Efficacy of CD40 Agonists Is Mediated by Distinct cDC Subsets and Subverted by Suppressive Macrophages Open Access

Effective treatment of many cancer types consistently improved over recent decades (1). Despite checkpoint inhibitors cementing themselves as invaluable therapeutic interventions, only a minority of patients experience long-term efficacy (2). Therefore, identification of prognostic biomarkers and synergistic combination therapies that can increase the proportion of responsive patients are current focuses at the forefront of tumor immunology research (3).

Alternative therapies that aim to prime T cells rather than rescue dysfunctional T cells show great promise (4). The TNF-receptor superfamily member CD40 is an ideal target within this context, as CD40 ligation that occurs naturally during T-cell help via CD40-L results in the activation of antigen-presenting cells leading to increased T-cell priming (5–8). Preclinical results using CD40 agonist antibodies have been shown to slow the growth of murine tumors containing strong tumor antigens (9, 10), however their success in the clinic as a monotherapy was limited to a minority of patients with melanoma (11). An encouraging aspect of CD40 agonist therapy lies in the broad potential for synergistic combinations that have been shown to reduce tumor growth, including antiangiogenic therapies, tumor-associated macrophage depletion, checkpoint inhibitors, chemotherapy, and radiotherapy (12–19). As such, CD40 agonists have achieved beneficial clinical outcomes in pancreatic cancer when combined with chemotherapy (20). Although the results of these combinations are encouraging, they also hint at the importance of understanding which combination of therapies should be applied in which context.

Most antitumor effects of CD40 agonists have been shown to rely on the function of CD8⁺ T cells. However, critical cellular mediators must have activated these CD8⁺ T-cell responses. The prime candidate that has been identified as critical to CD40 efficacy are type 1 conventional dendritic cells (cDC1) that are essential for CD8⁺ T-cell priming (18, 21, 22). However, studies have also implicated macrophages and other monocyte-derived cells as critical components of successful CD40 agonist-mediated antitumor immunity (23, 24). Encouraging combinations investigated so far involve Flt3L treatment-mediated DC boosting therapies prior to CD40 agonist therapy, with or without radiotherapy, which have been able to slow tumor growth of orthotopic and subcutaneously implanted pancreatic ductal adenocarcinoma tumors, respectively (25, 26).

Altogether, these results underline the importance of understanding both the cancer- and immune-specific contexture, relating to successful CD40 agonist therapy. To shed further light on how the tumor microenvironments predict optimal responses to CD40 agonist therapy, and which combinatory interventions can resensitize nonresponsive tumors, in this study we performed single-cell RNA (scRNA-seq) sequencing on tumor-infiltrating immune cells to identify the cellular mediators of anti-CD40 (αCD40) therapy. We also utilized the Xcr1^wt/dtr mouse model to temporally deplete cDC1s and show that whereas the therapeutic effect of αCD40 therapy in B16F10 tumors relied on the initial function of cDC1s prior to therapy, cDC2s could be responsible for the subsequent activation, but not expansion, of antitumor T cells in response to αCD40 therapy. When comparing the αCD40-responsive B16F10 melanoma with the αCD40-resistant Lewis lung carcinoma (LLC), we identified that the highly immunosuppressive microenvironment of LLC tumors as well as their poor immunogenicity limited αCD40 efficacy. By reducing suppression through αCSF1R treatment and increasing immunogenicity by combination with immunogenic cell death (ICD)-inducing chemotherapy, we could resensitize subcutaneous and orthotopic LLC tumors to αCD40 therapy.

Female C57BL/6 mice were purchased from Janvier. Xcr1^wt/dtr mice were provided by Christian Kurts (University of Bonn) with the permission of Tsuneyasu Kaisho. Csf3r^−/− mice were provided by Sebastian Jaillon and Paola Allavena (Humanitas University). CD45.1 mice were purchased from Charles River. Itgax-DTR mice were obtained from in-house breeding. In all experiments involving transgenic or knockout mice, wild-type (+/+) littermate mice were used as controls as specified in the figures and figure legends.

All procedures followed the guidelines of the Belgian Council for Laboratory Animal Science and were approved by the Ethical Committee for Animal Experiments of the Vrije Universiteit Brussel (licenses 16–220–02, 18–220–19, 19–220–33, 20–220–32, 21–220–25, 22–220–12) and the Animal Ethics Committee of KU Leuven (ECD project P073/2022).

For the generation of bone marrow chimeras, female 6-week-old CD45.1 mice were lethally irradiated (8 Gy). After a six-hour rest period the mice were injected intravenously with 1.3 × 10⁶ BM cells obtained from Itgax-WT or Itgax-DTR littermate mice. The mice were used experimentally 8 weeks after BM reconstitution. Chimerism was confirmed by flow cytometry prior to tumor challenge and treatment.

LLC and B16F10 cell lines (from ATCC) were cultured in DMEM (Gibco) supplemented with 10% (v/v) heat-inactivated FCS (Capricorn Scientific), 300 μg/mL l-glutamine, 100 units/mL penicillin, and 100 μg/mL streptomycin. For the LLC-OVA (a kind gift from Dmitry Gabrilovich) cell line, DMEM was replaced by RPMI (Gibco).

For subcutaneous tumor implantation, 10⁶ LLC cells, 10⁶ B16F10, or 3×10⁶ LLC-OVA cells were injected subcutaneously into the right flank of syngeneic female C57BL/6 mice in 200 μL of Hank's Balanced Salt Solution (HBSS). For intravenous tumor injection, 0.25 × 10⁶ LLC-OVA or LLC cells were injected into the tail vain of syngeneic female C57BL/6 mice in a 100 μL volume of HBSS.

Tumor volumes were determined by caliper measurements and calculated using the formula: V = π × (d² × D)/6, where d is the shortest diameter and D is the longest diameter.

For CD40 agonist treatments, a single dose of 100 μg of αCD40 (clone: FGK4.5; BioXCell) agonist antibody or rat IgG2a isotype control (clone 2A3; BioXCell) was administered intraperitoneally in a volume of 100 μL HBSS when tumors reached approximately 100 mm³.

For macrophage depletions, 660 μg of αCSF1R (clone 2G2; provided by Roche) or murine IgG1 isotype control (clone MOPC-21; BioXCell) were administered intraperitoneally in a volume of 100 μL HBSS when tumors reached approximately 100 mm³ with additional doses being administered weekly, if applicable.

For B-cell depletions, 500 μg of αCD20 (clone 18B12; BioXCell) or murine IgG2a (clone 2A3; BioXCell) were administered intraperitoneally in a volume of 100 μL HBSS once at day 4 after tumor implantation.

To deplete cDC1s in Xcr1^wt/dtr mice, diphtheria toxin (D0564, Merck) was injected intraperitoneally in Xcr1^wt/wt and Xcr1^wt/dtr mice at a dose of 25 ng/g body weight for the first dose, with following doses administered at a dose of 5 ng/g body weight.

For CD8⁺ T-cell depletions, 200 μg of αCD8 (clone YTS169; Polpharma Biologics) was administered intraperitoneally in a volume of 100 μL HBSS every 2 to 3 days starting 1 day prior to tumor implantation.

For IL12 neutralization, 500 μg of αIL12 p40 (clone C17.8, BioXCell) or rat IgG2a (clone 2A3: BioXCell) was administered intraperitoneally in a volume of 100 μL HBSS daily starting 24 hours prior to αCD40 treatment and continuing until the end of the experiment.

To increase cDC numbers, 30 μg of Flt3L-Ig (hum/hum; clone Flt3L Fc-G1; BioXCell) was administered intraperitoneally in a volume of 50 μL HBSS every 24 hours between day 0 and day 8 post-LLC tumor implantation.

To deplete Tregs, 100 μg of αCD25 (ONCC4, kindly provided by Oncurious) was administered intraperitoneally in a volume of 100 μL HBSS every 48 hours between day 4 and day 10 after LLC tumor implantation, unless otherwise indicated.

Neutrophils were depleted using 75 μg αLy6G (clone 1A8; BioXCell) followed by 150 μg mouse anti-RAT (clone MAR18.5; BioXCell) administered intraperitoneally in volumes of 100 μL HBSS. Alternatively, neutrophils were depleted using a CXCR2 inhibitor (SB225002; Selleck Chemicals) administered intraperitoneally at a dosage of 4 mg/kg body weight.

Oxaliplatin (NSC 266046; Selleck Chemicals) was dissolved in HBSS containing 5% glucose and administered intraperitoneally at 1 mg/kg body weight every 48 hours between day 4 and day 14 after LLC tumor implantation. Vehicle control (5% glucose in HBSS) was administered according to body weight of mice at time of treatment. Volumes administered were equal to 2 μL × weight of mouse (g).

Blood was collected from mice in 1 mL syringes containing 0.5 mol/L EDTA. Tumors were excised, cut in small pieces, incubated with 10 U/mL collagenase I, 400 U/mL collagenase intravenously and 30 U/mL DNase I (Worthington) in RPMI for 20 minutes at 37°C, squashed, triturated, and filtered on a 70 μm cell strainer. Spleens were mashed through a 70 μm cell strainer, bone marrow was flushed out from the femurs into RPMI. Single-cell suspensions were then treated with ammonium–chloride–potassium (ACK) erythrocyte lysis buffer.

Single cell suspensions were resuspended in HBSS and samples for flow cytometry analysis were incubated with Fixable Viability Dye eFluor 506 (1:1,000, eBioscience) for 30 minutes at 4°C. Next, cell suspensions were washed with HBSS and resuspended in HBSS with 2 mmol/L EDTA and 1% (v/v) FCS. To prevent nonspecific antibody binding to Fcγ receptors, cells were preincubated with anti-CD16/CD32 (clone 2.4G2) antibody. Cell suspensions were then incubated with fluorescently labeled antibodies diluted in HBSS with 2 mmol/L EDTA and 1% (v/v) FCS for 20 minutes at 4°C and then washed with the same buffer. The following fluorochrome-conjugated antibody clones were used: CD45 (30-F11), CD11b (M1/70), Ly6G (1A8), SiglecF (E50–2440), MHC-II (M5/114.15.2), Ly6C (HK1.4), F4/80 (CI:A3–1), CD11c (HL3), XCR1 (ZET), NK1.1 (PK136), CD19 (1D3), TCRβ (H57–597), CD4 (RM4–5), CD8α (53–6.7), CD44 (IM7), CD62 L (MEL-14), PD-1 (RMP1–30), CCR8 (REA921), MMR (C068C2), LAG-3 (C9B7W), CXCR2 (5E8/CXCR2), CD200 (OX2), CCR7 (4B12), MHC-I (SF1–1.1), SiglecH (551), Dextramer (Immudex, Catalog No. JD2163).

For intracellular staining, after extracellular staining was complete, samples were spun and fixed using the eBioscience Intracellular Fixation & Permeabilization Buffer Set (Thermo Fisher Scientific, 88–8824–00) according to manufacturer's instructions. The following fluorochrome-conjugated antibody clones were used: FoxP3 (FJK-16s), GZMB (GB11), Ki67 (16A8), ARG1 (14D2C43), and IL12p40 (C17.8). To measure active caspase-3 we used the FITC Active Caspase-3 Apoptosis Kit (BD Biosciences, 550480).

Flow cytometry data were acquired using a BD FACSCanto II (BD Biosciences) and analyzed using FlowJo. The gating strategy to identify immune cell populations in tumors is shown in Supplementary Fig. S1. Samples with cell contamination from the tumor-draining lymph node (identified as outliers in B-cell and naive T-cell abundance) were excluded from further analyses.

For fluorescence-activated cell sorting, 7AAD− CD45⁺ immune populations were sorted into RPMI containing microcentrifuge tubes for single cell sequencing. For purification of tumor-residing neutrophils, samples were enriched for CD11b⁺ cells using magnetic cell separation (Miltenyi). 7-AAD staining was used to exclude dead cells. Cell subsets were then sorted into ME medium [RPMI with 10% (v/v) FCS, 300 μg/mL l-glutamine, 100 units/mL penicillin, 100 μg/mL streptomycin, 1% (v/v) MEM nonessential amino acids (11140050, Gibco), 1 mmol/L sodium pyruvate (Gibco), and 0.02 mmol/L 2-mercaptoethanol (Sigma-Aldrich)]. Fluorescence-activated cell sorting was performed using a BD FACSAria II (BD Biosciences).

Similarly sized tumors (collected at either day 15 or day 17 after tumor inoculation for LLC or B16F10, respectively) were pooled from three mice. The regular tissue processing procedure was followed, with the addiction of actinomycin D (Sigma-Aldrich, A1410–5MG) to each buffer. Tumor collection was performed in 30 μmol/L, enzyme incubation and subsequent filtering in 15 μmol/L, and all other steps in 3 μmol/L. For scRNA-seq, the single cell suspensions were stained with APC-Cy7-labeled anti-CD45 and 7AAD. Approximately 60,000 live CD45⁺ cells were sorted into ME medium using the BD FACSAria III (BD Biosciences). The sorted cells were centrifuged and resuspended in PBS containing 0.04% BSA at room temperature at an estimated final concentration of 1,000 cells/μL. The cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq) sample was counted and 1 million cells were isolated and centrifuged. The pellet was resuspended and incubated for 30 minutes on ice with 25 μL of staining mix in PBS + 0.04% BSA containing APC-Cy7 labeled mouse anti-CD45 and the mouse cell surface protein antibody panel containing 174 oligo-conjugated antibodies. Subsequently the cells were washed and 60,000 live CD45⁺ cells were sorted into ME medium. Next, the 10× genomics single-cell bead-in emulsions and scRNA-seq and CITE-seq libraries were prepared as described previously (27). The mean reads per cell for the LLC and B16F10 scRNA-seq data were 17,476 and 31,109, with a sequencing saturation metric of 38% and 42.7%, respectively. The LLC CITE-seq data yielded 11,624 mean RNA reads per cell, 28.4% RNA sequencing saturation, and 2,042 mean ADT reads per cell. For filtering of the low-quality cell barcodes, associated with empty droplets, the “emptyDrops” function of the DropletUtils package (v.1.8.0) has been applied on the RNA expression data, using an FDR cutoff of 0.01. The gene expression matrices were further filtered using the Scater package (v.1.16.2). The detection of outlier cells for percentage of mitochondrial genes per cell and removal of low-abundance genes were performed as described previously (28). Library size normalization and unsupervised Leiden clustering were performed with Seurat v.3.2.3. The obtained clustering was visualized in two-dimensional scatter plots via Uniform Manifold Approximation and Projection (UMAP). Differential expression analysis was done using Wilcoxon Rank Sum test with the “FindMarkers” function of Seurat to identify genes, specific for each cluster. Bonferroni correction has been applied for adjustment of the P values. The processing of the ADT expression matrix was done as described previously (28).

Trajectory inference was performed on the monocyte and TAM subsets of the mouse B16F10 and LLC tumors, using the Slingshot package (v.1.8.0; ref. 29). The B16F10 and the LLC datasets were merged using the “merge” command of Seurat, then monocyte and TAM clusters were subsetted and clustered using the same procedure as described above. Slingshot was run on the first 10 PCA embeddings of the monocyte/TAM subset. To identify differentially expressed genes along the identified trajectories, the package tradeseq was used (v.1.4.0), using five knots for fitting the model. To find the genes that vary significantly between the two lineages, the “diffEndTest” was used, whereas for Identifying genes that change along a lineage, the “associationTest” function was applied (30).

To predict the putative molecular pathways and functions of the genes that distinguish the B16F10 and LLC TAMs, we performed a GO analysis on the genes that varied significantly between the two lineages using the Metascape (http://metascape.org/) online tool with default parameters (31). We have selected the genes had Wald statistic >100 and LogFc >1.5 or LogFc < −1.5, respectively for the “diffEndTest” between lineage 1 and 2.

Zhang and colleagues analyzed CD45⁺ sorted tumors and tumor-draining lymph nodes from MC38 tumor-bearing mice treated with αCD40 antibodies (32). We have extracted the raw FASTQ data of the day 2 treated MC38 tumors with αCD40 or isotype control (ERR3498977, ERR3499108, ERR3498975, ERR3499106, ERR3499107, ERR3499050, ERR3498978, ERR3499109, ERR3507081, ERR3507082) from https://www.ebi.ac.uk/ena/browser/view/PRJEB34105. The single cell data have been analyzed as described above. DC clusters have been subsetted and reclustered.

Mice were anaesthetized with isoflurane (2% in 100% oxygen; Piramal Healthcare) and scanned in the supine position using in vivo μCT (Skyscan 1278; Bruker micro-CT) with the following parameters: 50 kVp X-ray source voltage, 350 μA current, 1 mm aluminum X-ray filter, 150 milliseconds exposure time per projection, acquisition of three projections per step was performed in 0.9° increments over a total of 220° angle, 10 cm field of view covering the whole body producing expiratory-weighted reconstructed data sets with 50 μm isotropic voxel size (33). Each scan took approximately 3 minutes and was associated with a measured radiation dose of 60 to 80 mGy (34).

Software provided by the manufacturer (TSort, NRecon, DataViewer, and CTan) was used to retrospectively gate, reconstruct, visualize, and process the μCT data. Quantification of nonaerated lung volume and total lung volume was carried out for a volume-of-interest covering the lung including the regions that had been manually delineated from coronal μCT images ensuring heart and main blood vessels were avoided (35, 36).

A total of 2 × 10⁵ neutrophils sorted from tumors were added to 2 × 10⁵ naïve C57BL/6 splenocytes stimulated with anti-CD3 (1 μg/mL) and anti-CD28 (2 μg/mL) and cultured in flat-bottom 96-well plates in ME medium for ex vivo cell culture described above. After 24 hours of culture, 1 μCi (0.037 MBq) ³H-thymidine was added and after another 18 hours of culture, the plates were frozen and stored at −20°C, after which, T-cell proliferation was measured as count per minute in a liquid scintillation counter.

The J774 macrophage-like myeloid cells were cultured in a media containing 10% heat inactivated FBS. After two passages, J774 cells containing genetic reporter constructs for detecting transcriptional activity of the NF-κB and IFN-stimulated response element (ISRE)-binding IFN regulatory factor (IRF) were enriched via antibiotic-based selection (using 5 μg/mL blasticidin and 100 μg/mL of zeocin). The J774 NF-κB and ISRE/IRF reporter myeloid cells (Invivogen) were plated with a density of 3 × 10⁴ cells per well in a 96-well plate. Cancer cells were plated in 10 cm dishes, and were treated with cisplatin (100 μmol/L), paclitaxel (100 μmol/L), doxorubicin (50 μmol/L), mitoxantrone (0.5 μmol/L), oxaliplatin (400 μmol/L), or left untreated. After 24 hours, the cancer cells were collected and counted. They were centrifuged at 15,000 rpm for 5 minutes and resuspended in J774 reporter myeloid media according to the manufacturer. These were then added on top of the J774 reporter myeloid cells, in a 1:1 ratio (in 200 μL final volume), within the 96-well plates. Stimulation with LPS (1,000 ng/mL) was used as a positive control. To measure the NF-κB transcriptional activity (marked by extracellular secretion of reporter alkaline phosphatase enzyme), after 24 or 48 hours of cancer cell-J774 coculture, 100 μL of media was transferred to a standard transparent-bottom 96-well plate. Herein, 100 μL of Quanti-BLUE substrate (Invivogen) for the alkaline phosphatase was added to each well and incubated for 4 to 8 hours. The absorbance was measured at an optical density of 655 nm with the Biotek Synergy H1M plate reader. To measure the ISRE/IRF expression (marked by extracellular secretion of reporter luciferase enzyme), another 100 μL of media was derived from the above cancer cell-J774 coculture in a white opaque-bottom 96-well plate. Herein, 50 μL of Quanti-LUC substrate (Invivogen) for the luciferase was added and bioluminescence was directly measured with 100 milliseconds of signal integration, with the Biotek Synergy H1M plate reader. To account for interassay baseline variability a fold change to the J774 myeloid cells alone was taken from all data derived from these reporter assays.

All schematic figures were created using BioRender.com.

All graphs show mean ± SEM. Statistical significance (P value < 0.05) was determined in GraphPad Prism 9.1.2 software. For relevant pairwise comparisons, unpaired t tests were performed. For the comparison of multiple groups, one-way ANOVA was performed, followed by a posttest. Tumor growth curves were compared by mixed-effects two-way ANOVA with multiple comparisons tests, where appropriate statistical tests with Welch correction were performed. For statistically significant differences, the P value is indicated in graphs as the following: *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

The data associated with this study are available in the main text or the Supplementary Materials and Methods. scRNA-seq raw data are deposited at GEO (NCBI) under accession code GSE209763.

B16F10 melanoma is a frequently used preclinical mouse model in immuno-oncology that is highly infiltrated by immune cells (Fig. 1A), of which, 13.0 ± 2.4% represent CD8⁺ T cells (Fig. 1B; Supplementary Fig. S1A). To assess the activation status of the tumor-infiltrating CD8⁺ T cells, we performed scRNA-seq on CD45⁺ immune cells from B16F10 tumors grown subcutaneously in C57BL/6 mice. Unsupervised clustering yielded 19 distinct clusters, identified on the basis of their expression of canonical marker genes (Fig. 1C; Supplementary Fig. S1B). Interestingly, both defined CD8⁺ T-cell clusters in B16F10 tumors expressed high levels of genes associated with an exhausted or dysfunctional T-cell phenotype including Pdcd1 (PD-1), Lag3 (CD223), and Tox (Fig. 1D). Anti-PD-1 mAb therapy has previously been shown to reinvigorate exhausted CD8⁺ T cells, but likely due to low Tcf7 (TCF7) expression in the CD8⁺ T-cell population (Fig. 1D), did not result in delayed tumor growth in the B16F10 model (Supplementary Fig. S1C; refs. 37, 38), despite Cd274 (PD-L1) gene expression within multiple different clusters (Supplementary Fig. S1D).

To investigate whether tumor growth could be arrested by targeting earlier steps in the tumor-immunity cycle, B16F10 tumor-bearing mice were treated with an anti-CD40 agonist antibody (αCD40; Fig. 1E), as CD40 was shown to enable DC licensing and maturation, resulting in subsequent priming of cytotoxic T cells (39). CD40 agonist monotherapy significantly reduced tumor growth and weight (Fig. 1F and G) whereas the relative infiltration of immune cells into B16F10 tumors increased (Fig. 1H). Ten days after αCD40 treatment, within the myeloid compartment, tumor-associated macrophages (TAM) were strongly decreased after successful αCD40 treatment, which is in line with previous observations showing that the presence of mature TAMs correlates with tumor size (40), whereas frequencies of cDC1 and cDC2 reduced slightly (Fig. 1I–M). Importantly, within the lymphocytes, the abundance of cytotoxic CD8⁺ T cells was strongly increased compared with CD4⁺ T cells, NK cells, and B cells (Fig. 1N–Q; Supplementary Fig. S1E and S1F), likely due to a higher proliferation rate (Fig. 1R). Moreover, in mice treated with αCD40, the CD8⁺ T cells displayed an effector T-cell phenotype as indicated by the increased CD44⁺ CD62L^– effector versus CD44^– CD62L⁺ naive T-cell ratio (Fig. 1S). The elevated abundance of activated CD8⁺ T cells was accompanied by both a decreased infiltration of FoxP3⁺ regulatory T cells (Tregs), as well as a reduced expression of CCR8 on the Tregs (Fig. 1T–U), indicative for reduced suppressive phenotype of these cells (41). Collectively, these results indicate that CD40 agonist monotherapy is sufficient to repolarize the immune infiltrate in B16F10 tumors delaying tumor growth.

To investigate the mechanisms underlying the reduced tumor growth upon αCD40 agonist treatment, we first set out to determine the impact of the increased abundance of cytotoxic CD8⁺ T cells on the inhibition of tumor progression. Systemic depletion of CD8⁺ T cells restored B16F10 tumor growth in αCD40 agonist treated mice to WT levels (Fig. 2A). Next, we interrogated our scRNA-seq data to assess which cell types were expressing Cd40 and were potentially driving antitumor CD8⁺ T-cell responses in B16F10 tumors. Cd40 expression was mainly found in B cells, cDCs including cDC1s, cDC2s, CCR7⁺ DCs [also termed migratory-DCs (MigDC), mature DCs, DC3 (42), or mregDCs (43)], and mononuclear myeloid cells including monocytes and different subsets of TAMs (Fig. 2B; Supplementary Fig. S2A). In accordance with the gene expression pattern, CD40 protein assessed via flow cytometry was only detected at the surface of Ly6C^high monocytes, TAMs, cDC1s, cDC2s, and B cells within B16F10 tumors (Fig. 2C; Supplementary Fig. S2B).

Next, we investigated whether any of these populations are involved in the therapeutic effect of αCD40 treatment. B-cell depletion using αCD20 antibodies did not alter the reduced tumor growth upon αCD40 therapy and increased abundance of effector CD8⁺ T cells (Supplementary Fig. S2C; Fig. 2D–F), indicating that the reduction in tumor growth mediated by the CD40 agonist was B-cell independent. Macrophage depletion using an αCSF1R antibody (Supplementary Fig. S2D) also did not revert the tumor growth, nor the increase in T-cell abundance and activation status or the decrease of immunosuppressive Tregs upon αCD40 therapy (Supplementary Fig. S2D; Fig. 2G–K), implying that the antitumor effect of CD40 agonist therapy was independent of TAMs.

Having excluded the requirement of B cells and TAMs for the generation of a therapeutic response upon αCD40 treatment in B16F10, we next investigated the role of cDC1s. Hereto, we employed Xcr1^wt/dtr mice, which allowed temporal control of systemic cDC1 depletion upon injection of diphtheria toxin (DT; Supplementary Fig. S3A–C; ref. 44). Strikingly, when cDC1 depletion was initiated 24 hours before αCD40 administration, the therapeutic effect of the CD40 agonist therapy was unaltered (Fig. 3A), suggesting that cDC1s did not play a major role in the αCD40-mediated immune response in established B16F10 tumors. Interestingly, in Xcr1^wt/dtr mice, αCD40 treatment reduced the abundance of CD8⁺ T cells to levels comparable with isotype-treated littermate control mice (Fig. 3B), highlighting the important role cDC1s play in increasing intratumoral CD8⁺ T cells. Despite the inhibited expansion of CD8⁺ T cells in Xcr1^wt/dtr mice, the CD8⁺ T cells still showed an effector T-cell phenotype in Xcr1^wt/dtr mice treated with αCD40, with CD44:CD62 L ratio's and granzyme B secretion similar to the T cells in αCD40-treated littermate controls (Fig. 3C; Supplementary Fig. S3D). These suggest that cDC1s are essential for CD8⁺ T-cell recruitment or expansion, whereas other cell types can also contribute to the proper activation of existing CD8⁺ T cells into antitumor effector cells. Importantly, the depletion of these CD8⁺ T cells in Xcr1^wt/dtr mice treated with αCD40 agonist restored tumor growth in αCD40 treated mice to isotype-treated littermate control levels (Fig. 3D; Supplementary Fig. S3E). Overall, these findings suggest that therapeutic responses induced by αCD40 were driven by CD8⁺ effector T cells, independent of cDC1-mediated activation. Indeed, when cDC1s were depleted 24 hours prior to tumor inoculation and depletion was maintained throughout tumor progression, the efficacy of αCD40 agonist therapy was abrogated (Fig. 3E). Consequently, only a nonsignificant trend towards higher CD8⁺ T-cell levels was seen in αCD40-treated Xcr1^wt/dtr mice, which was incapable of restricting B16F10 tumor growth (Fig. 3F).

Next, we aimed to unravel which other antigen-presenting cells were involved in the activation of existing CD8⁺ T cells upon CD40 agonist therapy in the absence of cDC1s. Because TAMs expressed CD40 in B16F10 tumors and CD40-activated macrophages were shown to be involved in CD40-mediated tumor responses (23), we depleted TAMs in Xcr1^wt/dtr mice (Supplementary Fig. S3f). The tumor progression and activation of CD8⁺ T cells in this experiment did not differ from the results obtained in mice depleted of cDC1s 24 hours before αCD40 administration, in which TAMs were present, suggesting that TAMs were not responsible for CD8⁺ T-cell activation in the absence of cDC1s (Fig. 3G; Supplementary Figs. S3F–S3H).

The only remaining immune cell types expressing CD40 that could be involved in CD8⁺ T-cell activation were cDC2s (Fig. 2B and C). Importantly, it was previously shown that the transcriptional program of cDC1s and cDC2s converged upon differentiation into MigDCs in various scRNA-seq analyses (43, 45). Two genes that showed specific upregulation in the MigDC cluster in our B16F10 data were Ccr7 and Cd200 (Fig. 3H and I; Supplementary Fig. S3I). After αCD40 administration, a higher proportion of both cDC1 and cDC2 expressed either receptor (Fig. 3J–M) and the frequency MigDCs, gated based on CD200 expression, was increased, whereas both cDC1 and cDC2 were reduced in frequency (Supplementary Figs. S3J–S3M). This is in line with the results we obtained when reanalyzing a publicly available scRNA-seq dataset of murine MC38-tumor bearing WT mice generated by Zhang and colleagues (32). Our analysis showed that the cDC1s and cDC2s were adopting a Ccr7 expressing MigDC profile 48 hours after αCD40 treatment (Supplementary Figs. S3N–S3O). This might suggest that cDC2 activated by αCD40-agonist could adopt a MigDC transcriptional phenotype and mediate the activation of preexisting CD8⁺ T-cell clones. Moreover, MHC-I levels on cDC2s were also increased upon αCD40 treatment (Supplementary Fig. S3P), further suggesting that as was shown in human (46, 47), cDC2s might be able to stimulate CD8⁺ T cells.

DC-derived IL12 was previously shown to stimulate T-cell immunity (16). In B16F10 tumors, IL12 was mainly upregulated in the MigDC cluster, both at the transcript and protein level (Fig. 3N; Supplementary Fig. S3Q). To parse the role of cDC2/MigDC-derived IL12 in effective αCD40 therapy, we depleted cDC1s and TAMs and treated Xcr1^wt/dtr mice with αCD40/αCSF1R while neutralizing IL12. Blockade of IL12 rendered the mice nonresponsive to αCD40 therapy and decreased the abundance and activation of tumor-infiltrating CD8⁺ T cells compared with αCD40/αCSF1R-treated mice, demonstrating that IL12 was essential to the therapeutic efficacy of αCD40 in the absence of both cDC1s and TAMs (Fig. 3O–Q). Next, to investigate whether depleting all CD11c⁺ cells (including cDC1, cDC2, and TAMs) within B16F10 tumors would abrogate the response to αCD40, we generated Itgax-DTR and Itgax-WT bone marrow chimeras to allow for continued depletion of CD11c⁺ cells (Supplementary Figs. S3R–S3U). Interestingly, we found that upon depleting CD11c⁺ cells, no differences were observed between isotype control and αCD40/αCSF1R treated mice, whereas tumor growth and the increase of effector CD8⁺ T cells was still significantly reduced in WT reconstituted mice treated with αCD40/αCSF1R (Fig. 3R–U).

Overall, our data indicate that, although cDC1s play an important role in expanding CD8⁺ T cells during early phases of tumor progression, they are dispensable for the activation of existing antitumor CD8⁺ T-cell clones driving therapeutic αCD40 responses. On the other hand, our data suggest that cDC2s are capable of stimulating antitumor CD8⁺ T cells in an IL12-dependent manner to reduce tumor growth upon CD40 agonist treatment.

Although αCD40 strongly reduced B16F10 tumor growth, long lasting antitumor responses were absent and eventually all mice lost tumor control approximately 5 days after the αCD40 treatment (Fig. 4A). Administration of a second dose of αCD40 5 days after the first dose did not provide any therapeutic benefit compared with mice that only received one αCD40 dose (Fig. 4A). When comparing the tumor immune infiltrate of the response phase on day 16 after tumor inoculation (tumor volume <400 mm³) to the regrowth phase on day 21 after tumor inoculation (tumor volume >600 mm³) upon αCD40 treatment, the myeloid compartment was more prominent in the latter at the expense of the CD8⁺ T-cell infiltrate (Fig. 4B). Moreover, there was an enrichment of MMR⁺ TAMs during the delayed regrowth phage, with MMR being a marker associated with a more protumor TAM phenotype (Fig. 4C). These data suggest that CD40 agonist therapy provides a short-term switch that polarizes the TME into an immunopermissive environment, but eventually the cytotoxic response subsides, resulting in therapy resistance and tumor regrowth.

Given that MMR⁺ TAM have been shown to stimulate tumor relapse after therapy (48) and that in several preclinical tumor models αCD40/αCSF1R combination was able to reduce tumor growth synergistically (14, 15), we wondered whether TAMs would be contributing in the delayed regrowth after αCD40 treatment. Therefore, we treated mice with αCD40 + αCSF1R when tumors reached 100 mm³. Indeed, TAM depletion on top of αCD40 treatment significantly delayed tumor growth (Fig. 4D), resulting in a prolonged survival compared with mice that received the αCD40 as monotherapy (Fig. 4E). The TME in αCD40 + αCSF1R-treated mice contained fewer TAMs compared with αCD40-monotherapy treated tumors (Supplementary Figs. S4A and S4B), from which the latter included fewer TAMs that expressed ARG1 and MMR (Supplementary Figs. S4C and S4D). Consequently, the abundance of CD8⁺ T cells and their effector T-cell phenotype was increased in tumors of αCD40 + αCSF1R-treated mice (Supplementary Figs. S4E and S4F). Depletion of CD8⁺ T cells 4 days after αCD40/αCSF1R administration prevented the protective effect generated by αCSF1R, showing that this effect was CD8⁺ T-cell dependent (Supplementary Fig. S4G). These results indicate that although TAM depletion was not able to further improve the therapeutic effect of αCD40 during the response phase, αCSF1R treatment could prolong the antitumor responses to CD40 agonist therapy during the delayed regrowth phase.

Given the protumor role played by B16F10 TAMs upon αCD40 treatment during the delayed regrowth phase, we wondered whether the response to αCD40 would differ in preclinical models heavily infiltrated by TAMs during early tumor growth. Therefore, we utilized the LLC model, for which we previously showed the prominence of the myeloid compartment (40, 49). To investigate how the myeloid compartment differs between B16F10 and LLC tumors, we performed a scRNA-seq on the CD45⁺ fraction of LLC tumors at similar tumor volumes as for the B16F10 scRNA-seq experiment (Fig. 5A and B; Supplementary Fig. S5A). Interestingly, the TME of LLC was characterized by a considerable heterogeneous myeloid infiltrate exemplified by expression of Itgam, whereas B16F10 tumors harbored more lymphocytes as indicated by expression of Cd3e (Supplementary Figs. S5B and S5C).

To explore the TAM heterogeneity between both models, we first subclustered the mononuclear populations, containing monocytes and TAMs and subsequently performed trajectory analysis. Some populations such as the monocyte, IFN-signature, and the two hypoxic TAM clusters were represented in both tumor models, whereas other TAM populations such as the TAM-1 and TAM-4 clusters appeared to be unique to B16F10 or LLC tumors, respectively (Fig. 5C and D). On the basis of differentially expressed (DE) genes between these clusters, we found that the TAMs, enriched in B16F10, expressed high levels of H2-DMb1, Cxcl9, and Cxcl10, which are associated with an MHC-II^high M1-like inflammatory TAM phenotype. The TAM clusters enriched in LLC expressed high levels of genes associated with anti-inflammation such as Mrc1, Folr2, and Spp1 (Fig. 5E–J; Supplementary Figs. S5D and S5E). Interestingly, both LLC and B16F10 tumors harbored hypoxic TAM clusters expressing high levels of Arg1, Vegfa, Bnip3, and Hildpa (Fig. 5K–N; Supplementary Fig. S5F).

Trajectory inference using the Slingshot method predicted 3 distinct pseudotime lineages within TAMs (Fig. 5O). Lineage 1 was mainly represented by LLC TAMs, lineage 2 by B16F10 TAMs, and lineage 3, which contained the hypoxic TAMs, was shared by both models, indicating that the distinct monocyte-TAM lineages are tumor-type driven. Of note, cell percentages for each trajectory were calculated to correct for the fact that LLC tumors contained considerably more monocytes/TAMs (Supplementary Fig. S5G). Next, we performed gene ontology (GO) analysis on the DE genes at the end points of lineage 1 versus lineage 2 to further unravel the divergences between TAMs from B16F10 versus LLC tumors. For the genes specific for the LLC TAM trajectory, GO analysis highlighted besides “inflammatory response,” terms related to cell adhesion, response to wound healing, angiogenesis, and negative regulation of cell population proliferation. In contrast, the GO terms “antigen processing and presentation,” “positive regulation of T-cell activation,” and “response to IFNγ” were highlighted for the B16F10 TAM trajectory (Fig. 5P and Q).

Overall, these results demonstrate that B16F10 tumors are enriched with lymphoid cells compared with LLC tumors and hint that monocyte to TAM differentiation and reprogramming is tumor-model specific with B16F10 TAMs developing toward T-cell stimulating cells, whereas LLC TAM develop towards potential wound-healing cells.

On the basis of the inherent differences between the B16F10 and LLC TME, we wondered whether LLC could represent a model with an inherent resistance to αCD40 therapy. Indeed, treatment of LLC tumor-bearing mice with αCD40 as a monotherapy did not reduce tumor progression (Fig. 6A and B). Given the high protumor TAM infiltration into LLC tumors, we combined αCD40 with αCSF1R therapy. This resulted only in a small reduction in tumor growth, nonetheless, slightly repolarized the remaining TAM towards an MHC-II^hi phenotype and increased the neutrophil, CD4⁺ and CD8⁺ T-cell infiltrate, without altering the percentages of Tregs (Fig. 6A–E; Supplementary Figs. S6A–S6C).

We hypothesized that distinct immune players could be responsible for the resistance of LLC tumors towards αCD40 + αCSF1R therapy. When comparing B16F10 and LLC tumors, we observed a >4-fold increase in the abundance of tumor-infiltrating neutrophils in LLC (Fig. 6F). In addition, Tregs were strongly decreased upon αCD40 + αCSF1R treatment only in B16F10 (Fig. 6G). Both Tregs and neutrophils were shown to suppress CD8⁺ T cells in LLC tumors (41, 49), which could be responsible for the lower initial abundance of CD8⁺ T cells in LLC tumors and their inability to expand upon αCD40 + αCSF1R treatment (Fig. 6H).

First, to assess whether expanding the CD8⁺ T-cell number in αCD40 + αCSF1R-treated LLC would result in a therapeutic response, we employed an optimized Flt3L treatment schedule to increase cDC numbers prior to therapy (Supplementary Figs. S6D and S6E). However, although this resulted in an increased CD8⁺ T-cell abundance in αCD40 + αCSF1R-treated mice, tumor growth remained unaltered (Fig. 6I–J). Similarly, when depleting Tregs using an optimized αCD25 antibody regimen (Supplementary Fig. S6f), CD8⁺ T cells were increased upon αCD40 + αCSF1R treatment but did not result in reduced tumor growth (Fig. 6K and L).

Finally, we addressed whether neutrophils would represent a resistance mechanism to αCD40 + αCSF1R therapy. Depletion of neutrophils pharmacologically in LLC tumors using αLy6G/αMAR regimens or CXCR2 inhibitors were unsuccessful when combined with αCSF1R (Supplementary Figs. S6G–S6J). To understand why neutrophils in LLC were not well depleted using CXCR2 inhibitors, we analyzed CXCR2 expression on neutrophils from bone marrow (BM), blood, spleen, and tumor in naïve or LLC tumor-bearing mice. We found that in both the naïve and tumor bearing scenario, approximately 50% of the neutrophils in the BM expressed CXCR2, a receptor required for neutrophil maturation and release from the BM (Fig. 6M; ref. 50). As expected, all neutrophils in blood and spleen expressed CXCR2, but surprisingly 50% of the neutrophils downregulated CXCR2 when reaching the TME. Interestingly, we saw that CXCR2⁺ neutrophils suppressed T cells more compared with CXCR2^– neutrophils (Fig. 6N). In αCD40 + αCSF1R-treated LLC-tumors, the CXCR2⁺ neutrophil population was increased, emphasizing the therapeutic potential of neutrophil depletion in αCD40 + αCSF1R-treated mice (Supplementary Fig. S6K). However, unfortunately, when using Csf3r^–/– mice, in which neutrophils are unable to egress from the BM, neutrophil depletion did not affect tumor growth of αCD40 + αCSF1R-treated mice (Fig. 6O; Supplementary Fig. S6L), implying that still other compensatory mechanisms are responsible for the therapy resistance of LLC tumors.

Finally, we aimed to understand whether alternative complementary therapies could alleviate LLC tumor resistance and sensitize tumors to αCD40 + αCSF1R immunotherapy. Therefore, we inoculated mice with LLC cells expressing the chicken ovalbumin antigen as surrogate tumor antigen (LLC-OVA). In contrast to the results obtained in LLC, tumor growth was significantly reduced in LLC-OVA upon treatment with αCD40 and this response was even improved by the combination of αCD40 + αCSF1R, with the latter treated mice displaying prolonged survival (Fig. 7A–C). This antitumor effect was accompanied with a strong increase in CD8⁺ T-cell abundance and a trend towards an increase in antigen-specific CD8⁺ T cells, together with a repolarization of the remaining TAMs (Fig. 7C–F). Importantly, similar results were observed in mice bearing orthotopic LLC-OVA tumors. Micro-CT scanning revealed that the nonaerated lung volume and total lung volume, both parameters associated with lung tumor burden (35), were both reduced in mice treated with αCD40 + αCSF1R compared with mice treated with isotype, αCD40, or αCSF1R alone (Supplementary Figs. S7A–S7C). These results demonstrate that the presence of strong tumor antigens could resensitize resistant models to CD40 agonist therapy.

Next, we wanted to assess whether treating LLC tumors with ICD-inducing chemotherapy would recapitulate the results obtained in LLC-OVA tumors. ICD inducers have been reported to facilitate DC-based immunogenic phagocytosis of cell corpses, resulting in subsequent antigen specific T-cell activation (51). Hereto, we first evaluated which chemotherapies could induce most potent ICD in LLC. Oxaliplatin, generated the highest NF-κB and type I IFN responses in J774 macrophages cocultured with LLC cells (Fig. 7G and H). In LLC tumor-bearing mice treated with oxaliplatin, cancer cells indeed showed increased caspase-3 activity, indicative for an increased cancer cell death (Fig. 7I). Hence, to increase the rate of immunogenic LLC cell death, we combined αCD40 + αCSF1R with oxaliplatin. Oxaliplatin could significantly reduce LLC tumor progression when used in combination with αCD40 + αCSF1R and increased the proportions of CD8⁺ T cells expressing an effector phenotype and granzyme B (Fig. 7J–N). Moreover, the remaining TAM were repolarized towards an MHC-II^high phenotype and less suppressive Tregs infiltrated tumors treated with the combination therapy (Fig. 7O–R). We found this antitumor effect to be mediated by the combination of the three therapies, as exclusion of αCSF1R resulted in the loss of antitumor immunity (Supplementary Fig. S7D–S7J). Interestingly, when depleting cDC1 using Xcr1^wt/dtr mice during the αCD40 therapeutic window in LLC, we observed reduced sensitivity to the therapy and reduced CD8⁺ T-cell numbers, highlighting the important role cDC1 play in antitumor CD8⁺ T-cell responses in this model (Supplementary Figs. S7K–S7P). The therapeutic effect of the triple therapy was also recapitulated in orthotopic LLC tumors. Mice that had received oxaliplatin + αCD40 + αCSF1R had lower nonaerated lung volume, a proxy for the tumor burden, compared with all other groups (Supplementary Figs. S7Q–S7S).

Overall, these findings show that ICD-inducing chemotherapy could subvert αCD40 + αCSF1R therapy resistance and thereby resensitize resistant tumor models.

Cancer therapies that aim to activate a patient's own immune cells hold a great deal of clinical promise. However, due to the potential to generate extreme adverse events, clinical application of agonist therapies must be performed with caution (52). In the case of CD40 agonists, MTDs have been identified and their clinical use appears safe for patients with solid tumors (53). To optimize CD40 agonist outcome, context-dependent cellular inputs are required for efficacy needs to be identified, as successful αCD40 therapy has been shown to rely on multiple cell types (14, 18, 23). Finally, investigation of appropriate combinatory approaches will ensure that a highest possible proportion of patients can potentially benefit from αCD40 agonist therapy.

In B16F10 tumors, the involvement of CD8⁺ T cells and cDC1s was essential to generate CD40-mediated therapeutic responses. This is in accordance with previous findings that showed, using Batf3 KO mice, that αCD40-mediated responses rely on the cDC1-CD8⁺ T-cell axis in pancreatic ductal adenocarcinoma and bladder cancer (18, 54). However, Batf3 KO mice have two potential drawbacks. First, they genetically lack critical cDC1 functions throughout all stages of tumor progression, making it challenging to evaluate cDC1 contribution in a temporal manner. Second, BATF3 input has been shown recently to be critical for memory CD8⁺ T-cell development, indicating that these mice may also have intrinsic issues in memory CD8⁺ T-cell formation, regardless of their dysfunctional cDC1 pool (55).

Surprisingly, using Xcr1^wt/dtr mice, we observed that cDC1 function was only essential prior to αCD40 administration in the responsive B16F10 model, and that cDC1 were redundant during the actual therapeutic phase of αCD40.

Strikingly, during the later stages of tumor progression, we showed that another cell type, likely cDC2s that upon αCD40-activation adopt a MigDC phenotype, was able to activate CD8⁺ T cells in the absence of cDC1 and TAMs. A transcriptional heterogeneity of cDC2s within mice was demonstrated in multiple cancer types, with similar counterparts also being identified in human cancers (56). Moreover, a population of inflammatory CD64⁺ cDC2s, capable of priming both CD4⁺ and CD8⁺ T cells within a respiratory viral infection setting were recently identified (57), suggesting that cDC1s are not the only cell types able to cross present antigens to CD8⁺ T cells. As such, in the cancer context, human circulating inflammatory CD88⁻CD1c⁺CD163⁺ DCs were shown to regulate tumor immunity (58) and human cDC2s were proposed as critical mediators of cross-presentation of tumor antigens thereby promoting potent antitumor CTL responses (46, 47). Curiously, despite their different functional specializations, both cDC1 and cDC2 adopt an overlapping transcriptional signature upon activation and differentiation to MigDCs (27, 43, 45). Whether the ontogeny-related functions of cDC1 or cDC2 persist despite the altered signature has yet to be proven, although as cDC2 and cDC2-derived MigDCs are not depleted in Xcr1^wt/dtr mice, our data strongly suggest that these cells could be responsible for CD8⁺ T-cell activation in response to αCD40 therapy. As such, although the presence and function of intratumoral cDC1s has been shown an important player for the success of αCD40 therapy, the antitumor functions performed by cDC2s in response to CD40 agonist therapy can be more important than initially thought.

In response to CD40 agonist therapy, B16F10 tumor growth was controlled for multiple days. However, all mice would eventually lose tumor control and display a delayed tumor regrowth occurrence. Similar observations have been made using a combination of CD40L, TNFα, and an antibody against the melanoma antigen TRP1 in which the B16F10 cancer cells formed cell-in-cell structures to avoid immune recognition (59). This process was suggested to be mediated by IFNγ-activated CD8⁺ T cells, and once T cells were no longer present, the cancer cells would disassociate from one another and continue growing. Interestingly, when we depleted TAMs, we observed a higher proportion of CD8⁺ T cells associated with prolonged survival and delayed tumor regrowth. This could suggest that the presence of CD8⁺ T-cell suppressive TAM could be associated with a faster disruption of the CD8⁺ T-cell-mediated cell-in-cell structures and subsequent tumor regrowth.

Because of their plasticity, TAMs represent an important therapeutic target. CD40 therapy has also been shown to rely on the presence and subsequent repolarization of TAMs to generate antitumor immunity (14, 23). The effect of αCD40 therapy on primary tumor growth was not enhanced with TAM depletion in the B16F10 model, yet TAM might have undergone a rapid reprogramming in response to CD40 agonist before their depletion as was shown in other models (14). Interestingly, when merging scRNA-seq data of untreated B16F10 TAMs and TAMs from the heavily infiltrated LLC model, we found considerable tumor-specific heterogeneity and polarization. B16F10 TAMs appeared to be more immune-stimulatory compared with LLC TAMs. Nevertheless, when B16F10 tumors started to regrow, TAMs adopted a MMR⁺ protumor phenotype, at which point the net depletion of TAMs had a beneficial effect on survival. These results hint that a more nuanced approach to TAM depletion may result in better antitumor effects, as broad depletion using αCSF1R may also deplete antitumor TAM populations (32, 49), although more antitumoral TAMs have been shown to rely less on CSF1R for their survival (60).

Finally, with regard to the clinical application of CD40 agonists, our data suggest that different tumor types would benefit from different combinations of therapies. While stimulating CD8⁺ T-cell responses against tumors is a critical facet of any successful immunotherapy, our results indicate that an underlying CD8⁺ T-cell response must exist in order for αCD40 to function as a monotherapy. Therefore, if patient stratification occurs based on tumor CD3 complexity, further assessment should be made to determine whether the T cells present can recognize relevant tumor antigens or not since the intratumoral CD8⁺ T-cell pool has been shown to contain considerable irrelevant CD8⁺ T cells (61, 62). Patients with immune desert tumors would require more nuanced combination therapies, such as chemotherapy or radiotherapy, that would aim to generate an antitumor immune response that could be amplified with αCD40 therapy (20, 25, 26). Although CD40 agonists have achieved beneficial clinical outcomes in pancreatic cancer when combined with chemotherapy (20), the potential for toxicity remains a major caveat limiting their use (53, 63, 64). Encouragingly, although we suggest DCs as one of the critical mediators of antitumor immunity in response to αCD40, DCs have recently been shown to not be involved in αCD40-associated toxicity (65), suggesting that another layer of potential therapeutic combinations exist that could offset tissue damage while preserving the antitumor function of αCD40.

A. Murgaski reports grants from Fonds Wetenschappelijk Onderzoek (FWO) and Kom op tegen Kanker (KOTK) during the conduct of the study. E. Hadadi reports grants from Research Foundation Flanders (FWO) during the conduct of the study. S. Hoves reports a patent 20200392234 pending, a patent 20190284284 pending, a patent 20180346581 pending, a patent 20170051065 pending, a patent 20160220669 pending, a patent 9192667 issued, a patent 20150073129 pending, and a patent 20140314771 pending; and also reports employment with Roche Diagnostics GmbH and ownership of Hoffmann La Roche stock. S. Deschoemaeker reports grants from Stichting tegen Kanker post-doctoraal mandaat (2021-023) during the conduct of the study. M. Schmittnaegel reports employment with Roche Diagnostics GmbH. C.H. Ries reports personal fees from Roche during the conduct of the study; personal fees from Macomics Ltd. and Dr. Carola Ries Consulting outside the submitted work; also has a patent for EP12158519 issued. D. Laoui reports other support from Roche and grants from FWO, VUB, and Collen-Francqui Foundation during the conduct of the study. No disclosures were reported by the other authors.

A. Murgaski: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. M. Kiss: Conceptualization, investigation, writing–review and editing. H. Van Damme: Investigation, writing–review and editing. D. Kancheva: Data curation, software, formal analysis, visualization, methodology, writing–review and editing. I. Vanmeerbeek: Data curation, formal analysis, validation, investigation, visualization, methodology. J. Keirsse: Investigation. E. Hadadi: Investigation, writing–review and editing. J. Brughmans: Investigation. S.M. Arnouk: Investigation. A.E. Hamouda: Investigation. A. Debraekeleer: Investigation. V. Bosteels: Resources, investigation, methodology, writing–review and editing. Y. Elkrim: Investigation. L. Boon: Resources. S. Hoves: Resources, writing–review and editing. N. Vandamme: Resources, data curation, software, investigation, methodology. S. Deschoemaeker: Resources, investigation, writing–review and editing. S. Janssens: Resources, methodology, writing–review and editing. A.D. Garg: Data curation, formal analysis, validation, investigation, visualization, methodology, writing–review and editing. G. Vande Velde: Investigation, methodology. M. Schmittnaegel: Resources, writing–review and editing. C.H. Ries: Resources. D. Laoui: Conceptualization, resources, data curation, supervision, funding acquisition, writing–review and editing.

The authors thank Kasia Błażejczyk, Ella Omasta, Marie-Therese Detobel, Nickey Van Riebeek, Nadia Abou, and Christopher Stanley for technical assistance and administrative assistance. They thank Pascal Merchiers, Akiko Iwasaki, and Orr-El Weizman for providing reagents. The authors thank the VIB singularity platform for support and access to scRNA-seq technologies and Ria Roelandt for the library preparations. They thank Mikaël Pittet, Patrick De Baetselier, Jo Van Ginderachter, and Kiavash Movahedi for insightful discussions. A. Murgaski, H. Van Damme, I. Vanmeerbeek, and S.M. Arnouk are supported by an FWO doctoral fellowship (1S16718N, 1S24117N, 1S06821N, and 1S78120N). S. Janssens and V. Bosteels are funded by an EOS grant (G0G7318N) and FWO predoctoral fellowship. A.D. Garg was supported by FWO (G0B4620N; EOS grant 30837538), KU Leuven, Kom op Tegen Kanker, and VLIR-UOS. G. Vande Velde was supported by KU Leuven Internal Funds (C24/17/061 and STG/15/024). D. Laoui was supported by grants from FWO (12Z1820N), Kom op Tegen Kanker, Stichting tegen kanker, and Vrije Universiteit Brussel.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.