Ethical approvals

Studies conducted at the Swiss Federal Institute of Technology in Lausanne (EPFL; De Palma laboratory) were approved by the Veterinary Authorities of the Canton Vaud according to Swiss law (protocols VD3154, VD3154.1 and VD3785). Studies conducted at the University of Geneva (UNIGE; Migliorini laboratory; glioma models) were approved by the Veterinary Authorities of the Canton Geneva according to Swiss law (protocol VD3717c). Studies conducted at the German Cancer Research Center (DKFZ; Heikenwälder laboratory; liver cancer models) were approved by the Regierungspräsidium Karlsruhe according to German law (protocols G275/18, G5/19 and DKFZ332). All studies were compliant with the humane endpoints established in the above authorizations.

Design and construction of LVs

To generate bicistronic constructs for the expression of either FLT3L or IL-12 in combination with a marker gene (GFP or dLNGFR), or for the co-expression of FLT3L and IL-12 without marker genes, we used the P2A peptide70. To co-express FLT3L and GFP, we modified an LV transfer construct containing the spleen focus-forming virus (SFFV) promoter7. A synthetic complementary DNA (cDNA) sequence encoding for the mouse FLT3L, in which an IgK secretion signal (MDFQVQIFSFLLISASVIMSRG) replaced the native signal peptide, was obtained from GenScript and cloned downstream to the SFFV promoter. Then, a P2A-GFP sequence was cloned downstream to the FLT3L cDNA to obtain the ‘FLT3L-P2A-GFP’ LV. To co-express IL-12 and GFP, we obtained from GenScript a synthetic cDNA sequence encoding for the single-chain bioactive murine IL-12 containing the P40 and P35 subunits separated by a linker71. The IL-12 sequence was then cloned upstream to the P2A-GFP sequence to obtain the ‘IL-12-P2A-GFP’ LV. To co-express IL-12 with dLNGFR or FLT3L, we obtained full-length, mouse-optimized DNA sequences from GenScript and cloned them downstream to the SFFV promoter to obtain the ‘IL-12-P2A-dLNGFR’ and ‘IL-12-P2A-FLT3L’ LVs. Monocistronic LVs expressing GFP or dLNGFR from the SFFV promoter were described previously7.

LV production and titration

Third-generation self-inactivating LV particles were produced according to published protocols72,73. In brief, 293T cells were transiently transfected with a mix of packaging plasmids and the desired transfer construct as described previously72,73. Conditioned medium was collected after 48 h and 72 h and concentrated by ultracentrifugation with a Beckman ultracentrifuge72,73. To titrate concentrated LVs, 293T cells were transduced by serially diluted LV particles. The frequency of marker gene (GFP or dLNGFR/CD271)-positive cells was measured by flow cytometry 5–8 days after transduction, and the titer was calculated as described previously73. The IL-12-P2A-FLT3L LV was titrated by ELISA of the capsid protein p24 (OriGene), IL-12 (BD Biosciences) or FLT3L (Invitrogen).

Design and production of retroviral vectors

The anti-GD2 CAR retroviral vector was generated as follows. The murine-optimized anti-GD2 scFv 14g2a was obtained from GenScript. The scFv fragment was cloned in the pMSGV retroviral vector in frame with mouse CD8a hinge and transmembrane segments, mouse 4–1BB intracellular domain and mouse CD3ζ intracellular domain. Phoenix-Eco cells were transfected with the GD2 CAR plasmid and pCL-Eco-packaging plasmid using lipofectamine (Invitrogen). Cell culture supernatant containing retroviral vector particles was collected after 48 h and 72 h and concentrated as described above for LV particles.

Cell lines

The 293T cells were obtained from L. Naldini (San Raffaele Institute, Milan, Italy). Phoenix-Eco cells were obtained from ATCC (cat. no. CRL-3214). MC38 colorectal carcinoma cells and B16F10 melanoma cells modified to express OVA were obtained from P. Romero (University of Lausanne, Switzerland). SB28 glioma cells were generated and provided by H. Okada (University of California, San Francisco, CA, USA); these cells express both luciferase and GFP. Cell lines were cultured in DMEM (ThermoFisher) with 10% FBS (Gibco), 1% l-glutamine (Gibco) and 1% penicillin–streptomycin (Pen–Strep; ThermoFisher). While original stocks of cancer cell lines were authenticated, we did not perform further authentication in the past several years. However, the cell lines appeared authentic based on morphology and growth behavior in vitro and in vivo, especially with respect to the ability to form tumors in mice, expression of defined fluorescent or bioluminescent genes and gene expression. In particular, 293T and Phoenix-Eco cells supported high-titer LV and retroviral vector production; B16F10 cells were validated as melanoma cells by RNA-seq analysis; and SB28 cells expressed luciferase and GFP and formed invasive gliomas in mice. All cell lines were negative for Mycoplasma contamination in tests performed in the laboratory.

Modification of cell lines

SB28 cells were modified to express GD2 by transduction of cancer cells with LVs encoding GD2S and GD3S synthases, as described previously7. GD2+ (transduced) cancer cells were sorted using a FACSAria II SORP (Becton Dickinson).

To generate Ifngr1-knockout B16F10 cancer cells, we used a previously described CRISPR–CAS9 system that does not involve stable expression of immunogenic CAS9 or resistance genes by the modified cells74. The guide RNA sequences were as follows:

Ifngr1-KO_Fw: CACCGATTAGAACATTCGTCGGTAC

Ifngr1-KO_Rv: CTAATCTTGTAAGCAGCCATGCAAA

The targeting plasmid was transfected in parental B16F10 cells (which express OVA) using lipofectamine (Invitrogen), and GFP+ cells were sorted 72 h after transfection. Cells were kept in culture and passaged until expression of GFP was lost. To isolate Ifngr1-knockout cells, sorted cells were stimulated with IFNγ (PeproTech; 20 ng ml–1) for 24 h to induce expression of H-2Kb on the Ifngr1-proficient population7 and stained with an antibody to H-2Kb/SIINFEKL (BioLegend). H-2Kb/SIINFEKL-negative cells were then sorted. To further validate loss of IFNGR expression in Ifngr1-knockout cells, sorted Ifngr1-knockout and Ifngr1-proficient cells were treated with IFNγ or left untreated and then stained with an anti-B2m antibody (BioLegend) to assess responsiveness to IFNγ, or lack thereof.

Isolation of mouse BM cells

Long bones of CD45.1 or CD45.2 C57BL/6 mice were excised and decontaminated. BM cells were isolated by a pulse of high-speed centrifugation and passed through a 70 μm cell strainer to remove debris and aggregates. Red blood cells were removed with RBC Lysis Buffer (Sigma) and BM cells were filtered to obtain single-cell suspensions. BM marrow cells were washed once in complete RPMI 1640 Medium (ThermoFisher) supplemented with 10% FBS (Gibco), 1% glutamine and 1% Pen–Strep before transfer to specific media.

Generation of mouse DCPs

DCPs were generated from mouse BM cells isolated from long bones as described above. The two-step procedure involved short-term expansion of HSPCs followed by partial differentiation under conditions that promote CDP and cDC1 lineage commitment14. For the expansion phase, BM cells were cultured at a density of 2 × 106–3 × 106 cells per ml in 10 cm plates for 2–3 days in complete RMPI medium containing recombinant murine cytokines: 100 ng ml–1 SCF, 40 ng ml–1 TPO, 50 ng ml–1 FLT3L, 30 ng ml–1 IL-3, 30 ng ml–1 IL-6 and 30 ng ml–1 IL-1b (all from PeproTech) (referred to as ‘HSPC medium’). For the differentiation phase, floating cells were then collected and cultured at a density of 2–3 × 106 cells per ml in six-well plates for 4–5 days in complete RMPI medium containing 200 ng ml–1 FLT3L and 5 ng ml–1 GM–CSF (PeproTech) (cDC1 medium). The DCPs were then enriched by depleting lineage-positive cells using EasySep Mouse Hematopoietic Progenitor Cell Isolation Kit (STEMCELL Technologies).

Differentiation of DCPs in the presence of T cell cytokines

Enriched DCPs (106 cells per ml) were resuspended in ‘cDC1 medium’ (complete RMPI medium with 200 ng ml–1 FLT3L and 5 ng ml–1 GM–CSF) supplemented with various recombinant murine cytokines (IL-2, IL-12, IL-15, IL-18, IL-21, IL-23, IL-27 or none; all from PeproTech except for IL-18, IL-23 and IL-27, from BioLegend). Each cytokine was tested at the concentration of 2 ng ml–1, 8 ng ml–1 or 20 ng ml–1. The phenotype of DCP-derived cells was examined after 15 days of culture.

Generation of mouse moDCs and cDC1-like cells

To generate moDCs, BM cells were cultured in complete RPMI supplemented with 100 ng ml–1 GM–CSF and 40 ng ml–1 IL-4 (referred to as ‘moDC medium’) at a concentration of 2 × 106–3 × 106 cells per ml for 2 days, as previously described7,15. Non-adherent and loosely adherent cells were gently collected, replated at a concentration of 2 × 106–3 × 106 cells per ml in six-well plates and cultured for four to six additional days. To generate cDC1-like cells, we used a published protocol14 with some modifications. In brief, BM cells were cultured in cDC1 medium at a density of 2 × 106–3 × 106 cells per ml in six-well plates for up to 14–18 days but adding fresh medium every 3–4 days.

Co-culture of OT-I and OT-II T cells with antigen-loaded cDC1-like cells

BM cells were differentiated into cDC1-like cells and incubated for 4 h in 96-well U-bottom plates (5 × 104 per well) in 200 µl of medium in the presence of 2 mg OVA protein (10 mg ml–1; vac-stova; InvivoGen) and individual interleukins (IL-2, IL-12, IL-15, IL-18, IL-21, IL-23 or IL-27). The concentration of each interleukin was 2 ng ml–1, 8 ng ml–1 or 20 ng ml–1 (in 200 µl). Meanwhile, OT-I CD8+ and OT-II CD4+ cells were isolated using EasySep kits (STEMCELL Technologies) and resuspended in T cell medium (complete RPMI medium with 50 μM beta-mercaptoethanol, minimal non-essential amino acids, 1 mM sodium pyruvate and 10 mM HEPES). OVA-loaded cDC1-like cells were then washed and co-cultured with OT-I or OT-II cells in T cell medium in the presence of the aforementioned interleukins. Co-culture of OVA-loaded cDC1-like cells with OT-I or OT-II cells was continued for 3 days and 5 days, respectively. T cell activation was measured by intracellular staining with antibodies against IFNγ (clone XMG1.1, BD Biosciences) using BD Golgi Stop kit (BD Biosciences).

Mouse DCP and moDC transduction with LVs

Freshly enriched DCPs were cultured in cDC1 medium at a concentration of 1.5 × 106 DCPs per ml in six-well plates and transduced 2 h later with LVs at the multiplicity of infection of 350 (determined by GFP or dLNGFR titer, or ELISA, as described above). Non-adherent or loosely attached moDCs were collected, plated in six-well plates at a concentration of 1.5 × 106 moDCs per ml and transduced with LVs at the multiplicity of infection of 100. Transduction was measured 3 days after transduction by flow cytometry or ELISA. For adoptive transfer to mice, transduced cells were collected 12–14 h after transduction and resuspended in PBS before infusion.

CAR-T cell production

Spleens from C57BL/6 mice were smashed on a 70 μm cell strainer. CD8+ T cells were purified using the EasySep Mouse T Cell Isolation Kit (STEMCELL Technologies), then 0.5 × 106 T cells were seeded in 48-well plates in complete RPMI medium supplemented with 10% fetal calf serum (FCS) and 50 IU ml–1 rhIL-2 (Bio-Techne). T cells were activated with Activator CD3/CD28 Dynabeads (Gibco/ThermoFisher) at a ratio of two beads per cell. Retroviral vector transduction was conducted after 24 h in 48-well plates coated with 20 μg ml–1 recombinant human fibronectin (Takara Clontech). The medium was replaced on day three with medium containing 10 IU ml–1 rhIL-2, 10 ng ml–1 rhIL-7 (PeproTech) and 10 ng ml–1 rhIL-15 (PeproTech). Cells were passaged every second day. CAR expression was determined on day eight by flow cytometry before T cell collection for downstream experiments.

CAR-T cell killing assay

SB28 cells, both unmodified or modified to express GD2, were seeded in 96-well plates at 5 × 104 cells per well. Cells were allowed to adhere to the plate for 2 h, followed by the addition of non-transduced T cells and anti-GD2 (14g2a) mouse CAR-T cells. Three different effector:target ratios were tested: 1:1, 5:1 and 10:1. The cytotoxic assay was conducted for 72 h, then the proportion of dead cancer cells was evaluated by flow cytometry. SB28 cells were identified by GFP expression while dead cells were detected using the LIVE/DEAD Fixable Violet Dead Cell Stain Kit (Invitrogen). The percentage of specific lysis was calculated using the following formula: % specific lysis = ((% sample lysis − % control lysis)/(100 − % control lysis)) × 100, where ‘% sample lysis’ corresponds to T cells + cancer cells and ‘% control lysis’ corresponds to cancer cells alone.

Mice

All studies used C57Bl/6 mice (CD45.2 wild type; CD45.1 wild type; Batf3−/−; Rag1−/−). C57Bl/6 (CD45.2) mice were purchased from Charles River Laboratories (France). C57Bl/6 (CD45.1), Batf3−/− and Rag1−/− mice were maintained as stable colonies in the EPFL mouse facility. All mice were housed and bred under specific pathogen-free conditions in the EPFL, UNIGE and DKFZ mouse facilities. Mice were housed in groups of up to five mice per cage at 18–24 °C with 40–60% humidity and maintained on a 12:12 h light:dark cycle (06:00–18:00 h). Experiments involving subcutaneous tumor models used cohorts of 6–9-week-old female mice. Glioma studies were performed twice: once in female mice and once in male mice (all 6–8 weeks old). The liver cancer studies used 8–9-week-old male mice. BM cells were isolated from female mice.

Adoptive DC transfer studies

For adoptive transfer of DCs to tumor-free mice, DCPs, moDCs and cDC1-like cells were prepared from CD45.1 mice (or CD45.2 mice in the case of Batf3−/− donor mice). Congenic CD45.2 mice (or CD45.1 mice in the case of Batf3−/− donor cells) received 2 × 106 cells twice (3 days apart) through the tail vein, without prior conditioning. Recipient mice were killed 4 days after the second cell dose, and the spleen was removed to analyze donor-derived cells.

For adoptive transfer of DCPs, moDCs and cDC1-like cells to tumor-bearing mice, CD45.1 or CD45.2 cells, generated and transduced as described above, were infused twice (2–3 days apart) in recipient mice (either CD45.2 or CD45.1). The DC dose and time of injection after tumor initiation was dependent on the tumor model, as shown in the figures. When LVs were used that expressed only one cytokine, transduced cells were mixed before injection through the tail vein in a 1:2 ratio for IL-12 and FLT3L-expressing cells, respectively (1 × 106 IL-12-expressing cells plus 2 × 106 FLT3L-expressing cells in most experiments, excluding intracranial DC delivery; see ‘intracranial glioma model’ model below). Control cells expressing marker genes only (GFP or dLNGFR) were mixed to corresponding ratios and numbers, and infused. When LVs were used that expressed both cytokines coordinately, 1 × 106 cells (most experiments) or 2 × 106 cells (for moDC in dose escalation study) were infused (excluding intracranial DC delivery; see ‘intracranial glioma model’ model below). Recipient mice were killed several days or weeks after the second DCP injection, depending on the tumor model.

Subcutaneous tumor models

B16F10 and MC38 cells were passaged at least three times to obtain actively growing cells. Cancer cells were then resuspended in PBS (5 × 106 cells per ml for MC38 and 2 × 106 cells per ml for B16F10 cancer cells), and 100 µl of cell suspension was injected subcutaneously into the right flank of C57BL/6 mice. DCPs, moDCs and cDC1-like cells were infused on days 3 and 5 after tumor injection, except in Fig. 5a,b (days 11 and 13). The mice and tumor growth were monitored three times per week. The tumors were allowed to grow up to 1 cm3 in size. Upon reaching the endpoint tumor size, the experiments could continue for an additional 48 h, provided that all health parameters detailed in a health score sheet remained normal. Long (D) and short (d) tumor diameters were measured with a caliper and the tumor volume calculated using the following formula: tumor volume = ½ × d2 × D.

Liver tumor models

To obtain KrasG12D; Trp53−/− liver tumors, we used HDTVi of previously described oncogenic plasmids75. To obtain mice with KrasG12D; Trp53−/− liver tumors, we prepared 2 ml of 0.9% NaCl solution containing the following plasmids (for one mouse): 5 µg of pT3-EF1a-KrasG12D-IRES-GFP (gift from D. Tschaharganeh, DKFZ, Heidelberg), 10 µg of px330-sg-tp53 (gift from T. Jacks, MIT, Boston; addgene, plasmid no. 59910) and 1.25 µg of pCMV-sleeping beauty 13 (pCMV-SB13) transposase-encoding plasmid76 (gift from D. Tschaharganeh, DKFZ, Heidelberg). To obtain mice with Myc; Trp53−/− liver tumors, we prepared 2 ml of 0.9% NaCl solution containing the following plasmids (for one mouse): 5 µg of pT3-EF1a-MYC-IRES-luciferase34 (addgene, plasmid no. 129775), 10 µg of px330-sg-tp53 and 1.25 µg of pCMV-SB13 transposase-encoding plasmid. The plasmid mix was delivered by HDTVi with its volume adjusted to 10% of the body weight of each mouse.

Both HDTVi models were initiated in male C57BL/6 mice at the age of 8–9 weeks. Seven days after HDTVi, the liver was imaged by magnetic resonance tomography (MRT) for the KrasG12D; Trp53−/− model, and by in vivo imaging system (IVIS) analysis of luciferase for the Myc; Trp53−/− model. For MRT (KrasG12D; Trp53−/− model), anesthetized mice were screened using a Pharmascan 7T MRT (Bruker) with ParaVision 5.1 software in FLASH scan mode without fat suppression, using an echo-time of 2.2 ms for out-phase and 2.9 ms for in-phase. For IVIS, imaging was performed using an IVIS Spectrum system (PerkinElmer). Mice were injected intraperitoneally with fresh d-luciferin (150 mg kg–1; ThermoFisher Scientific) and anesthetized with 3% isoflurane. Ten minutes after luciferin injection, mice were scanned in the IVIS imaging chamber. The luciferase signal was analyzed with IVIS software (v4.7.3; PerkinElmer). Randomization was performed based on MRT or IVIS imaging results.

Transduced DCPs were infused in two doses on days 11 and 13 after HDTVi. DCPs were mixed before injection (1 × 106 IL-12-expressing cells and 2 × 106 FLT3L-expressing cells). For each model, one mouse cohort was killed on day 23 (KrasG12D; Trp53−/−) or day 21 (Myc; Trp53−/−) for flow cytometry. Another cohort was monitored for survival for up to 90 days. The number of liver nodules was determined post-mortem by inspection under a stereoscope. Termination criteria were defined by authority-confined endpoints (morbidity; non-physiological posture; indication of jaundice, cramps, or emaciation; weight loss of more than 20% of the initial weight).

Intracranial glioma model

For the intracranial glioma model, 6–8-week-old C57BL/6 mice were implanted intracranially with SB28 cells using a stereotaxic apparatus. Mice were anesthetized with isofluorane and received subcutaneous injections of 2 μg of buprenorphinum (Bupaq; Streuli) in 100 μl of PBS and 50 μl of lidocaine 1% (Streuli) before surgery. Next, 1.6 × 103 SB28 cells were injected in 2 μl PBS in the pallidum. Seven days after tumor engraftment, we measured the tumor burden by IVIS (PerkinElmer) and randomized the mice according to tumor burden. Eight and eleven days after tumor engraftment, the mice received transduced DCPs intravenously (either 1 × 106 IL-12-expressing cells plus 2 × 106 FLT3L-expressing cells, or 1 × 106 DCPs co-expressing IL-12 and FLT3L). Additionally, 10 days after tumor engraftment, mice received 0.5 × 106 transduced DCPs and 1 × 106 CAR-T cells intracranially in the same location as the injected tumor. Mice were monitored three times per week, imaged by IVIS two times per week and killed when meeting authority-confined endpoints (15% weight loss over 1 week; compromised ability to walk, eat or drink; dyspnea; hunched posture; or lethargic behavior).

Treatment of mice with cell-depleting or neutralizing antibodies

Cell-depleting or neutralizing antibodies were administered starting 1 day before DCP transfer, followed by injection twice per week until the end of the experiment. We used the following monoclonal antibodies: CD8a+ cell-depleting antibody (12 mg kg–1; clone 53–6.7, rat IgG2a, Bio X Cell), CD4+ cell-depleting antibody (10 mg kg–1; clone GK1.5, rat IgG2b, Bio X Cell), CSF1R+ cell-depleting antibody (30 mg kg–1; clone AFS98, rat IgG2a, Bio X Cell), NK1.1+ cell-depleting antibody (20 mg kg–1; clone PK136, rat IgG2a, InVivoPlus) and IFNγ-neutralizing antibody (12 mg kg–1; clone XMG1.2, rat IgG1, Bio X Cell). All antibodies were prepared in sterile PBS (100 μl) and injected intraperitoneally.

Treatment of mice with cisplatin and PD-1 blocking antibodies

For the treatment of MC38 tumor-bearing mice, cisplatin (5 mg kg–1) was administered once on day eight after tumor injection. The mice then received a PD-1 blocking antibody (10 mg kg–1; rat IgG2a, clone RMPI-14, BE0146, Bio X Cell) or rat IgG2a isotype control antibody (10 mg kg–1; clone 2A3, BE0089, Bio X Cell) twice per week starting 1 day after DCP transfer. For the treatment of liver tumor-bearing mice, cisplatin (5 mg kg–1) was injected intraperitoneally in mice on day nine after HDTVi. These mice were also treated with a PD-1 blocking antibody or control IgG2a (3 mg kg–1) three times per week starting 1 day after DCP transfer. Cisplatin and antibodies were prepared in sterile PBS (100 μl) and injected intraperitoneally.

Flow cytometry analysis of mouse cells

Tissue samples were processed as described in the Reporting Summary. Single-cell suspensions were incubated in PBS with Fc block (1:100; BD Biosciences) and fixable live–dead colors, and stained with antibodies (see Reporting Summary and Supplementary Table 1). In experiments with non-fixable live–dead colors, cells were resuspended in live–dead color 7-AAD (1 µg ml–1; BioLegend) or DAPI (0.1 µg ml–1; Sigma-Aldrich) before acquisition by flow cytometer. Stained samples were analyzed with the following flow cytometry machines: BD LSRFortessa (Becton Dickinson); BD LSR II SORP (Becton Dickinson); and Attune NxT (Invitrogen). Analysis of flow cytometry data used FlowJO v10.1. Immune cell populations were identified as indicated in the Reporting Summary and Supplementary Figs. 1–6.

Immunofluorescence staining of tumor and liver

Freshly isolated tumors and livers were snap-frozen in OCT Compound and stored at −80 °C. Tissue sections were fixed in methanol for 20 min at −20 °C, washed and incubated in PBS with Fc block, 1% BSA and 5% FBS for 1 h at room temperature (20–25 °C). Sections were incubated overnight at 4 °C in blocking solution containing primary antibodies, followed by staining with secondary antibodies in some cases (see Reporting Summary and Supplementary Table 1). After staining, nuclei were labeled with DAPI. Image acquisition used an Axioscan microscope (Zeiss) as detailed in the Reporting Summary.

Immunofluorescence staining of brain

Upon euthanasia, mice were perfused by cardiac injection of PBS. Whole brains were removed, placed in OCT and snap-frozen. Sections of 10 μm were fixed in 80% methanol. Following incubation in Tris-NaCl-blocking buffer, slides were incubated overnight with the antibodies listed in the Reporting Summary and Supplementary Table 1. Slides were mounted with Fluoromount and DAPI (Invitrogen). Images were acquired with an Axioscan microscope (Zeiss).

Generation and transduction of human DCPs

CD34+ cells purified from cord blood were purchased from STEMCELL Technologies. CD34+ cells purified from plerixafor and G-CSF-MPB were purchased from STEMCELL Technologies and Lonza. CD34+ cells were cultured in StemSpan SFEMII medium (STEMCELL Technologies) with StemSpan CD34+ Expansion Supplement (STEMCELL Technologies) and 1 μM UM729 (STEMCELL Technologies) for 7 days at a concentration of 5 × 104 cells per ml in U-bottom 96-well plates (medium was replaced every 2–3 days). Day-seven cultures contained CD34+CD115+ cells (comprising DCPs) and additional progenitor-cell populations identified as indicated in the Reporting Summary. To transduce human DCPs, day-one CD34+ cells were transferred to retronectin (Takara) coated wells and dmPGE2 (STEMCELL Technologies) was added to a final concentration of 10 μM. After 2 h, the cells were transduced with a dLNGFR-encoding LV7 at the multiplicity of infection of 300. The medium was replaced on the following day, and the cells were cultured for five additional days.

Purification of human DCPs and differentiation into DCP-derived DCs and APCs

Day-seven cultures were processed to FACS-sort CD34+CD115+ DCPs (or mock-sorted) using a BD FACSAria II SORP apparatus. Sorted DCPs were cultured in StemSpam SFEMII medium (STEMCELL Technologies) supplemented with 50 units per ml of penicillin (Gibco), 50 μg ml–1 streptomycin (Gibco), 20 ng ml–1 hGM-CSF, 100 ng ml–1 hFLT3L, 20 ng ml–1 hSCF (all from PeproTech) and 1,000 IU ml–1 hIFNa2b (InvivoGen) for 7 days. Day-14 cultures contained differentiated cell populations identified as indicated in the Reporting Summary.

Generation of human moDCs

Blood from healthy donors was obtained from the Blood Transfusion Center (University of Lausanne, Switzerland) under Project P_297. Peripheral blood mononuclear cells were isolated by density gradient centrifugation on Lymphoprep (STEMCELL Technologies). To generate moDCs, CD14+ cells were isolated with magnetic beads (Miltenyi) and cultured in RPMI 1640 containing 10% FBS, 100 U ml–1 penicillin, 100 μg ml–1 streptomycin, 2 mM glutamine, 50 ng ml–1 hGM-CSF and 50 ng ml–1 hIL4 (PeproTech; 200-04), at a density of 106 cells per ml for 7 days.

Human T cell stimulation assays

The A2/CMV/pp65495-504-specific CD8+ T cell line42 was provided by D. Speiser (University of Lausanne, Switzerland) and kept in culture in 8% human serum (from AB+ donors; Blood Transfusion Center, Lausanne), 100 U ml–1 penicillin, 100 μg ml–1 streptomycin, 2 mM glutamine, 1% NEAA (Gibco), 1 mM sodium pyruvate (Gibco) and 55 μM 2-mercaptoethanol (Gibco), in RPMI 1640 supplemented with 150 U ml–1 hIL-2 (PeproTech).

To evaluate antigen presentation, HLA-A2+ moDCs or DCP-derived DCs and APCs (DCP-progeny) were pulsed with 1 μg ml–1 CMV/pp65495-504 peptide (provided by the Peptide & Tetramer Core Facility; Ludwig Institute for Cancer Research, Lausanne) for 1 h at 37 °C. After washing, cells were co-cultured with A2/CMV/pp65495-504-specific CD8+ T cells at a 1:1 ratio in RPMI 1640 containing 10% FBS, 100 U ml–1 penicillin and 100 μg ml–1 streptomycin. After 30 min at 37 °C, brefeldin A (1:1000; BD Biosciences, GolgiPlug) was added and cells were further cultured overnight before flow cytometry analysis. To assess cross-presentation, HLA-A2+ moDCs or DCP-progeny were pulsed with 10 μg ml–1 CMVpp65 protein (Abcam) for 2 h at 37 °C before adding A2/CMV/pp65495-504-specific CD8+ T cells at a 1:1 ratio. Co-cultures were kept at 37 °C overnight and for another 4 h in the presence of brefeldin A before flow cytometry analysis. To assess cross-dressing, extracellular vesicles isolated from human melanoma cell lines43 were directly pulsed with 1 μg ml–1 CMV/pp65495-504 peptide for 1 h at 37 °C and washed with PBS before adding them to HLA-A2− moDCs or DCP-progeny at 1 μg ml–1 concentration in RPMI 1640 containing 5% extracellular-vesicle-depleted FBS, 100 U ml–1 penicillin and 100 μg ml–1 streptomycin and glutamine. Extracellular vesicles were pulsed overnight at 37 °C and, after two washes, A2/CMV/pp65495-504-specific CD8+ T cells were added to the cells at a 1:1 ratio. After 30 min at 37 °C, brefeldin A was added and cells were cultured for another 5 h before flow cytometry analysis. As negative controls, we used non-pulsed moDCs or DCP-derived DCs and APCs, or T cells alone. As a positive control, T cells were stimulated with 10 ng ml–1 PMA and 500 ng ml–1 ionomycin.

scRNA-seq

After filtration through a 40 μm Flowmi strainer (Bel-Art), single-cell tumors were resuspended in PBS with 0.04% BSA, checked for the absence of doublets or aggregates and loaded into a Chromium Single Cell Controller (10× Genomics, Pleasanton) in a chip together with beads, master mix reagents (containing reverse transcriptase and poly-dT primers) and oil to generate single-cell-containing droplets. Single-cell expression libraries were prepared using Chromium Single-Cell 3′ Library & Gel Bead Kit v3.1 (PN-1000268; protocol CG000315, Rev C). Quality control was performed with a TapeStation 4200 (Agilent) and QuBit dsDNA high-sensitivity assay (ThermoFisher). Sequencing libraries were loaded onto an Illumina HiSeq 4000 paired-end flow cell and sequenced using read lengths of 28 nt for read1 and 91 nt for read2, at a depth of about 35 k reads per cell. Cell Ranger Single-Cell Software Suite v6.1.1 was used to perform sample demultiplexing, barcode processing and 3′ gene counting using 10× Genomics custom annotation of murine genome assembly mm10. After mapping with Cell Ranger 7.0 (X, with parameters force = 15,000) on the mouse reference refdata-gex-mm10-2020-A, cells were considered for further processing using Seurat (v4.1.1). Using Seurat (v4.1.1), cells with more than 20% mitochondrial content were removed; only cells in which at least 200 genes were expressed were retained; and only genes detected in at least ten cells were retained, for a final set of 1,524 cells and 17,901 genes. Samples were independently log-normalized and integrated using 4,000 most variable features. Unsupervised clustering was performed by applying the graph-based clustering approach and Louvain algorithm, and uniform manifold approximation and projection dimensional reduction was performed based on the previously computed neighbor graph using the top 30 principal components. Manual annotation based on selected markers was used to annotate cell clusters by unsupervised clustering. Wilcoxon test statistics were used to examine differences between DCP-treated and control samples at the cell level. Overrepresentation was performed on statistically significant genes using the Hallmark and Reactome gene sets collection with ClusterProfiler (v4.4.1), using a hypergeometric test followed by Benjamini–Hochberg P value correction.

Bulk TCR sequencing

Sample preparation and TCR-seq were performed as described77, with some modifications as follows. TCR products were purified, quantified and loaded on a NextSeq instrument (Illumina) for the deep sequencing of the TCRβ chain. The TCR sequences were further processed using ad hoc Perl scripts to: (1) pool all TCR sequences coding for the same protein sequence; (2) correct for amplification and sequencing errors using 9mers UMI; (3) filter out all out-frame sequences; and (4) determine the abundance of each distinct TCR sequence. TCRs with a single read were removed for the analysis. TCR clones that were considered out of frame were filtered out from the datasets. Analysis of the TCR clones was performed by using the Immunarch library in R v4.2.1. TCR diversity was estimated by computing the TCR richness of each tumor, defined as the number of unique clonotypes in each dataset. Tumors of mice treated with DCPs were pooled together, and TCR richness was compared with tumors of mice that did not receive DCPs. The proportion of each V gene was compared between the two groups. Tumors were correlated based on V gene usage using the Jansen–Shannon divergence and clustered by multidimensional scaling and k-means clustering.

Statistics and reproducibility

Studies involving independent cohorts of mice were typically performed once, with several exceptions stated in the figure legends. No specific statistical tests were used to predetermine the sample size, and our previous experience with different tumor models provided guidance. Based on these considerations, we typically employed experimental cohorts of five to ten mice. Studies conducted in parallel may share selected mouse cohorts to limit the number of experimental mice. Consequently, some datasets may be shown more than one time in separate figures to facilitate presentation of the data (for example, Fig. 3g and Extended Data Fig. 7m; Fig. 4j and Extended Data Fig. 5b). Studies involving human primary blood-derived cells used several independent donors in experiments aimed to characterize key properties of the cells (for example, DCP expansion and yield). In those studies, we used five to seven different donors to verify reproducibility of the key results across independent donors. Further studies that aimed to evaluate qualitative differences (for example, behavior of moDCs and DCPs) used only one to three independent donors, and statistical analyses were not performed.

Tumor-bearing mice were randomized before treatment by allocating mice to alternate treatment groups. Endpoints for experiments with mice were selected in accordance with institutional-approved criteria; fixed time points of analysis shown in the figures indicate time elapsed from tumor injection. The investigators were blinded when acquiring tumor volumetry data (both at randomization and endpoint of analysis), flow cytometry data and immunofluorescence staining data, but were not blinded when analyzing flow cytometry data. In some cases, selected samples were excluded from specific analyses because of technical flaws during sample processing or data acquisition; this was the case, for example, when analyzing tissue samples that did not provide sufficient numbers of cells for multi-panel flow cytometry. Outliers were not excluded from the analyses.

Graphs were generated and statistical analyses performed with Prism (GraphPad Software). Error bars indicate s.e.m. unless indicated otherwise. The number of biological (non-technical) replicates and applied statistical analysis are indicated in the figure legends. In brief, comparison between two unpaired groups was performed by the non-parametric Mann–Whitney test. For multiple comparisons involving one variable, one-way ANOVA was performed followed by Tukey’s multiple comparison test, unless otherwise stated in the figure legends. Simultaneous analysis of two variables (tumor growth over time) among multiple groups was performed by two-way ANOVA followed by Tukey’s (three groups or more) or Sidak’s (two groups) multiple comparison test. Other statistical tests were applied in selected cases, as detailed in the figure legends.

Reporting summary

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