displayed. E Representative flow cytometry plots for the gating of CD103⁺ tetramer-OVA⁺ T_Act in the tdLN of WT, XCR1^DTA and CD11c^IRF4 mice bearing KP-OVA tumors. Black tetramer-OVA⁺ cells overlay grey tetramer-OVA^- cells. F Quantification of CD103⁺ tetramer-OVA⁺ T_Act in the tdLN of WT, XCR1^DTA, and CD11c^IRF4 mice bearing KP-OVA tumors (n = 4 WT, n = 4 XCR1^DTA, and n = 4 CD11c^IRF4 mice, one-way ANOVA with Tukey’s multiple comparisons test). Data are presented as mean values ± SD with individual replicates displayed. G Representative flow cytometry plots for the gating of transferred tdLN CD103⁺ OT1 in the tdLN of WT, XCR1^DTA, and CD11c^IRF4 mice bearing KP-OVA tumors as described in Supplementary Fig. 5O. Black CD45.1⁺ OT1 overlay grey endogenous CD45.2⁺ CD8⁺ T cells. H Quantifications of transferred tdLN CD103⁺ OT1 in the tdLN of WT, XCR1^DTA, and CD11c^IRF4 mice bearing KP-OVA tumors as described in Supplementary Fig. 5O (n = 8 WT, n = 7 XCR1^DTA and n = 4 CD11c^IRF4 mice, one-way ANOVA with Tukey’s multiple comparisons test). Data are presented as mean values ± SD with individual replicates displayed. I Representative flow cytometry plots for the gating of T_Act subsets in the tdLN of WT, XCR1^DTA, or CD11c^IRF4 mice bearing KP-tumors. J Quantifications of total T_Act in WT, XCR1^DTA, or CD11c^IRF4 mice bearing KP-tumors (n = 12 WT, n = 8 XCR1^DTA, and n = 6 CD11c^IRF4 mice, one-way ANOVA with Tukey’s multiple comparisons test). Data are presented as mean values ± SD with individual replicates displayed. K Quantifications of T_Act subsets in WT, XCR1^DTA, or CD11c^IRF4 mice bearing KP-tumors (n = 12 WT, n = 8 XCR1^DTA, and n = 6 CD11c^IRF4 mice, one-way ANOVA with Tukey’s multiple comparisons test). Data are presented as mean values ± SD with individual replicates displayed. L Quantification of resident DC1, resident DC2, migratory DC1, and migratory DC2s in the tdLN of KP-bearing Mgl2^DTR mice treated with PBS or diphtheria toxin at day 8 and analyzed at day 9 post-KP tumor injections (n = 4 Mgl2^DTR + PBS and n = 4 Mgl2^DTR + DT mice, two-sided unpaired t-tests). Data are presented as mean values ± SD with individual replicates displayed. M–O WT or Mgl2^DTR mice bearing KP-OVA tumors were treated with diphtheria toxin at days 2, 5, and 7, and tdLN was analyzed at day 9. M Schematic of the experiment. Created in BioRender. Guermonprez, P. (2026) https://BioRender.com/t3f9mbm. N Representative flow cytometry plots for the gating of tdLN tetramer-OVA⁺ T_Act subsets. Black tetramer-OVA⁺ cells overlay grey tetramer-OVA^- cells. O Quantifications of tdLN tetramer-OVA⁺ T_Act subsets (n = 3 WT + DT and n = 5 Mgl2^DTR + DT mice, two-sided unpaired t-tests). Data are presented as mean values ± SD with individual replicates displayed. DC: dendritic cell; KP: (Kras^G12D, p53^-/-) lung adenocarcinoma; tdLN: tumor-draining lymph node; T_Act: activated CD8⁺ T cell; XCR1^DTA: XCR1^cre x Rosa^{Lox-Stop-Lox-DTA} mice; CD11c^IRF4: CD11c^cre x Irf4^flox mice; DT: diphteria toxin.

We next sought to validate these findings using an alternative strategy to deplete migratory DC2s. Because MGL2 is a lectin notably expressed by lung DC2s⁵⁵, we assessed its expression among tdLN DCs in our model and found that it was specifically expressed by migratory DC2s (Supplementary Fig. 5P). Consistently, DT administration in Mgl2^DTR mice⁵⁵ resulted in substantial depletion of migratory DC2s (Fig. 4L). As a consequence, only tumor-specific Tet-OVA⁺ CD103⁺CXCR6^- and CD103⁺CXCR6⁺ T_Act subsets were reduced in KP-OVA tdLNs, while CD103^- Tet-OVA⁺ T_Act were left unchanged (Fig. 4M–O).

Notably, XCR1^DTA but not CD11c^IRF4 mice displayed increased tumor growth in both the KP-OVA and AT3-OVA models (Supplementary Figs. 5Q and 6A–D). However, the direct contribution of T_RM-like TILs to tumor control in these murine models could not be established due to lack of tools for their specific depletion. Altogether, these results demonstrate that:

- DC1s have a non-redundant function in the activation of all tumor-specific CD8⁺ T cell populations.

- Both DC1s and migratory DC2s are required for the differentiation of CD103⁺CXCR6^+/- T_Act in tdLN.

Efficient generation of T_RM-like TILs requires both DC1s and DC2s at the level of tdLN

We next sought to address the consequences of impaired T_RM specification at the level of tdLN on the generation of lung T_RM-like TILs.

In a first approach, we engrafted KP tumors in WT, Flt3L^-/- deficient in all classical DC subsets, and Ccr2^-/- mice deficient in inflammatory monocytes, and assessed CD8⁺ T_RM-like content. In accordance with a non-redundant role of classical DCs in CD8⁺ T cell immunity^29,48, we found that infiltration of all PD1⁺ activated CD8⁺ T cell subsets, including T_RM-like was deeply inhibited in DC-deficient Flt3L^-/- mice (Supplementary Fig. 7A, B). However, T_RM-like TILs accrual was not modified in Ccr2^-/- mice, excluding a role for monocyte-derived/inflammatory phagocytes in T_RM-like generation. To further define the role of DC1 and DC2, we analysed the differentiation of polyclonal CD8⁺ T cells in XCR1^DTA and CD11c^IRF4 mice engrafted with KP tumors. In XCR1^DTA mice, a major defect in the infiltration of all CD8⁺ T cells subsets was observed (Fig. 5A, B). These results highlight general defects in CD8⁺ T cell activation in DC1-lacking mice^29,32,54 and correlate with the deep impairment in T cell activation observed in tdLN (Fig. 4E–K). By contrast, CD11c^IRF4 mice only had a partial and selective defect in both TIM3^- and TIM3⁺ T_RM-like populations, while other populations of CD8⁺ TILs were left unaffected.

Fig. 5. Efficient generation of T_RM-like TILs requires both DC1s and DC2s at the level of tumor-draining lymph nodes.

A Representative flow cytometry plots for the gating of lung CD8⁺ T cell populations in WT, XCR1^DTA, or CD11c^IRF4 mice bearing KP-tumors. B Quantifications of lung CD8⁺ T cell populations in WT, XCR1^DTA, or CD11c^IRF4 mice bearing KP-tumors (n = 22 WT, n = 10 XCR1^DTA, and n = 22 CD11c^IRF4 mice, one-way ANOVA with Tukey’s multiple comparisons test). Data are presented as mean values ± SD with individual replicates displayed. C, D Representative flow cytometry plots for the gating of lung tetramer-OVA⁺ CD8⁺ T cell populations in WT, XCR1^DTA, or CD11c^IRF4 mice bearing KP-OVA tumors. D Quantifications of lung tetramer-OVA⁺ CD8⁺ T cell populations in WT, XCR1^DTA, or CD11c^IRF4 mice bearing KP-OVA tumors (n = 4 mice per group, one-way ANOVA with Tukey’s multiple comparisons test). Data are presented as mean values ± SD with individual replicates displayed. E Representative flow cytometry plots for the gating of lung tetramer-OVA⁺ CD8⁺ T cell populations in WT or Mgl2^DTR mice bearing KP-OVA tumors and treated with diphtheria toxin as described in Fig. 4M. F Quantifications of lung tetramer-OVA⁺ CD8⁺ T cell populations in WT or Mgl2^DTR mice bearing KP-OVA tumors and treated with diphtheria toxin as described in Fig. 4M (n = 3 WT + DT and n = 5 Mgl2^DTR + DT mice, two-sided unpaired t-tests). Data are presented as mean values ± SD with individual replicates displayed. G Experimental design for the adoptive transfer of total tdLN. Donor CD45.2⁺ mice (WT, XCR1^DTA, or CD11c^IRF4) and host CD45.1⁺ mice were injected with KP cells. Seven days later, total donor tdLN were transferred intravenously in host mice. Lungs from host mice were analyzed 7 days post-transfer. Created in BioRender. Guermonprez, P. (2026) https://BioRender.com/t3f9mbm. H Representative flow cytometry plots for the gating of donor CD45.2⁺ total CD8⁺ T cells and donor CD45.2⁺ T_RM in the lungs of host CD45.1⁺ mice bearing KP tumors 7 days post-transfer of WT, XCR1^DTA, or CD11c^IRF4 tdLN. I Quantifications of donor CD45.2⁺ total CD8⁺T cells and donor CD45.2⁺ T_RM in the lungs of host CD45.1⁺ mice bearing KP tumors 7 days post-transfer of WT, XCR1^DTA, or CD11c^IRF4 tdLN (n = 6 WT, n = 4 XCR1^DTA, and n = 6 CD11c^IRF4 donor mice, one-way ANOVA with Tukey’s multiple comparisons test). Data are presented as mean values ± SD with individual replicates displayed. KP: (Kras^G12D, p53^-/-) lung adenocarcinoma; tdLN: tumor-draining lymph node; T_RM: tissue-resident memory-like CD8⁺ T cell; T_Ex: exhausted-like CD8⁺ T cell; T_Effector: effector-like CD8⁺ T cell.; XCR1^DTA: XCR1^cre x Rosa^{Lox-Stop-Lox-DTA} mice; CD11c^IRF4: CD11c^cre x Irf4^flox mice; DT: diphtheria toxin.

Next, we intended to address the role of DC subsets on CD8⁺ T cell populations in tumor antigen-specific settings using KP-OVA tumors engrafted in WT, XCR1^DTA, and CD11c^IRF4 mice. We found that after 14 days, OVA-specific lung CD8⁺ T cells were severely reduced in XCR1^DTA but not in CD11c^IRF4 mice (Fig. 5C, D, and Supplementary Fig. 7C, D). By contrast, only TIM3⁺ T_RM-like were significantly decreased among Tet^-OVA⁺ CD8⁺ TILs in CD11c^IRF4 mice, evidencing a differential regulation of T_RM-like compared to other CD8⁺ TILs in antigen-specific context as well. Similar conclusions were obtained using an alternative model of DC2 depletion: DT-treated Mgl2^DTR mice displayed reduced T_RM-like among Tet^-OVA⁺ CD8⁺ TILs, while other populations were not significantly impacted (Fig. 5E, F, and Supplementary Fig. 7E, F). In addition, we found that both XCR1^DTA and CD11c^IRF4 mice had reduced T_RM-like among Tet^-OVA⁺ CD8⁺ TILs when engrafted with AT3-OVA tumors (Supplementary Fig. 7G–K).

In sum, we have shown above that both migratory DC subsets are required for the generation of tdLN CD103⁺ T_Act, which are poised to generate T_RM-like TILs (Fig. 4E–O). To assess the contribution of this tdLN-localized phenomenon on lung T_RM-like accrual, we adoptively transferred total CD45.1⁺ tdLN cell suspensions into KP-bearing CD45.2⁺ recipient mice (Supplementary Fig. 7L). Tumor-draining but not unrelated LNs gave rise to T_RM-like TILs in KP recipient mice (Supplementary Fig. 7M–P). We conclude that this system is relevant to test the differentiation potential of activated T cells acting as T_RM-like progenitors. We therefore transferred tdLNs from WT, XCR1^DTA, or CD11c^IRF4 CD45.2⁺ mice to KP-bearing WT CD45.1⁺ host mice (Fig. 5G). This setting allowed us to normalize the lung tumor-microenvironment in recipient mice. Indeed, we observed similar responses in endogenous CD45.1⁺ CD8⁺ T cells (Supplementary Fig. 7Q, R). As anticipated, total donor CD8⁺T cells from XCR1^DTA but not CD11c^IRF4 tdLNs were decreased in host lungs 7 days post-transfer (Fig. 5H, I). However, the generation of T_RM-like TILs was impaired when tdLN progenitors originated from XCR1^DTA or CD11c^IRF4 tdLNs. This finding further substantiates the relevance of CD103⁺ T_Act tdLN cells as progenitors of T_RM-like infiltrating tumors.

These results demonstrate that T_RM-like TILs activation requires the presence of both migratory DC1s and migratory DC2s in tdLN. By contrast, the generation of non-T_RM-like, exhausted CD8⁺ T cells or PD1^- CD8⁺ T cell effector populations is more selectively dependent on DC1s and less impacted by DC2 deficiency.

Low TCR triggering by migratory cross-priming DC2s favors CD103 expression, while migratory DC1-derived IL-12 supports CXCR6 expression

Because both types of migratory DCs were required for T_RM-specification in tdLN, we wondered which could be the mechanism accounting for their respective contribution. We noticed that both migratory DC1 and DC2 (but not resident DCs) had uptaken tumor-associated antigens, although migratory DC1 were more able to transport tumor antigens to the tdLN (Fig. 6A). Migratory DC1 expressed much higher levels of IL-12 (78.3%) than migratory DC2 (15.9%) in IL-12^YFP reporter mice (Fig. 6B) consistent with higher Il12b expression in scRNAseq cluster 8 (Supplementary Fig. 8A). We found that both migratory DC1 and DC2 cross-presented tumor-associated OVA antigen, although DC1s did it more efficiently than DC2s (319.8 vs 132.8 MFI H2K^b-OVA) (Fig. 6C). Of note, no signal was detected on resident DCs. Functionally, we evaluated the cross-presentation ability of migratory or resident DCs sorted from KP-OVA tdLNs by co-cultivating them with cell trace violet (CTV)-loaded OVA-specific OT-1 CD8⁺ T cells. Migratory but not resident DCs were able to induce OT-1 proliferation and CD44 upregulation (66.4% vs 2.82% of CTV^lowCD44⁺) (Supplementary Fig. 8B). Ex vivo addition of synthetic peptide demonstrated that resident DCs are competent in driving OT-1 proliferation when loaded with antigens (Supplementary Fig. 8C, D). We thus focused our study on migratory DC subsets only. FACS-sorted migratory DC1s and DC2s from KP-OVA-bearing tdLNs (hereafter respectively called DC1 KP-OVA and DC2 KP-OVA) were co-cultured with CTV-loaded OT-1 for 2 days (Fig. 6D). We found that DC1 KP-OVA induced a vigorous proliferation of OT-1 that were almost entirely CD103^-, while they expressed homogenously the CD44 activation marker (Fig. 6E, F). By contrast, activation by DC2 KP-OVA was milder, but CD44⁺ proliferating OT-1 expressed high levels of CD103. Addition of DC2 KP-OVA to DC1 KP-OVA OT-1 co-culture did not increase proliferation but promoted CD103 expression on proliferating OT-1 (Fig. 6G, H). This demonstrates that DC1s are the main antigen-presenting cells and DC2s provide factors upregulating CD103 expression. In a setting where DC1 KP-OVA were too few for efficient OT-1 priming, addition of large numbers of DC2 KP-OVA restored OT-1 proliferation (Supplementary Fig. 8E–G). Surprisingly, addition of large numbers of DC2 KP with low numbers of DC1 KP-OVA also sustained efficient OT-1 proliferation, as well as strong CD103 expression (Supplementary Fig. 8E–G). Taken together, these ex vivo cross-presentation assays show that the priming of OT-1 is mainly DC1-dependent and that DC2 can promote CD103 expression in an antigen-independent fashion (Supplementary Fig. 8E–G).

Fig. 6. Low TCR triggering by migratory cross-priming DC2s favors CD103 expression while migratory DC1-derived IL-12 supports CXCR6 expression.

A Representative flow cytometry plot and quantification of ZsGreen⁺ resident and migratory DC1 and DC2 in tdLN of mice bearing KP-ZsGreen tumors (full line) (n = 5 mice, one-way ANOVA with Tukey’s multiple comparisons test). Dotted lines represent control KP-injected mice for each subset. Data are presented as mean values ± SD with individual replicates displayed. B Representative flow cytometry plot and quantification of IL-12-YFP⁺ resident and migratory DC1 and DC2 in tdLN of IL-12^YFP mice bearing KP tumors (n = 5 mice, one-way ANOVA with Tukey’s multiple comparisons test). Data are presented as mean values ± SD with individual replicates displayed. C Representative flow cytometry plot and quantification of H2K^b-OVA expression on resident and migratory DC1 and DC2 in tdLN of mice bearing KP-OVA tumors (full line). H2Kb-OVA MFI were normalized with KP-injected mice (dotted line) (n = 4 mice, one-way ANOVA with Tukey’s multiple comparisons test). Data are presented as mean values ± SD with individual replicates displayed. D Experimental design for the ex vivo cocultures. Migratory DC subsets were FACS-sorted from tdLN of mice bearing KP-OVA tumors at day 9 and cocultured with CTV-loaded naïve OT-1 cells for 2 days. Created in BioRender. Guermonprez, P. (2026) https://BioRender.com/t3f9mbm. E Representative flow cytometry plots of CTV-loaded naïve OT-1 cells cultured for 2 days with flow cytometry-sorted migratory cDC1 or migratory DC2 from tdLN of mice bearing KP-OVA tumors in the presence or not of recombinant IL-12p70 or anti-IL-12p40 neutralizing antibody. F Quantifications of of division and CD103 expression among CTV-loaded naïve OT-1 cells cultured for 2 days with flow cytometry-sorted migratory cDC1 or migratory DC2 from tdLN of mice bearing KP-OVA tumors in presence or not of recombinant IL-12p70 or anti-IL-12p40 neutralizing antibody (n = 6 DC1, n = 1 DC1 + anti-IL12, n = 6 DC2, n = 6 DC2 + recIL12, repeated measures one-way ANOVA with Tukey’s multiple comparisons test). Data are presented as mean values ± SD with individual replicates displayed. G Representative flow cytometry plots for the gating of CTV-loaded naïve OT-1 cells cultured for 2 days with 1250 migratory cDC1 alone or together with 1250 migratory DC2 sorted from tdLN of mice bearing KP-OVA tumors. H Quantifications of division and CD103 expression among CTV-loaded naïve OT-1 cells cultured for 2 days with 1250 migratory cDC1 alone or together with 1250 migratory DC2 sorted from tdLN of mice bearing KP-OVA tumors. (n = 6 DC1, n = 6 DC1 + DC2, two-sided paired t-tests). Data are presented as mean values ± SD with individual replicates displayed. I CTV-loaded OT-1 cells were cultured for 2 days with flow cytometry-sorted migratory DC2 from tdLN of mice bearing KP tumors in the presence of increasing SIINFEKL peptide concentrations (1 pM, 10 pM, 10² pM, and 10⁴ pM). Divided OT-1 and CD103 expression were quantified (n = 4 mice, repeated measures one-way ANOVA with Tukey’s multiple comparisons test). Data are presented as mean values ± SD with individual replicates displayed. See also Supplementary Fig. 8H. J Naïve OT-1 cells were cultured for 4 days with SIINFEKL peptide and low-affinity peptides SIIQFEKL (Q4), SIITFEKL (T4) and SIIGFEKL (G4) at 10² pM. CD44⁺ activated OT1 and CD103 expression were quantified (n = 3 replicates, repeated measures one-way ANOVA with Tukey’s multiple comparisons test). Data are presented as individual paired replicates. See also Supplementary Fig. 8I. K–M Mice were injected i.v. with KP tumors and treated with anti-IL-12p40 neutralizing antibody every 3 days from day 0 to 14, as depicted in Supplementary Fig. 8J. K Representative flow cytometry plots for the gating of CD103⁺CXCR6⁺ T_Act. L Quantifications of T_Act and CD103⁺CXCR6⁺ T_Act (n = 7 control and n = 5 anti-IL12 treated mice, two-sided unpaired t-tests). Data are presented as mean values ± SD with individual replicates displayed. M Quantifications of CXCR6 median fluorescence intensity within CD103⁺CXCR6⁺ T_Act (n = 7 control and n = 5 anti-IL12-treated mice, two-sided unpaired t-test). Data are presented as mean values ± SD with individual replicates displayed. N Representative flow cytometry plots showing CTV-loaded OT-1 cells cultured for 4 days in the presence of increasing SIINFEKL peptide concentrations (1 pM, 10 pM, 10² pM, 10³ pM, and 10⁴ pM) and cytokines TGF-β, IL-12, or both. O Quantifications of CTV-loaded OT-1 cells cultured for 4 days in the presence of increasing SIINFEKL peptide concentrations (1 pM, 10 pM, 10² pM, 10³ pM, and 10⁴ pM) and cytokines TGF-β, IL-12, or both (n = 3 replicates). CD103 and CXCR6 expression were analyzed. Data are presented as mean values ± SD. DC: dendritic cell; KP: (Kras^G12D, p53^-/-) lung adenocarcinoma; tdLN: tumor-draining lymph node; T_Act: activated CD8⁺ T cell; TCR: T cell receptor.

We next wondered which factor could explain the differential ability of DC2s to promote CD103 expression. We showed that IL-12 was highly produced by migratory DC1 (Fig. 6B) and IL-12 silences TGF-β-activated CD103 expression via TBET competition at SMAD binding elements within the CD103 promoter⁵⁶. Therefore, we tested the role of IL-12 ex vivo and found that addition of recombinant IL-12 blunted CD103 expression without affecting the proliferation of naïve OT-1 by DC2s KP-OVA (Fig. 6E, F). By contrast, IL12 blockade during DC1-dependent activation did not restore CD103 on proliferating T cells (Fig. 6E, F). We therefore reasoned that another parameter, distinct from IL-12, was associated to cross-priming by DC1s and limited CD103 expression in activated T cells. We hypothesized that higher levels of MHCI-peptide complexes on DC1s (Fig. 6C) could intrinsically limit CD103 expression on activated T cells. To test this, we loaded DC2 KP with increasing amounts of SIINFEKL synthetic OVA peptide ex vivo in the presence of naïve OT1s. We found that peptide add-back was sufficient to impair CD103 expression in CD44⁺CTV^low OT-1 T cells while it increased proliferative expansion, thereby mimicking T cell activation by DC1s (Fig. 6I and Supplementary Fig. 8H). We next addressed the role of TCR avidity for MHCI-peptide complexes in controlling CD103 persistence on activated T cells. To this end, we primed naïve OT1s with various altered peptide ligands acting as weaker agonists for the OT1 TCR⁵⁷. We observed a decrease of OT-1 activation with low-affinity peptides (SIINFEKL > Q4 > T4 > G4), associated to an increased expression of CD103 on CD44⁺ divided OT-1 cells (Fig. 6J and Supplementary Fig. 8I). Overall, we conclude that a mild TCR triggering and low levels of IL-12 are both required for the persistence of CD103 on T cells during their cross-priming by migratory DCs. However, CXCR6 expression is also a hallmark of early-activated T_RM-like cells in tdLNs (Fig. 3). We found that IL-12 blockade in vivo blunted T cell activation while inhibiting CXCR6 expression in tdLNs (Fig. 6K–M and Supplementary Fig. 8J). Consistently, IL-12 also promoted CXCR6 expression during the priming of naïve OT1 in vitro (Fig. 6N, O). This shows that IL-12, despite its inhibitory effect on CD103, is required to acquire the T_RM-like phenotype. Given the paramount role of TGF-β signaling in the specification of T_RM⁵⁸, we further tested the role of IL-12 in combination with TGF-β in vitro. We found that IL-12 inhibited TGF-β signaling, supporting CD103 expression (Fig. 6N, O). Vice versa, TGF-β counteracted (i) TCR-induced CD103 inhibition and (ii) IL-12-induced CXCR6 expression (Fig. 6N, O). As a result, optimal generation of CD103⁺CXCR6⁺ OT1s required both IL-12 and TGF-β and intermediate levels of antigen presentation, ensuring mild TCR triggering (Fig. 6N, O). Consistently, both migratory DC1 and migratory DC2 expressed αV-β8 and GARP TGF-β presentation molecules, but migratory DC2 displayed significantly higher membrane LAP expression, showing a superior pro-TGF-β production (Supplementary Fig. 8K–M).

Altogether, we conclude that (i) both migratory DC1 and migratory DC2 (but not resident DCs) cross-present KP-associated antigens to T cells in tdLNs. (ii) DC1s are the most efficient at antigen uptake, cross-presentation, and in triggering CD8⁺ T cell proliferative expansion concomitant with the loss of CD103, which is dependent on both efficient TCR triggering and IL-12 release. (iii) IL-12 low DC2s are inefficient at uptake, cross-presentation, and in triggering CD8⁺ T cell proliferative expansion, but the low efficiency of antigen cross-presentation by migratory DC2s is required for CD103 persistence on CD44⁺ activated T cells by delivering a milder TCR signalling as compared to DC1s. (iv) Despite its inhibitory effect on CD103 expression, IL-12 is crucially needed for CXCR6 expression.

Cross-priming by migratory DC2s recapitulates the T_RM transcriptional program

We next sought to analyse how the identity of migratory DCs subset controlled the transcriptional phenotype of CD8⁺ T cells during cross-priming. To this end, we performed bulk RNAseq on OT1 cells that did or did not upregulate CD103 after ex vivo cross-priming by migratory DC1 or DC2 sorted from tdLN of KP-OVA-bearing mice (Fig. 7A). PCA analysis revealed a stronger segregation between DC1 and DC2-activated OT1 across PC1 (84% of the variance) compared to CD103 expression (Fig. 7B). We conclude that the identity of migratory DCs cross-priming T cells is the main factor driving their transcriptional phenotype. PC1 was positively driven by T_RM-associated genes (e.g., Itgae, Cd226) and stemness genes (e.g., Klf2, Sell, Slamf6, and Il7r) and negatively driven by effector-related genes (e.g., Il12rb1 and Il12rb2, Tbx21, Gzmb, Eomes) and proliferation-associated transcription factor Myc. T_RM-associated signatures found in tdLNs pre-T_RMs (by scRNASeq, DEG Cluster 8 Cxcr6⁺ or bulk RNA-Seq, CD103⁺ T_Act > CD103^- T_Act) or lung T_RM-like TILs (bulk RNA-Seq, TIM3⁺ T_RM > TIM3⁺ T_Ex) associated primarily with CD103⁺ OT1 crossprimed by DC2s (CD103⁺ OT1_DC2-Act, Fig. 7C). By contrast, signature of genes downregulated in pre-T_RM (bulk RNA-Seq, CD103^- T_Act > CD103⁺ T_Act) was poorly expressed by CD103⁺ OT1_DC2-Act. GSEA analysis of CD103⁺ OT1_DC2-Act compared to all other OT1 populations, hereafter referred as Rest, further confirmed these results (Fig. 7D). CD103⁺ OT1_DC2-Act were enriched in T_RM-associated genes Itgae, Cd226, Cxcr6, or Inpp4b, stemness-related genes Il7r and Tcf7 and TGF-β receptor genes Tgfbr1 and Tgfbr2 (Fig. 6E). Conversely, CD103⁺ OT1_DC2-Act expressed lower levels of effector-related genes Gzmb, Klrk1, Eomes, or Tbx21, proliferation-associated transcription factor Myc and Il12r genes Il12rb1 and Il12rb2 (Fig. 7E). Consistently, GSEA scoring of a memory CD8 T cell signature revealed that CD103⁺ OT1_DC2-Act had increased expression of memory-related genes (NES = 1.46, Fig. 7F). We also evaluated a signature of SMAD4 target genes, a central mediator of TGF-β signalling, and found that it was enriched in CD103⁺ OT1_DC2-Act (NES = 1.32, Fig. 7F). Conversely, MTORC1 signaling (NES = − 1.58) and MYC targets (NES = − 1.57) signatures were strongly reduced in CD103⁺ OT1_DC2-Act, consistent with reduced proliferation activity (Fig. 7G).

Fig. 7. Cross-priming by migratory DC2s recapitulates the T_RM transcriptional program.

A CTV-loaded naïve OT-1 cells were cultured for 2 days with flow cytometry-sorted migratory DC1 or migratory DC2 from tdLN of mice bearing KP-OVA tumors. CD103⁺ and CD103^- divided OT1 activated by DC1 or DC2 were FACS-sorted, and bulk RNA sequencing was performed. B Principal component analysis (PCA). Top high and low-loading genes are indicated for PC1 and PC2. C Scoring of previously defined gene signatures for tdLN CD103⁺ T_Act (bulk RNA-seq), tdLN cluster 8 Cxcr6⁺ (scRNA-seq), lung T_RM-like TILs (bulk RNA-Seq), and tdLN CD103^- T_Act (bulk RNA-seq) (n = 4 replicates per group, repeated measures one-way ANOVA with Tukey’s multiple comparisons test). Data are presented as mean values ± SD with individual replicates displayed. D GSEA of pairwise comparisons of CD103⁺ OT1_DC2-Act vs all other OT1 populations (termed as Rest) for gene signatures of tdLN CD103⁺ T_Act (bulk RNA-seq), tdLN cluster 8 Cxcr6⁺ (scRNA-seq), lung T_RM-like TILs (bulk RNA-Seq) and tdLN CD103^- T_Act (bulk RNA-seq). Enrichment significance was assessed by phenotype permutation testing, and false discovery rate (FDR) q-values were used to correct for multiple hypothesis testing. E Volcano plot representing differentially expressed genes (DEGs) between CD103⁺ OT1_DC2-Act and all other OT1 populations (Rest) (two-sided Wald test, p-values adjusted for multiple testing using the Benjamini–Hochberg method). Genes with a Log2(fold change, FC) > ± 0.25 and a false discovery rate (FDR)-adjusted p-value of less than 0.05 were considered significant. F GSEA of pairwise comparisons of CD103⁺ OT1_DC2-Act vs all other OT1 populations (Rest) for gene signatures of memory CD8 T cells⁸⁵ and SMAD4 targets⁸⁶ extracted from the MSigDB website. G GSEA of pairwise comparisons of CD103⁺ OT1_DC2-Act vs all other OT1 populations (Rest) for gene signatures of Hallmark MTORC1 signaling and MYC targets⁸⁷ extracted from the MSigDB website. For GSEA analysis in (F, G), enrichment significance was assessed by phenotype permutation testing, and false discovery rate (FDR) q-values were used to correct for multiple hypothesis testing. KP: (Kras^G12D, p53^-/-) lung adenocarcinoma; tdLN: tumor-draining lymph node; T_Act: activated CD8⁺ T cell; T_RM: tissue-resident memory-like CD8⁺ T cell; T_Ex: exhausted-like CD8⁺ T cell; GSEA: gene set enrichment analysis.

Taken together, these results indicate that migratory DC2s recapitulate the differentiation of naïve OT1s into pre-T_RM-like similar to those found in tdLNs of tumor-bearing mice. The specification of the T_RM program within naïve T cells is enabled by a stronger activation of the TGF-β signalling pathway by migratory DC2s as compared to migratory DC1s.

Discussion

Infiltration of solid tumors with CD8⁺ TILs with a T_RM phenotype is predictive of improved survival and better response to immunotherapies^1,3,5–9. Besides, T_RM-like TILs show signs of superior anti-tumor activity as compared to other T cell populations^16,17,20,21. Therefore, we have sought to address the mechanisms governing the acquisition of the T_RM phenotype during T cell activation in tumors. Using a combination of transcriptomic and flow cytometry approaches, we identified a population of T_RM-like TILs specifically expressing canonical T_RM markers and sharing the expression of exhaustion molecules with T_Ex TILs. Fate mapping experiments show that CD103 expression during the early phases of T cell activation in tdLNs defines CD8⁺ T cells committed to generate T_RM-like TILs. Previous studies similarly identified CD103 as a marker of T_RM-commitment within naïve T cells in a model of skin vaccination²⁴. Using unsupervised analysis of activated T cells in the tdLN, we improved this phenotypic characterization and identified a subpopulation of CXCR6⁺CD103⁺ cells with T_RM characteristics. We further addressed the dendritic cell requirements for T_RM phenotype specification. Induction of CD103⁺CXCR6^+/- T_RM-like precursor cells in tdLNs is dependent on both migratory DC1s and DC2s. Consistently, lung T_RM-like TILs accrual requires both types of migratory DCs at the level of tdLN. By contrast, effector and exhausted CD8⁺ T cells infiltrating tumors solely rely on DC1s. Finally, we asked which factors provided by DCs subsets could control T_RM-like TILs generation. We provide evidence that T_RM specification in tdLNs:

• i)Requires sub-optimal TCR triggering by weak cross-presenter DC2s that display reduced densities of MHCI-peptide complexes.
• ii)Is inhibited by high amounts of IL-12 mainly produced by DC1s that inhibits TGF-β signalling and its CD103 downstream target.
• iii)Necessitates some levels of IL-12 as evidenced by the IL-12 function in upregulating CXCR6 during T cell cross-priming.
• iv)Relies on the proliferative expansion uniquely achieved by cross-priming DC1s.

The existence of precursors pre-committed to differentiate into T_RM has been demonstrated within naïve CD8⁺ T cells epigenetically poised to express T_RM-associated genes²⁴ and within blood-circulating activated T cells after skin vaccination²⁶. Mani et al. showed that TGF-β activation by migratory DCs was required for T_RM preconditioning at the level of naïve T cells in the LN. Whether this mechanism requires one or multiple DC subsets has not been addressed. Also, it is unknown which DC subset participate to the activation of T_RM from T_RM-poised naïve progenitors. Our study identifies a subset or early activated CD8⁺ T cells expressing selectively a transcriptional signature of tumor-infiltrating T_RM in tdLN. We demonstrate that the activation of tdLN CD103⁺ CD8⁺ T cells relies on both migratory DC1s and DC2s. By contrast, activation of CD103^- CD8⁺ T cells is largely independent on the presence of migratory DC2s. Furthermore, we demonstrate by adoptive transfer that genetically labelled CD103⁺ T cells from tdLNs uniquely give rise to lung T_RM-like TILs. Therefore, we establish that the specification of lung T_RM-like TILs is initiated within lymph nodes by a process orchestrated by both migratory DCs subsets. This does not exclude the existence of additional pathways of T_RM-like generation that may occur directly within the tissue, which was not addressed in this study.

These findings might have a relevance in humans since CD1c⁺ DCs have been shown to participate to the activation of CD8⁺CD103⁺ T_RM-like cells^39,40. Of note, the contribution of DC2s to the generation of memory but not effector T cell responses had been previously noted in the influenza infection model³⁸ and to the maintenance of anti-infectious T_RM at mucosal sites in HSV infection²⁷. Altogether, our results further extend the major contribution of DC1s in anti-tumor CD8 T cell immunity to T_RM^{29,30,34,53,54}. Our findings are in line with pioneer studies highlighting the crucial role of DC1s in the activation of vaccine-induced T_RMs³⁷. However, it is important to underline that the role of DC2s had not been addressed in this context. In addition, we evidence a contribution of DC2s to T_RM generation at the level of tdLNs.

T_RM development dependency on TGF-β has been well established in several studies^24,25,27. However, both migratory DC subsets express TGF-β activating molecules at the transcriptomic level (e.g., Itgav)²⁴. Therefore, migratory DC2s might contribute to T_RM specification (i) by providing a milder TCR signal strength because they cross-present tumor-associated antigens inefficiently; (ii) by not providing IL-12 in contrast to DC1s; (iii) by providing specific signals yet to be identified. Here, we provide evidence for (i) and (ii). We show that sub-optimal cross-presentation by DC2s enables CD8⁺ T cells to maintain CD103 expression during T cell activation in vitro. These results are in line with previous studies highlighting a negative impact of strong TCR signalling on T_RM differentiation in vivo in the context of viral infections^59,60. More recently, Shakiba et al. have shown that low-affinity tumor-associated antigens prime T cells expressing higher levels of CD103 while presenting less features of dysfunctional/exhausted T cells as compared to T cells primed with high-affinity antigens⁶¹. Finally, both mild TCR signalling⁶² and migratory DC2s contribute to the priming of central memory CD8⁺ T cells⁶³. In addition to optimal cross-priming efficiency, migratory DC1s also provide high levels of IL-12, which acts as an inhibitory signal erasing CD103 expression induced by IRF4-dependent migratory DC2s. These findings are in line with the regulatory role of TBET, downstream of IL-12 signalling at SMAD3 binding sites in the CD103 locus^56,64. Bergsbaken et al. have noticed that IL-12 promotes CD103^- but not TGF-β-dependent CD103⁺ T_RM by a STAT4-dependent mechanism^65,66. Our results clearly establish that the TGF-β-dependent program is induced more efficiently by migratory DC2s than migratory, IL-12-producing DC1s. TGF-β transpresentation might play a role in this process, together with the regulatory role of IL-12 on TGF-β signalling (that can be visualized by IL-12 ability to counteract TGF-β-induced CD103).

T_RM have been extensively studied in the context of infections or vaccination when memory is settled and maintained after antigen clearance. In this study, we have not addressed the functional ability of T_RM-like cells to recirculate or generate memory. Recent studies addressing the functional ability of T_RM-like cells to recirculate in a mouse model of E0771 breast cancer suggest that the T_RM phenotype does not necessarily correlate with tumor residence or correspond to neo-generated T_RM-like cells from circulating precursors¹⁵. However, T_RM-like cells in this model do not express CD103, which may be a major contributor of residence within the tumor through direct interaction with epithelial E-Cadherin, which is not expressed in the E0771 cells investigated in this study^67,68. Evaluating this possibility would require additional experiments.

From a functional perspective, although migratory DC2-deficiency mice did not impair tumor control in our models, the prognostic value of tumor-associated T_RM-like has been well established in multiple studies in humans^{1,3,5–7,9,16} and murine models of cancer^16,20,21. T_RM-like could represent a quiescent pool of responsive cells with poor access to antigens⁶⁹, therefore explaining the lack of responses in the absence of restimulation^3,9,20. Deeper functional studies are required to understand the determinants of the prognosis value of the T_RM-like phenotype in cancer and how their functionality could be activated.

In conclusion, our study highlights the fine-tuning of anti-tumor CD8⁺ T cell diversity by multiple migratory DCs subsets in tdLNs. We show that the cross-priming of CD8⁺ T cells reaching a dysfunctional/exhausted phenotype within tumors is uniquely dependent on DC1s and entirely independent on migratory DC2s. By contrast, induction of T_RM-like cells requires both DC1 and DC2 subsets. These findings challenge the relevance of the “division of labor” across DCs subsets, ascribing discrete, specific, and exclusive immune functions to specific DCs subsets. Instead, we evidence a high level of cooperativity relevant for the induction of memory precursors in tumor-draining lymph nodes. In sum, our findings support the view that instruction of residency program and proliferative expansion rely on complementary yet distinct DCs subsets.

These findings have important implications for the design of effective immunotherapies harnessing multiple types of CD8⁺ T cell immunity, id est effector vs tissue-resident memory cells. One correlates of the current findings that remains to be tested is that optimal priming of diverse T cell phenotypes might rely on antigen delivery to multiple DCs subsets.

Methods

Study design

The aim of this study was to understand the dendritic cell requirements for T_RM specification in lung cancer. To that end, we intravenously injected KP lung tumors into several mouse models in which specific dendritic cell populations can be manipulated. We thoroughly characterized polyclonal and antigen-specific CD8 T cell populations to model antigen OVA using a combination of transcriptomics (bulk RNA-seq, scRNA-seq, and CITE-Seq) and flow cytometry approaches. To address the role of dendritic cells in the lymph nodes, we designed a series of adoptive cell transfer and ex vivo OT1-dendritic cell coculture experiments.

T cell nomenclature

Lung CD44⁺CD62L^-CD8⁺ T cell subsets were initially defined by single-cell RNA-sequencing, then identified using a corresponding flow cytometry gating strategy and validated by bulk RNA-sequencing.

PD1^- T_Eff were defined by their enrichment in the intravascular compartment, high expression of effector-associated signatures and markers (KLRG1), and low expression of exhaustion-associated signatures and markers (TIM3, LAG3, TIGIT, CTLA4, and TOX). CD69⁺CD103⁺ T_RM-like cells were defined by high expression of tissue-residency-associated signatures and markers (CD103, CD69, CD49a, and CXCR6) together with high expression of the transcription factor Hobit. By contrast, T_Ex-like cells were defined as CD69^- and/or CD103^- and displayed higher expression of circulation-associated signatures (including S1pr1 and Gzmk) together with increased expression of the transcription factor EOMES.

Both T_RM-like and T_Ex -like populations were further separated into TIM3^- and TIM3⁺ subsets. TIM3^- subsets were characterized by higher expression of stemness-associated signatures and markers (TCF1, SLAMF6, and CD127), whereas TIM3⁺ subsets exhibited higher expression of exhaustion-associated signatures and markers (LAG3, TIGIT, CTLA4, and TOX).

Tumor-draining lymph node CD8⁺ T cells were initially classified into CD62L⁺CD44^- naive T cells (T_Naive), CD62L⁺CD44⁺ central memory T cells (T_CM), and CD62L^-CD44⁺ activated T cells (T_Act). Based on single-cell RNA-sequencing analysis, T_Act were further subdivided into CD103^- T_Act, CD103⁺CXCR6^- T_Act, and CD103⁺CXCR6⁺ T_Act subsets, which were subsequently characterized by transcriptomic and flow cytometry analyses.

Mice

Mice were housed at the INSERM U1149 CRI, Medicine school site Bichat or at the Paris Institut Pasteur conventional animal house facilities, under controlled environmental conditions with ambient temperature maintained at 22 ± 2 °C, a relative humidity between 40 and 60%, and on a 12 h light/dark cycle. For each experiment, sex-matched animals were used between 8 and 12 weeks of age, and experimental and control groups were co-housed. Mice were euthanized by CO₂ inhalation or cervical dislocation in accordance with institutional guidelines. The study was approved by the local ethics committee, Comité d’éthique Paris Nord no. 121, and the Ministère de l’enseignement supérieur, de la recherche et de l’innovation under the authorization number APAFIS#15373. Animal care and treatment were conducted with national and international laws and policies (European Economic Community Council Directive 86/609; OJL 358; December 12, 1987). All experiments were performed in accordance with the Federation of European Laboratory Animal Science Associations (FELASA) guidelines, institutional guidelines, and French law.

Wild-type C57BL/6J and CD45.1 mice were purchased from Janvier and kept in our facility. XCR1^cre (Xcr1^{tm1(icre)Ciphe51}) and Rosa26-lox-stop-lox-DTA (B6.129P2 Gt(ROSA)26Sor^{tm1(DTA)Lky/J70}) mice were kindly provided by Dr. Marc Dalod (Center of Immunology Marseille-Luminy, Marseille). OT-1 RAG2^-/- CD45.1 mice were kindly provided by Dr. Sébastian Amigorena (Curie Institute, Paris)^71,72. Flt3L^-/- (C57BL/6-Flt3l^{tm1Imx/TacMmjax73}) and Zbtb46^cre (B6.Cg-Zbtb46^{tm3.1(cre)Mnz}/J⁴⁸) mice were kindly provided by Dr. Guillaume Darrasse-Jèze (UMR-S 979, Paris). CD11c^cre (B6.Cg-Tg(Itgax-cre)1-1Reiz/J⁷⁴), IRF4^flox (B6.129S1-Irf4^tm1Rdf/J75), and CCR2^−/− (Ccr2^tm1/fc76) mice were kindly gifted by Dr. Emmanuel Gautier (UMR-S U1166, Paris). Hobit-tdTomato-DTR [B6-Tg (Zfp683-tdTomato-P2A-Cre-P2A-DTR)¹⁸] mice were kindly provided by Dr. Klaas Van Gisbergen (Champalimaud Foundation, Lisbon). The reporter/deleter cassette disrupts the Hobit allele, in consequence, heterozygous Hobit tdTomato-DTR mice were used in the experiments. KRAS^LSL-G12D/+;p53^fl/fl mice (B6.129-Kras^{tm4Tyj Trp53tm1Brn/J77}) were kindly provided by Dr. Kairbaan Hodilvala-Dilke (Bart Institute, Queen Mary University, London). CD103^cre-ERT2 (Itgae-creER^T245) and Rosa^LSL-tdTomato (B6.Cg-Gt(ROSA)26Sor^{tm14(CAG-tdTomato)Hze}/J⁷⁸) mice were kindly provided by Dr. Tessa Bergsbaken (Rutgers, New Brunswick). IL-12^YFP (C.129S4(B6)-^{Il12btm1.1Lky/J79}) mice were kindly provided by Dr. Mikaël Pittet (University of Geneva). Mgl2^DTR (B6(FVB)-Mgl2^{tm1.1(HBEGF/EGFP)Aiwsk}/J⁵⁵) mice were kindly provided by Dr. Yosuke Kumamoto (Rutgers, New Jersey). Ms4a3^cre (C57BL/6J-Ms4a3^em2(cre)Fgnx/J⁸⁰) mice were kindly provided by Dr. Camille Bleriot (Institut Gustave Roussy, Paris) and Dr. Florent Ginhoux (Institut Gustave Roussy, Paris). Lyz2^cre (B6.129P2-Lyz2^tm1(cre)Ifo/J⁸¹) mice were kindly provided by Dr. Loredana Saveanu (INSERM U1149, Paris).

Tumor cell lines

The KP line has been isolated from primary lung tumors of C57BL/6 KP mice (KRAS^LSL-G12D/+;p53^fl/fl)⁸². The line was kindly provided by Dr. Federica Benvenuti (ICGEB, Trieste) and has been used in other studies^35,83. KP-OVA were generated by retroviral transduction of KP cells with a pBabe-OVA-IRES-Cherry vector. KP-ZsGreen were generated by lentiviral transduction of KP cells with a pLVX-IRES-ZsGreen vector. Tumor cell lines were cultured in RPMI 1640 medium, Glutamax (Thermo Fisher) supplemented with 10% fetal bovine serum (FBS) (Thermo Fisher), penicillin-streptomycin (Thermo Fisher), and 55 μM β-mercaptoethanol (Thermo Fisher) (complete RPMI), and maintained at 37 °C and 5% CO₂. The AT3-OVA line was kindly provided by Dr. Julie Helft (Institut Cochin, Paris).

Tumor models

For the grafted KP tumor models, the KP, KP-OVA, and KP-ZsGreen tumor cell lines were grown in RPMI-Glutamax with 10% FCS, 55 μM 2-mercaptoethanol, penicillin streptomycine and used for experiments when in the exponential growth phase. The cells were detached using Trypsine/EDTA 0.25%, washed two times in PBS, and counted in trypan blue. 8 × 10⁵ tumor cells were intravenously injected in 100 µL RPMI without supplement.

For the autochthonous KP model, C57BL/6 KP mice (KRAS^LSL-G12D/+;p53^fl/fl) at 8 weeks of age were intranasally inoculated with 2.5 × 10⁷ infectious particles of a replication-deficient adenoviral vector with a Cytomegalovirus promoter driving the expression of the Cre recombinase protein in order to sporadically induce mutations in lung cells and promote lung tumor development^41,42,77. Mice were euthanized 16 weeks after the adenovirus inoculation.

For the AT3-OVA orthotopic tumor model, cells were grown in RPMI-Glutamax with 10% FCS, 55 μM 2-mercaptoethanol, penicillin streptomycine and used for experiments when in the exponential growth phase. The cells were detached using Trypsin/EDTA 0.25%, washed two times in PBS, and counted in trypan blue. 5 × 10⁵ tumor cells were injected orthotopically in the mammary gland in 100 µL RPMI without supplement. Tumor dimensions were recorded every 2–3 days using a digital caliper, and tumor volume (V) was calculated using the formula V = (L × W × W)/2, with volume expressed in mm³. Mice were monitored daily, and endpoints were established based on signs of distress, including pain, lethargy, dehydration, ulceration, or tumor necrosis. Mice were euthanized when these criteria were met or when tumor volume reached 1500 mm³.

In vivo treatments

For the FTY treatment, mice received intraperitoneal injections of 1 mg/kg FTY720 (fingolimod) (Sigma-Aldrich) in 150 μl of non-deionised H₂0 (Versol) every 3 days from day 1 after tumor injection.

For IL-12p40 blocking, mice received intraperitoneal injections of 200 µg anti-IL-12p40 blocking antibody (Biolegend) in 200 µl of PBS every 3 days from day 0 after tumor injection.

For chronic ICB immunotherapies, mice received intraperitoneal injections of 200 µg control isotype (clone 2A3, BioXCell BE0089), anti-CTLA-4 (clone 9H10, BioXCell BE0131), anti-PD-1 (clone 29F.1A12, BioxCell BE0273), or both anti-PD-1 and anti-CTLA-4 antibodies in 200 µl of PBS every 3 days from day 7 after KP-OVA tumor injection and mice were euthanized at day 17 for T cell analysis.

For acute ICB immunotherapy, mice received one intraperitoneal injection of 200 µg anti-PD-1 antibody at day 14 post-tumor injection, and mice were euthanized 2 days later (day 16).

For diphtheria toxin (DT) treatments, Mgl2^DTR mice were injected intraperitoneally with PBS or 700 ng DT (Sigma D0564) in 200 µl of PBS at day 8 post-KP tumor injection, and depletion efficiency was assessed 1 day later (day 9). For chronic depletion, WT or Mgl2^DTR mice were injected with DT at days 2, 5, and 7 post-KP-OVA tumor injection, and T cells were analyzed at day 9.

Mouse tissues processing

Mouse lungs were harvested and transferred to 3 ml digestion buffer (Hank’s Balanced Salt Solution (HBSS) with calcium and magnesium (Thermo Fisher) and with 75 μg/mL of Liberase TL (Roche) and 0.02 mg/ml DNase I (Thermo Fisher)). Lungs were dissociated using gentleMACS Octo Dissociator (Miltenyi) (program m_lung_01), incubated at 37 °C for 45 min and dissociated again on the gentleMACS Octo Dissociator (Miltenyi) (program m_lung_02). AT3-OVA tumors were harvested, mechanically dissociated, and similarly digested. The cell suspensions were passed through 70 μm cell strainers (Corning), and red blood cells were lysed using ACK lysing buffer (Thermo Fisher). The mediastinal lymph nodes were smashed in FACS buffer on a 70 μm cell strainer. The absolute number of live immune cells in each tissue cell suspension was assessed using AccuCheck Counting Beads (Thermo Fisher) along with anti-CD45 and DAPI staining on BD FACS Fortessa 20 (BD Biosciences). For OT-1 CD8⁺T cell preparation, LNs from OT-1 CD45.1 RAG mice were passed through 70 μm cell strainers. The LN suspensions were stained with anti-CD8-APCCy7 and counted by FACS using counting beads (AccuCheck beads, Invitrogen).

Flow cytometry analysis

After tissue processing and cell counting, Fc-receptor were blocked using FcBlock (2.4G2, BD) for 15 min at 4 °C. For dead cell identification, the cells were stained with 7-AAD (BioLegend) for extracellular staining only or Zombie dye (BioLegend) for intracellular staining in PBS. Cells were then stained with fluorophore-conjugated antibodies in FACS buffer (see antibody list in the Key resource table), PBS 3% FCS (Gibco) 2 mM EDTA (Gibco), during 30 min at 4 °C. When needed, cell suspensions were subsequently fixed and stained in for nuclear protein (30 min fixation and 30 min intracellular staining) using a fixation/permeabilization kit (eBiosciences transcription factor staining buffer) according to the manufacturer’s instructions. Labeling with H2K^b-SIINFEKL tetramers (MBL) was performed at room temperature for 30 min. Data acquisition was performed using an LSR-Fortessa X20 (BD), and analysis were done using FlowJo software (TreeStar).

Ex vivo T cell restimulation assay

After tissue processing and cell counting, 1 million total CD45⁺ cells were incubated for 5 h at 37 °C in a restimulation cocktail composed of 50 ng/ml PMA (Sigma), 1 µg/ml Ionomycin (Sigma), 2 µg/ml Golgi Plug (BD), and 2 µg/ml Golgi stop (BD) in complete RPMI medium. Cells were then processed for flow cytometry analysis as described before.

Ex vivo tumor-T cell coculture assay

1000 KP-OVA tumor cells were seeded in 96-well plates. One day later, 3000 PD1⁺ T_Ex or 3000 PD1⁺ T_RM-like were sorted from the lungs of KP-OVA-bearing mice at day 14 and cocultured for 4 days with KP-OVA tumors cells in complete RPMI medium. Four days later, supernatants were harvested for ELISA quantification of IFNγ (BioLegend 430804) and Granzyme B (R&D DY1865). T cells were harvested, and KP-OVA tumor cells were detached using Trypsin/EDTA 0.25% and counted by FACS.

Bulk RNA-sequencing

After tissue processing and cell count, cell suspensions were washed in cold FACS buffer (PBS + 3% BSA + 2 mM EDTA), Fc-receptors were blocked 15 min at 4 °C, using purified anti-CD16/32 (2.4G2), and then stained with fluorophore-conjugated antibodies for 30 min at 4 °C. CD8⁺ T cell subsets from tumor-bearing lungs and tdLNs were isolated using an Aria III (BD) and a MELODY (BD) cell sorter, respectively, and directly collected on lysing TCL buffer (QIAGEN) containing 0.1% of beta-mercaptoethanol before storage at −80 °C. RNA were extracted and isolated using the Single Cell RNA purification kit (Norgen, Cat#51800) according to the manufacturer’s instructions. After extraction, total RNA was analyzed using the Agilent RNA 6000 Pico Kit on the Agilent 2100 Bioanalyzer System. RNA quality was estimated based on capillary electrophoresis profiles using the RNA integrity number (RIN). RNA sequencing libraries were prepared using the SMARTer Stranded Total RNA-Seq Kit v2 - Pico Input Mammalian (Takara). The input quantity of total RNA was comprised between 1 and 22 ng. This protocol includes a first step of RNA fragmentation, using a proprietary fragmentation mix at 94 °C. The time of incubation was set up for each sample based on the RNA quality and according to the manufacturer’s recommendations. After fragmentation, indexed cDNA synthesis was performed. Then the ribodepletion step was performed, using probes specific to mammalian rRNA. PCR amplification was finally achieved to amplify the indexed cDNA libraries, with a number of cycles set up according to the input quantity of tRNA. Library quantification and quality assessment was performed using a Qubit fluorometric assay (Invitrogen) with the dsDNA HS (High Sensitivity) Assay Kit and LabChip GX Touch using a High Sensitivity DNA chip (Perkin Elmer). Libraries were then equimolarly pooled and quantified by qPCR using the KAPA library quantification kit (Roche). Sequencing was carried out on the NovaSeq 6000 (Illumina), targeting between 10 and 15 M reads per sample and using paired-end 2 ×100 bp. Subsequent analysis were performed using the conventional DESeq2 pipeline.

Gene set enrichment analysis

To statistically evaluate the enrichment of previously reported gene signatures and DEG calculated from pairwise comparison in this study (Gene Sets), we used pairwise comparisons using the gene set enrichment analysis (GSEA)⁸⁴ method from the Massachusetts Institute of Technology (https://www.broadinstitute.org/gsea). Statistical analysis was performed by evaluation of nominal p-value and false discovery rate (q-value) based on 1000 random permutations. Results were considered significant when the p-value was below 0.05 and when the q-value was below 0.25 (false discovery rate below 25%) accordingly to the recommendation from the software developers.

scRNA-seq analysis

For cell sorting experiments, tumor-bearing lungs and tdLN were collected and processed as described above. The strategy used for the sort is depicted in the figures. The droplet-based approach of the 10× Genomics platform was used to perform scRNA-seq. For each scRNA-seq sample, cell suspensions from three individual mice were pooled and sorted in PBS containing 0.04% BSA. Ten thousand cells were loaded on a Chromium Controller using the Chromium Next GEM Single Cell 3 Reagent Kits v3.1 according to manufacturer instructions.

Raw sequencing reads were processed using the 10x Genomics software Cellranger. To obtain a cell count matrix, reads were mapped to the mouse GRCm38.84 reference genome. scRNA-seq analysis was performed using Seurat v4. For each dataset, cells with low genes per cell or with a high percentage of mitochondrial genes were removed from downstream analyses. Following data QC, we obtained 4421 cells from the lungs KP CD8⁺ T cells dataset, 1404 cells for the tdLN KP WT CD8⁺ T cells dataset, 3340 cells for the tdLN KP CD11c^IRF4 CD8⁺ T cells dataset, 2076 cells for the tdLN KP WT CD11c⁺MHCII⁺ cells dataset and 3150 cells for the tdLN KP CD11c^IRF4 CD11c⁺MHCII⁺ cells dataset. RNA and ADT expression were normalized separately using the default Seurat LogNormalize and Centered Log Ratio (CLR) approaches, respectively. The top 2000 most variable genes were then determined by applying a VST (Variance Stabilizing Transformation). Datasets were analyzed separately or were integrated together before analysis. For integration, the top 2000 most variable genes across datasets were used as anchors with the LogNormalize method. Dimensionality reduction was performed using PCA, and the top informative principal components were determined after visualization with an elbow plot and dimensional reduction heatmaps, and used for the non-linear dimensionality reduction technique UMAP. Louvain clustering of the data was performed across a range of resolution parameters visualized using a clustree. Cluster of cells with low number of genes and in which no T lymphocyte cell-identifying genes could be detected were not considered in the analysis. Subsequent analyses were performed using Seurat default parameters.

Adoptive transfers

Mediastinal lymph nodes from donor mice (CD45.2⁺ or CD45.1⁺ tumor-bearing or CD45.1⁺ OTI x Rag1^-/- tumor-free mice) were mechanically dissociated on a 70 μm cell strainers in sterile FACS buffer. After two washes in PBS, the LN cell suspensions were counted and resuspended in 100 μL of RPMI without supplement and intravenously injected in recipient mice (CD45.1⁺ or CD45.2⁺ tumor-bearing mice), according to the experimental designs described in Figs. 2K and 5G, Supplementary Figs. 5O, 7L.

Ex vivo mouse T cell priming assay

OT-1 cell preparation

OT-1 CD8⁺ T cells were isolated from lymph nodes of CD45.1⁺ OTI x Rag1^-/- mice and labeled for proliferation assay with 2 μM of CellTrace Violet (CTV, Invitrogen) at 37 °C for 15 min, washed, and counted before culture with DC subsets.

Dendritic cells isolation from LNs

For the isolation of mediastinal lymph node DCs from KP- or KP-OVA-bearing mice, minced mediastinal tumor-draining LNs were digested in 500 μl HBSS with calcium and magnesium (Thermo Fisher) with 75 μg/mL of Liberase TL (Roche) and 0.02 mg/ml DNase I (Thermo Fisher) and incubated at 37 °C for 20 min under agitation. Cell suspensions were passed through a 70 μm cell strainers. Cell suspensions were then blocked in FcBlock and stained with CD11c-biotin followed by SAV-beads (Miltenyi). Cell suspension was applied on a LS column (Miltenyi) according to the manufacturer’s instructions. The CD11c-enriched fraction retained in the column was then stained for CD11c, MHCII, CD19, XCR1, and CD11b. Dendritic cells subsets were FACS-sorted on a Melody (BD) according to the gating strategy depicted in Fig. 4A, B.

DCs and T cell co-culture

2500 DCs were plated in V-bottom 96-well plates and incubated with 25000 CTV-loaded OT-1 in RPMI medium (10% FCS + 1% pencillin/streptomycin + 1 mM sodium pyruvate 55 μM 2-mercaptoethanol and 10 mM HEPES). When indicated, various doses of the SIINFEKL OVA peptide or mutated peptides (Q4, T4, and G4) were added to the cultures at the indicated concentration. For some experiments, recombinant mouse TGF-β (20 μg/mL) (BioLegend), recombinant mouse IL-12p70 (20 μg/mL) (BioLegend), or anti-IL-12p40 blocking antibody (20 μg/mL) (BioLegend) were added to the well.

T cell activation analysis by flow cytometry

After 48 h or 96 h of co-culture as indicated in the figures, T cells were stained for CD45.1, CD8, CD44, CD62L, PD-1, and CD103 to analyze their activation and polarization status. CTV dilution was analyzed to assess T cell proliferation.

Quantification and statistical analysis

Statistical analyses were performed using Prism 10 (GraphPad Software Inc., USA). Each dot displayed on the graphs represents one biological replicate, and error bars indicate standard deviation (SD). Comparisons between two experimental groups were performed using two-sided unpaired or paired Student’s t tests, as appropriate. Comparisons among three or more groups were performed using one-way or two-way analysis of variance (ANOVA) (with or without repeated measures), as appropriate. When one-way ANOVA revealed significant differences, Tukey’s multiple comparisons test was used for post hoc pairwise comparisons. When two-way ANOVA revealed significant differences, Šídák’s multiple comparisons test was used for post hoc comparisons. A p < 0.05 was considered statistically significant, and exact p-values are reported in the corresponding figure panels.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization: N.V., P.B., and P.G. Methodology: N.V., P.B., and P.G. Investigation: N.V., P.B., M.S., A.O., L.G., A.C., M.R., A.S., R.S., Y.G., F.L.R.D.C., M.V., A.R., J.B., A.A., M.G., J.W., S.B., K.P., Z.G., and Ph.B. Software: N.V., A.A., S.B., and Y.G. Ressources: F.G., C.B., Y.K., G.D.J., E.L.G., M.D., T.B., K.V.G., K.H., M.M., L.S., J.H., F.B., and P.G. Funding acquisition: E.T., J.H., and P.G. Supervision: L.S., J.H., F.B., and P.G. Writing: N.V., P.B., and P.G.

Peer review

Peer review information

Nature Communications thanks Francesca Granucci, Lukas Heger, and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

P.G. is a CNRS investigator (Centre National de la Recherche Scientifique). N.V. is supported by a PhD fellowship from Université de Paris and by a PhD fellowship from Fondation pour la Recherche Médicale (FRM). P.B. is supported by a postdoctoral fellowship from Fondation pour la Recherche sur le Cancer (ARC). This work was supported by grants from Cancer Research UK, Institut National du Cancer (INCA PL-BIO22-147), Fondation pour la Recherche sur le Cancer (ARC PJA2021060003913), Fondation pour la Recherche Médicale (EQU202203014687), Fondation Chercher Trouver, Fondation BMS for Immunooncology, Université Paris Cité Infibrex “Initiative of Excellence” IdEx program (ANR-18-IDEX-0001).

Data availability

The bulk and single-cell RNA-seq datasets have been deposited in GEO under the accession codes GSE337601 and GSE337798. All data are included in the Supplementary Information or available from the authors, as are the unique reagents used in this Article. The raw numbers for charts and graphs are available in the Source data file whenever possible. Differential gene expression analyses and signature lists are available in the Supplementary Files. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77641-7.