The C5a/C5aR1 Axis Promotes Migration of Tolerogenic Dendritic Cells to Lymph Nodes, Impairing the Anticancer Immune Response

Dendritic cells (DC) are a heterogeneous group of antigen-presenting cells endowed with potent immunostimulatory properties and the ability to induce, maintain, and regulate T cell–mediated antitumor responses (1). DCs can be subclassified according to their ontogeny, phenotype, tissue distribution, and function into conventional DCs (cDC), which include type 1 cDCs (cDC1) and type 2 cDCs (cDC2), plasmacytoid DCs (pDC), and monocyte-derived DCs (moDC; ref. 2). cDC1s have a unique ability to cross-present MHC I–associated antigens to CD8⁺ T cells, both within the tumor microenvironment (TME) and after migration to tumor-draining lymph nodes (LN), which is essential for CD8⁺ T cell–mediated antitumor responses and cancer immunotherapy (3). cDC2s efficiently present MHC II–associated antigens to CD4⁺ T cells, although their role in cancer immunology is less understood (4). pDCs are producers of type I IFNs and are involved in antiviral and antitumor immune responses (4). Finally, moDCs are prominent in inflammatory sites, such as the TME (4). Several antitumor strategies have been developed to increase the abundance and functionality of DCs (5). These treatments have shown promising results in preclinical studies, but mixed outcomes in clinical trials. A better understanding of DC biology in the context of cancer may provide valuable insights for improving DC-driven therapies aimed at enhancing antitumor immunity.

The complement system, a component of the innate immune response, profoundly influences the characteristics of the TME (6, 7). The anaphylatoxin C5a, a key biological factor released upon activation of all complement pathways, is involved in multiple aspects of tumor biology, including immune evasion, progression, and metastasis (8, 9). C5a exerts its tumor-promoting activities through its canonical G protein–coupled receptor, C5a receptor 1 (C5aR1, also known as CD88), which is expressed on various cell populations of the TME (8), but primarily on components of the myeloid compartment (10, 11). Our group and others have reported the impact of the C5a/C5aR1 axis on the recruitment and modulation of tumor-associated macrophages (TAM) and myeloid-derived suppressor cells (MDSC), both in the primary tumor and in the metastatic niche (12–16). Activation of C5aR1 in myeloid cells contributes to the establishment of an immunosuppressive microenvironment, thereby impeding the efficacy of cancer immunotherapies (12, 13). Accordingly, inhibition of C5a signaling significantly improved the antitumor activity of PD-1/PD-L1 blockade in preclinical cancer models (17). Combined inhibition of PD-1/PD-L1 and C5aR1 has also been translated into clinical trials (NCT03665129; NCT04812535).

In addition to MDSCs and TAMs, DCs constitute the third element of the myeloid compartment within the TME. However, the function of C5a/C5aR1 in tumor-associated DCs remained completely unexplored. In the present study, we found that tumor-infiltrating DCs overexpress C5aR1. This expression was restricted to tolerogenic cDC2s and moDCs and drove the migration of C5aR1-expressing DCs to tumor-draining LNs, where they could inhibit cDC1-mediated activation and proliferation of CD8⁺ T cells. We used this knowledge to propose an effective combined immunotherapy based on the activation of DCs with the Toll-like receptor (TLR) 3 agonist poly I:C in combination with PD-1 blockade. The addition of C5aR1 inhibition to poly I:C and PD-1 blockade helped establish a more favorable environment for effective tumor rejection, resulting in a more potent antitumor effect. Our findings support the therapeutic potential of modulating the complement system in DCs to enhance immune responses against tumors.

Lewis lung carcinoma (LLC) cells were obtained in 2008 from the ATCC, (CRL-1642, RRID: CVCL_4358). LLC-OVA cells were generated after stable transduction of LLC cells with a lentiviral vector encoding OVA and were kindly provided by Dr. David Escors (Navarrabiomed Biomedical Research Center, Pamplona, Spain; ref. 18). 393P lung adenocarcinoma cells, derived from Kras^LA1/+;p53^R172HΔG mice, and B16.F10 melanoma cells were kindly provided by Dr. Silve Vicent in 2015 and Dr. Fernando Aranda in 2022 (both of Cima Universidad de Navarra), respectively. Cells were cultured in RPMI-1640 supplemented with GlutaMAX (Gibco), 10% FetalClone (Thermo Fisher Scientific), 100 U/mL penicillin, and 100 µg/mL streptomycin (Invitrogen). This medium is referred to as complete medium. No cell line authentication was performed within the last year. Cells were passaged approximately every 3 days to ensure that cultures remained subconfluent. The number of cell passages was not recorded. Cells were routinely tested for Mycoplasma.

PMX53, a cyclic hexapeptide that inhibits C5aR1, was synthesized as described previously (19). Mouse C5a, human C5a, and mouse CCL19/MIP-3 were purchased from R&D Systems. Dexamethasone, fedratinib, and ruxolitinib were obtained from MedChemExpress. IL4, IL6, IL10, and GM-CSF were purchased from PeproTech. Recombinant human Flt3L-Fc was purchased from Bio X Cell. iTAg Tetramer/APC - H-2Kb OVA (SIINFEKL) was purchased from MBL International Corporation. Invitrogen provided SYTOX Blue. Poly I:C was obtained from Amersham. Anti–PD-1 (clone RMP1-14, BE0146, RRID: AB_10949053) was obtained from Bio X Cell. FTY720 was purchased from Sigma-Aldrich. The antibodies targeting mouse antigens used in this study and their RRIDs are listed in Supplementary Table S1.

A multiplex immunofluorescence protocol based on tyramide signal amplification was used for the simultaneous detection of CD11c, C5aR1, and DAPI in two formalin-fixed paraffin-embedded (FFPE) tissue blocks of human lung adenocarcinoma according to a previously described procedure (20). The study was carried out according to The Code of Ethics of the World Medical Association (Declaration of Helsinki). The protocol was approved by the Research Ethics Committee of the University of Navarra (Reference 108-2013), and all patients gave written informed consent. Patients were recruited at the Cancer Center Clínica Universidad de Navarra. Lung primary tumors were fixed in 10% buffered formalin after surgical removal and paraffin-embedded using standard procedures. For multiplexed immunofluorescence staining, 4-μm-thick FFPE sections were deparaffinized and rehydrated in a BOND staining system (Leica Biosystems). Each section was subjected to two sequential rounds of antibody staining, including heat-induced antigen retrieval in BOND epitope retrieval solution pH 9, protein blocking with Antibody Diluent/Block (Akoya Bioscience), and incubation with primary antibodies (CD11c, 1:500, AB52632, Abcam; C5aR1, 1:10, NBP3-12085, Novus Biologicals) and anti-rabbit secondary antibodies (Akoya Biosciences), followed by tyramide signal amplification with the fluorophores Opal 620 (CD11c) or Opal 690 (C5aR1) diluted in Plus Amplification Diluent (Akoya Biosciences). Subsequently, nuclei were counterstained with spectral DAPI (Akoya Biosciences), and sections were mounted with Faramount Aqueous Mounting Medium (Dako).

For LLC tumor samples, a seven-color multiplex immunofluorescence assay for the simultaneous detection of CD8, CD4, FoxP3, CD11b, and DAPI (FP-1490, PerkinElmer) was performed. The antibodies used and their RRIDs are listed in Supplementary Table S1.

For analysis, spectral unmixing of the LLC tumors and human lung adenocarcinoma images and determination of autofluorescence using unstained samples were performed. Image analysis was carried out using the inForm software framework (version 2.4.8, Akoya Biosciences, RRID: SCR_019155).

Bone marrow progenitors were harvested from 8- to 12-week-old female C57BL/6 mice (Envigo). Erythrocytes were removed as described previously (17). The remaining cells were then plated in RPMI-1640 supplemented with GlutaMAX, 10% FetalClone, 100 U/mL penicillin, and 100 µg/mL streptomycin along with appropriate cytokines for differentiation. CD103⁺ bone marrow-derived DC (BMDC), CD11b⁺ CD103⁻ BMDCs, and monocytic BMDCs (moBMDC) were generated as previously described (21–23). Briefly, CD103⁺ BMDCs were obtained by incubating BM progenitors with 200 ng/mL Flt3L and 3.33 ng/mL GM-CSF for 12 to 14 days. CD11b⁺ CD103⁻ BMDCs were generated by culturing bone marrow progenitors with 100 ng/mL Flt3L for 8 days. moBMDCs were generated by using decreasing concentrations of GM-CSF (from 20 to 5 ng/mL) and 10 ng/mL IL4 for 10 days. In some experiments, after differentiation, C5a (10 nmol/L), dexamethasone (1 µmol/L), IL6 (150 ng/mL), IL10 (50 ng/mL), GM-CSF (20 ng/mL), or conditioned medium from LLC cells (diluted 1:5 in culture medium) was added. Cells were then incubated for another 24 hours and harvested for further studies.

Human-derived DCs were obtained from human peripheral blood mononuclear cells. To isolate peripheral blood mononuclear cells, heparinized blood was diluted in Ficoll-Paque (GE Healthcare) from a healthy donor and centrifuged at 800 g for 25 minutes at room temperature without brake. After centrifugation, mononuclear cells were collected from the boundary layer. Monocytes were allowed to adhere for 4 hours by culturing them at a concentration of 5 × 10⁶ cells/mL of complete medium at 37°C and 5% CO₂. Nonadherent cells were then removed and adherent cells were washed twice with PBS. Adherent cells were cultured for 6 days with cell culture medium supplemented with 100 ng/mL GM-CSF and 50 ng/mL IL4. On day 3, a fresh medium containing cytokines was added. On day 6, nonadherent cells were collected for further studies.

BMDCs were incubated with 2 µmol/L Fluo-4 (F-14201, Life Technologies) diluted in RPMI-1640 supplemented with 2 mmol/L glutamine (BE17-605E, Lonza) and 250 nmol/L probenecid (Invitrogen, P36400) at 37°C for 30 minutes. The cells were then analyzed using a FACSCanto II flow cytometer (BD Biosciences). After measuring basal calcium for 2 minutes, C5a (10 nmol/L) was added to cells preincubated with or without PMX53 (1.4 µmol/L), and cells were acquired for an additional 3 minutes. Data were analyzed using FlowJo software (v.10, Tree Star, RRID: SCR_008520).

Animal experiments were performed according to protocols approved by the institutional animal care committee of the University of Navarra (references 049-18 and 131-22). Sv/129 mice were obtained from Janvier and C57BL/6J mice from Envigo. Vert-X (B6(Cg)-Il10tm1.1Karp/J) mice, Kikume Green-Red (KikGR) mice, and OT-I and OT-II T-cell receptor (TCR) transgenic mice were purchased from The Jackson Laboratory. Xcr1-DTR mice were obtained from Riken BRC. C5ar1-deficient mice have been previously described (24).

To obtain Xcr1-DTR chimeras, naïve C57BL/6J mice were irradiated with two doses of 500 rad (5 Gy), 24 hours apart. Subsequently, 2 × 10⁶ bone marrow cells harvested from Xcr1-DTR mice were intravenously injected and allowed to proliferate for 5 weeks before tumor implantation. For XCR1⁺ DC depletion, a dose of 20 ng/g diphtheria toxin (Sigma-Aldrich) was administered intraperitoneally on the first day of treatment, followed by dosage of 4 ng/g every 3 days (21).

LLC cells (2 × 10⁶) were injected subcutaneously into the flank of 8- to 12-week-old female C57BL/6J, Vert-X or Xcr1-DTR mice. Tumor-bearing mice were treated with vehicle, 1-mg/kg PMX53 (s.c.; daily starting on day 6), 100 μg anti–PD-1 (i.p.; days 7, 10, and 14), and/or 50 μg poly I:C (i.t.; days 8, 11, 15, and 18).

In KikGR mice, LLC tumors were treated with vehicle or the C5aR1 inhibitor (12, 13, 14, and 15 days after 2 × 10⁶ LLC cell inoculation). On day 15, tumors were irradiated for 15 minutes with 200 milliwatts of violet light (405-nm wavelength) for KikGR photoconversion using a high-power UV-LED solo P lamp (Opsytec). Flow cytometry was used to quantify KikGR-red cells in tumor-draining LNs 24 hours later.

In vivo experiments with i.t. injection of BMDCs were performed as follows. BMDCs were differentiated with GM-CSF and IL4 as described above using bone marrow progenitors from C5ar1-wild-type or C5ar1-knockout mice. On days 7 and 10 after 2 × 10⁶ LLC cell implantation, BMDCs (1 × 10⁶ cells per mouse) were inoculated i.t. The immune infiltrate in the TME was analyzed by flow cytometry on day 13.

C57BL/6J mice were subjected to CD8⁺, CD4⁺, or NK cell depletion by intraperitoneal injection of 100-μg anti-mouse CD8α (clone 2.43; Bio X Cell, BE0061, RRID: AB_1125541), CD4 (clone GK1.5; Bio X Cell, RRID: AB_10013429) or NK1.1 (clone PK136; Bio X Cell, RRID: AB_1107737), respectively, as previously described (25). To prevent lymphocyte egress from secondary lymphoid organs, 20 μg of FTY720 (Sigma-Aldrich) was administered per animal intraperitoneally every 2 days, starting on day 6 after tumor implantation (26). Effective inhibition of lymphocyte egression was confirmed by quantification of TCRβ⁺ lymphocytes in the blood.

393P cells (4 × 10⁶) were injected subcutaneously into the flanks of 8- to 12-week-old female Sv/129 mice. After 10 days, 393P tumors were treated with the C5aR1 inhibitor PMX53 (from day 10), anti–PD-1 (days 11, 14, and 17), and/or poly I:C (days 11, 15, 18, and 22).

B16.F10 cells (0.5 × 10⁶) were injected subcutaneously into the flanks of female C57BL/6J mice aged 8 to 12 weeks and allowed to grow for 4 days. Mice were then treated with vehicle, PMX53 (s.c.; daily starting on day 4), anti–PD-1 (i.p.; days 5, 8, and 12), and/or poly I:C (i.t.; days 5, 9, 12, 15, and 19).

Rechallenge experiments were performed in cured mice at least 3 months after tumor regression using the same number of cells and conditions as described above.

Spleens and tumor-draining LNs were harvested from mice and mechanically disaggregated. In the case of tumors, single-cell suspensions were obtained by mechanical and enzymatic disaggregation using 1 mg/mL collagenase D (Roche) and 50 µg/mL DNase I (Roche) at 37°C for 30 minutes. EDTA (6 µmol/L) was then used to block collagenase and DNase I activities, and erythrocytes were removed as previously described (17). For flow cytometry analysis of the ex vivo generated DCs, 10⁶ nonadherent cells were harvested from tissue culture plates.

Single-cell suspensions were preincubated with a monoclonal antibody against mouse CD16/CD32 (1:200; Fc block; clone 2.4G2, BD Pharmingen) for 15 minutes at 4°C and stained on ice for 15 minutes with fluorochrome-conjugated diluted in FACS buffer (PBS, 5% FetalClone, and 2.5 mmol/L EDTA). For intracellular staining, cells were fixed and permeabilized with fixation/permeabilization buffer (eBioscience) and labeled for intracellular markers in permeabilization buffer (eBioscience). The antibodies used and their RRIDs are listed in Supplementary Table S1. Cell viability was assessed using PromoFluor 840 (1:2,560, PromoKine). Cells were analyzed using a Beckman Coulter CytoFLEX LX flow cytometer (Beckman Coulter). Data were analyzed using FlowJo software (v.10, Tree Star, RRID: SCR_008520). Gating strategies are shown in Supplementary Figs. S1–S4.

Ovalbumin (OVA) TCR-specific CD8⁺ and CD4⁺ T cells were isolated from the spleens of OT-I and OT-II transgenic mice using mouse CD8⁺ and CD4⁺ T-cell isolation kits (Miltenyi), respectively. Nonspecific CD8⁺ and CD4⁺ T cells were isolated from the spleens of C57BL/6 mice. To obtain tumor-infiltrating DCs, LLC-OVA or LLC tumors between 400 and 800 mm³ were harvested and processed under sterile conditions as described in the section on flow cytometry analysis. After erythrocyte lysis, cells were resuspended in 35% Percoll (GE Healthcare) and centrifuged at 870 g for 15 minutes without brake. The supernatant was discarded and cells were preincubated with Fc block for 15 minutes at 4°C and then labeled with fluorochrome-conjugated antibodies against CD45, MHC-II, CD11c, F4/80, and C5aR1 (Supplementary Table S1). Dead cells were labeled with SYTOX Blue (Invitrogen). C5aR1^high and C5aR1^low DCs were sorted using MoFlo Astrios (Beckman Coulter). Supplementary Fig. S5 shows the gating strategy used to classify cells as C5aR1^high or C5aR1^low. DCs and T cells (OT-I CD8⁺, OT-II CD4⁺, CD8⁺, or CD4⁺ T cells) were plated in a 1:3 ratio in rounded 96-well plates. For the nonspecific assays, wells were previously coated with anti-CD3 (clone 145–2C11, BD Pharmingen; 1 µg/mL) and anti-CD28 (clone 37.51, BD Pharmingen; 0.5 µg/mL). After 24 hours for OVA-specific assays or 48 hours for nonspecific assays, proliferation (Ki-67) and T-cell activation (granzyme B, IFNγ, GITR, and PD-1) were assessed by flow cytometry.

CD4⁺ T cells were isolated from spleens of C57BL/6 mice. Tumor-infiltrating DCs were isolated from LLC tumors as described above. CD4⁺ T cells (10⁵) and tumor-infiltrating DCs (10³) were plated in 96-rounded wells previously coated with anti-CD3 (1 µg/mL) and anti-CD28 (0.5 µg/mL). Human IL2 (100 U/mL, # 200-02, PeproTech) was added on days 0 and 3. The number of regulatory T cells (Tregs; CD4⁺CD25⁺FoxP3⁺) was quantified on day 5 by flow cytometry.

Tumor-infiltrating DCs were isolated from LLC tumors using an autoMACS Pro Separator with anti-CD11c microbeads (Miltenyi Biotech). Total DCs were incubated with FITC-conjugated dextran (FD40S, Sigma-Aldrich) at 1 mg/mL in complete medium for 2 hours at 37°C. Cells were washed with PBS, stained with fluorochrome-conjugated antibodies, as described above, and analyzed by flow cytometry.

For Luminex analysis, 20,000 tumor-infiltrating DCs were incubated in complete medium for 24 hours. The supernatants were then collected and cytokines were determined using the Mouse Cytokine & Chemokine Panel 1 (26 plex; Procarta) in a MAGPIX instrument (Luminex) following the manufacturer’s instructions. Quantification of murine IL10 and C5a was performed by ELISA (DY417-05 and DY2037, respectively; R&D Systems) according to the manufacturer’s instructions. Expression of signaling proteins was analyzed by Western blotting using anti-mouse phosphorylated and total ERK1/2, IKKα/β, and AKT antibodies. GAPDH was used as a protein loading control. Antibody sources, their RRIDs, and dilutions are listed in Supplementary Table S1.

Migration assays were performed with BMDCs or tumor-infiltrating DCs isolated from LLC tumor–bearing mice using an AutoMACS Pro Separator with anti-CD11c microbeads (Miltenyi Biotech). DCs (2 × 10⁵) were plated on the upper chamber of 5 μm pore size Transwells (Corning Costar) in RPMI-1640 supplemented with 0.5% FetalClone. Migration stimuli, either C5a or CCL19, were added to the bottom chamber in RPMI-1640 with 0.5% FetalClone. After 10 hours, DCs in the bottom chamber were quantified using a CytoFLEX LX flow cytometer (Beckman Coulter).

For bulk RNA sequencing (RNA-seq), LLC tumors were harvested, frozen, and mechanically disaggregated on dry ice. RNA was then extracted using the NucleoSpin RNA isolation kit (Macherey-Nagel) and subjected to quantity and quality control using the Qubit HS RNA assay kit (Thermo Fisher Scientific) and 4200 TapeStation with High Sensitivity RNA ScreenTape (Agilent Technologies). All RNA samples had an RNA integrity number >7. Library preparation was performed using the Illumina Stranded mRNA Prep Ligation kit (Illumina) according to the manufacturer’s protocol. All sequencing libraries were prepared from 100-ng total RNA. Library quality and quantity were verified as described above. Libraries were then sequenced on a NextSeq 2000 sequencer (Illumina). Twenty to thirty million pair-end reads (100 bp; Rd1:51; Rd2:51) were sequenced for each sample and demultiplexed using Cutadapt (RRID: SCR_011841). RNA-seq was performed at the Genomics Unit of Cima Universidad de Navarra. Quality control and data processing were performed as follows: (i) quality control of raw sequence data using FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/, RRID: SCR_017148), (ii) read trimming using Trimmomatic (27), (iii) alignment to the mouse reference genome (GRCm39) using STAR (28), (iv) quantification of read counts mapping exonic gene regions using featureCounts (29), and (v) gene annotation using GENCODE vM31 (30).

Statistical analysis of differential expression and graphs was performed using R (https://www.R-project.org/, RRID: SCR_001905). For analysis, gene expression data were filtered and normalized using edgeR (31) and voom (32). Only protein-coding genes were included in the analysis. Genes with low read counts were removed using the default settings of the filterByExpr function. limma (32) was used to identify genes with significant differential expression between experimental conditions. Genes with adjusted P values of <0.05 and log₂FC > 1 or < −1 were considered differentially expressed. Gene set enrichment analysis (GSEA) was performed using the R packages fgsea and the Gene Ontology: Biological Process gene set. Genes were ranked using the formula: sign (log₂FC) × −log₁₀ (adjusted P value).

Single-cell RNA sequencing (scRNA-seq) data from treatment-naïve patients with lung adenocarcinoma were downloaded from Gene Expression Omnibus (accession number: GSE131907, RRID: SCR_005012). Sample details and data processing have been described previously (33). Normalized counts from 10 primary lung adenocarcinomas (stages IA to IIIA) and seven metastatic LNs were analyzed (of note, the LUNG_T30 sample was excluded based on the visualization of the DC subset projection). Cells annotated as DCs were divided into C5AR1⁻ and C5AR1⁺ DCs based on the raw count matrix. Cells with at least one read for C5AR1 mRNA were considered C5AR1⁺. R was used for analysis and graphing. Differentially expressed protein-coding genes were identified using a scalable implementation of the Wilcoxon rank sum test from the presto package. Genes with an adjusted P value of <0.05 were considered to be differentially expressed. GSEA was performed as described above. Genes were ranked using the formula: −log₁₀ (adjusted P value).

The statistical tests used to analyze the RNA-seq data are indicated in the previous section. The Shapiro–Wilk test was used to determine the normal distribution of variables. A student t test or one-way ANOVA with the Tukey post hoc test were used to compare differences between two or more experimental groups, respectively. Expression scores were calculated from each sample as the mean of the z-scores of the genes included in the score. A P value of <0.05 was considered statistically significant. Statistical analyses were performed using GraphPad Prism 9 software (RRID: SCR_002798).

The bulk RNA-seq data were deposited in the NCBI Gene Expression Omnibus database under accession number GSE239485. The remaining data generated in this study are available in the article and its accompanying Supplementary Materials or upon request from the corresponding author.

We evaluated the expression of C5aR1 by flow cytometry in different immune populations belonging to the myeloid compartment from blood, spleen, and tumors of mice bearing LLC tumors. Tumor-infiltrating myeloid cells showed high expression of C5aR1 as compared with myeloid cells in spleen and blood (Fig. 1A; Supplementary Fig. S6), with the highest levels detected in polymorphonuclear MDCSs (PMN-MDSCs; F4/80⁻CD11c⁻CD11b⁺Ly6G^highLy6C^interm) and M2-like TAMs (CD11b⁺F4/80⁺CD206⁺). Although previous research has extensively investigated the role of the C5a/C5aR1 axis in TAMs and MDSCs (10, 11), nothing is known about the expression of C5aR1 and the significance of this pathway in cancer-associated DCs. Expression of C5aR1 in tumor-infiltrating DCs was also observed in human lung adenocarcinomas (Fig. 1B). Based on these findings, we aimed to characterize the phenotype and function of C5aR1-expressing tumor-infiltrating DCs.

We analyzed the transcriptome of human tumor-infiltrating DCs using scRNA-seq data from 10 treatment-naïve primary lung adenocarcinomas (33). Of the 1,424 cells categorized as DCs, a total of 350 cells expressed C5AR1 mRNA. Differentially expressed genes between C5AR1⁺ and C5AR1⁻ DCs were associated with pathways related to DC biology such as “phagocytosis” and “antigen processing and presentation” (Fig. 1C and D). We also examined the association between the expression of C5AR1 mRNA and the expression of a set of genes related to a tolerogenic DC phenotype (34). Most of these genes were upregulated in C5AR1⁺ DCs compared with C5AR1⁻ DCs (Fig. 1E). An analysis of cytokines characteristically expressed by DCs also revealed an immunosuppressive cytokine profile in C5AR1⁺ DCs (Fig. 1F).

Next, we used flow cytometry to quantify the levels of DC activation and maturation markers in DCs differentiated from bone marrow progenitor cells isolated from wild-type and C5ar1-knockout mice. C5ar1-deficient BMDCs showed significantly higher levels of DC activation and maturation markers compared with their wild-type counterparts (Fig. 1G). In addition, treatment of wild-type BMDCs with dexamethasone, a synthetic steroid that converts antigen-presenting cells into tolerogenic DCs (35), decreased the expression of DC activation markers but increased the expression of C5aR1 (Fig. 1H). The same result was obtained when DCs differentiated from human blood were treated with dexamethasone (Fig. 1I). Taken together, these data suggest that C5aR1 is expressed in tumor-infiltrating DCs that display a tolerogenic state.

Factors present in the TME may be inducing the expression of C5aR1 in tumor-infiltrating DCs. To test this hypothesis, we differentiated bone marrow progenitors into different DC subsets and treated these cells with conditioned medium from LLC tumors. Basal C5aR1 expression was significantly higher in moBMDCs than in CD103⁺ BMDCs or CD11b⁺ BMDCs (Supplementary Fig. S7A). Incubation with an LLC-conditioned medium significantly increased C5aR1 expression in all BMDC subsets (Fig. 2A). Incubation with IL6 or IL10, which have been reported to induce immunosuppression in tumor-infiltrating DCs (36, 37), also increased C5aR1 expression (Fig. 2A; Supplementary Fig. S7B). In contrast, GM-CSF, a proinflammatory cytokine that has been associated with DC-mediated CD8⁺ T-cell responses (38), had no effect. Incubation with the JAK inhibitors fedratinib and ruxolitinib blocked the induction of C5aR1 expression mediated by IL10, but not by IL6, suggesting the induction of different signaling mechanisms by these two cytokines (Supplementary Fig. S7C and S7D). Incubation with IL6 or IL10 of splenic DCs from naïve C57BL/6 mice induced the expression of C5aR1 in all DC subtypes but mainly in cDC2s (Fig. 2B).

Next, we evaluated the expression of C5aR1 in DC subpopulations isolated from tumors and tumor-draining LNs of LLC tumor–bearing mice. The proportion of DCs and the expression of C5aR1 in these cells were significantly higher in tumors than in tumor-draining LNs (Fig. 2C). cDC2s and moDCs were the predominant subpopulations of DCs expressing C5aR1 (Fig. 2D). The same result was obtained in B16.F10 tumor–bearing mice (Fig. 2E and F), although the proportion of DCs in the tumors and their level of expression of C5aR1 were noticeably lower in this model. We also analyzed the expression of C5aR1 in tumor-infiltrating DC subsets from the scRNA-seq study of human lung adenocarcinomas described in the previous section. We subcategorized the annotated DCs into cDC1s, cDC2s, moDCs, and pDCs using canonical markers (Fig. 2G). Consistent with the mouse data, C5AR1 mRNA expression was restricted to cDC2s and moDCs (Fig. 2H). Furthermore, C5aR1 expression was higher in tumor-infiltrating cDC2s and moDCs than in the same DC subtypes present in tumor-draining LNs (Fig. 2I). These data suggest that C5aR1 expression is induced by immunosuppressive cytokines found in the TME and is predominantly expressed by tumor-infiltrating cDC2s and moDCs.

We investigated the ability of C5aR1-expressing DCs to activate CD8⁺ T cells using the MHC class I–restricted OT-I system. Tumor-infiltrating DCs isolated from LLC-OVA tumors were separated into low and high C5aR1 expressers (C5aR1^low and C5aR1^high DCs, respectively). In contrast to C5aR1^low DCs, C5aR1^high tumor-infiltrating DCs failed to induce proliferation and activation of naïve OVA-specific CD8⁺ T cells ex vivo (Fig. 3A). To gain some mechanistic insight into this lack of cross-presentation capacity, we performed an analysis of the set of genes related to the biological processes “phagocytosis” and “antigen processing and presentation” previously identified by GSEA in C5AR1⁺ human DCs (Fig. 1D). Most of the genes presented in the “phagocytosis” pathway were overexpressed in C5AR1⁺ human DCs (Supplementary Fig. S8A), suggesting a higher phagocytic capacity in these cells. In the case of “antigen processing and presentation,” genes such as TAP1, TAP2, TAPBP (tapasin), and CD74, as well as HLA genes, were downregulated in C5AR1⁺ DCs, whereas others, such as cathepsins D, L, and S, were upregulated (Fig. 3B). Consistent with the observation in human DCs, LLC tumor–infiltrating C5aR1^high DCs showed a higher ability to phagocytose FITC-dextran than C5aR1^low DCs (Fig. 3C). In the case of antigen presentation, tumor-infiltrating C5aR1^high DCs isolated from LLC-OVA tumors showed an increased expression of H2K^b/SIINFEKL complexes (Fig. 3D). Moreover, expression of the costimulatory ligands CD80 and CD86 was higher in LLC tumor–infiltrating C5aR1^high DCs; however, a decrease in MHC-II and an increase of PD-L1 expression was observed (Fig. 3E).

Next, using both OVA-specific and nonspecific CD8⁺ T cells, we tested the effect of C5aR1^high DCs on the activity of C5aR1^low DCs. C5aR1^high DCs from LLC-OVA or LLC tumors inhibited the ability of C5aR1^low DCs to induce the proliferation of OT-I CD8⁺ T cells or nonspecific CD8⁺ T cells, respectively (Fig. 3F). Expression of the activation markers PD-1, GITR, and granzyme B was also reduced (Supplementary Fig. S8B). Even though C5aR1^high DCs reduced PD-1 expression, PD-1 blockade did not reverse the immunosuppressive capacity of C5aR1^high DCs (Supplementary Fig. S8C).

Similar effects were observed for CD4⁺ T cells. C5aR1^high DCs inhibited the ability of C5aR1^low DCs to activate OT-II CD4⁺ T cells or nonspecific CD4⁺ T cells (Fig. 3G). C5aR1^high DCs were able to block the induction of GITR, PD-1, and IFNγ expression mediated by C5aR1^low DCs, whereas no remarkable differences were observed for IL2 or IL4 (Supplementary Fig. S8D). Moreover, C5aR1^high DCs showed a significantly higher capacity to induce ex vivo Treg differentiation than C5aR1^low DCs (Fig. 3H).

Based on these results, we sought to evaluate the ability of C5aR1-expressing DCs to control tumor growth in vivo. Due to limitations in the number of tumor-infiltrating DCs that can be obtained from mice, we performed these experiments using BMDCs differentiated from wild-type or C5ar1-deficient mice. Tumors inoculated with C5ar1-deficient DCs were significantly smaller than those inoculated with wild-type DCs (Fig. 3I). In addition, these tumors showed a significant reduction in the number of tumor-infiltrating Tregs (Fig. 3J). No significant differences between experimental groups were observed for other immune cell subsets (Supplementary Fig. S9). Taken together, these experiments identify C5aR1-expressing DCs as mediators of tumor progression through their ability to induce suppression of antitumor T-cell responses.

To evaluate the functionality of C5aR1 in DCs, we first treated BMDCs with C5a and evaluated its downstream signaling. C5a treatment resulted in an increase in intracellular calcium, which was abolished by the addition of the C5aR1 antagonist PMX53 (Fig. 4A). Consistent with C5aR1 expression levels, the highest stimulation was observed in moBMDCs, followed by CD11b⁺ BMDCs. C5a treatment also resulted in the activation of MAPK and NF-κB pathways, as determined by ERK1/2 and IKKα/β phosphorylation, respectively (Fig. 4B). These results demonstrate that C5aR1 is a functional receptor in DCs.

Next, we tested the ability of C5a to affect antigen presentation or cytokine production by DCs. We did not observe any significant change in antigen uptake capacity or cytokine production in LLC tumor–infiltrating DCs treated with C5a (Supplementary Figs. S10A and S10B), suggesting that C5aR1 in cancer-associated DCs does not modulate their antigen-presenting activity.

Our differential expression analysis identified “myeloid leukocyte migration” and “cell chemotaxis” among the top differentially expressed processes between human C5AR1⁻ and C5AR1⁺ DCs (Fig. 1D), and genes related to these processes were generally upregulated in human C5AR1⁺ DCs (Supplementary Fig. S10C). To test the function of C5a in the migration of tumor-infiltrating DCs expressing C5aR1, we treated LLC tumor–infiltrating DCs with C5a and evaluated the migratory capacity of each DC subtype. C5aR1 engagement enhanced the migration of cDC2s and moDCs but not of cDC1s or pDCs (Fig. 4C). The CCR7 ligand CCL19, a known chemoattractant for DCs, was able to induce the mobilization of all DC subsets (Supplementary Fig. S10D). The combination of CCL19 with C5a resulted in an additive effect on migration, suggesting that these two chemoattractants acted independently (Supplementary Fig. S10E).

We next postulated that C5a may regulate the trafficking of DCs between tumors and secondary lymphoid organs, a key step in the generation of an antitumor immune response. In support of this idea, we found that tumor-draining LNs had significantly higher levels of C5a than LLC tumors (Fig. 4D). To test whether C5aR1-expressing DCs are chemoattracted to LNs by C5a, we treated LLC tumor–bearing mice with the C5aR1 inhibitor PMX53. As previously reported (13), C5aR1 inhibition had a modest effect on primary tumor growth (Supplementary Fig. S10F). Fifteen days after tumor implantation, we analyzed the proportion of DC subsets in tumors and LNs by flow cytometry, differentiating migratory and resident DCs by expression of MHC-II and CCR7 (Supplementary Fig. S10G; ref. 39). In primary tumors, no significant changes in the proportion of DC subsets were observed after C5aR1 inhibition (Fig. 4E). However, in LNs, C5aR1 blockade resulted in a significant reduction in the proportion of migratory cDC2s, with a consequent increase in the proportion of migratory cDC1s. No migratory moDCs or pDCs were detected in this experimental setting. In resident DCs, only a modest decrease in moDCs was observed (Fig. 4F).

To track the trafficking of tumor-infiltrating DCs to LNs, we used a genetically engineered mouse model based on the knock-in of the photoconvertible fluorescent protein KikGR (Fig. 4G; refs. 40, 41). C5aR1 blockade did not affect the number of cDC1s, moDCs, or pDCs in tumor-draining LNs but significantly reduced the presence of red KikGR cDC2s (Fig. 4H). Collectively, these results demonstrate the ability of C5a to induce the migration of tumor-infiltrating cDC2s from primary tumors to tumor-draining LNs.

C5a-mediated trafficking of tolerogenic cDC2s from the TME to tumor-draining LNs may interfere with the efficacy of immunotherapies that rely on the ability of cDC1s to prime T cells in these secondary lymphoid organs. To evaluate this possibility, we investigated whether C5aR1 inhibition could enhance the antitumor activity of poly I:C, an agonist of TLR3 capable of inducing the activity of cDC1s (Fig. 5A; ref. 42). We evaluated the combination of poly I:C and C5aR1 inhibition with PMX53 in three mouse tumor models with low (LLC), medium (393P), and high (B16.F10) sensitivity to poly I:C. C5aR1 blockade with PMX53 did not improve the therapeutic activity of poly I:C in any of these models (Supplementary Fig. S11). Considering the high expression of PD-L1 in the tolerogenic C5aR1-expressing DCs (Fig. 3E), we added an anti–PD-1 drug to the treatment combination. In this scenario, the C5aR1 inhibitor increased the efficacy of poly I:C/anti–PD-1 treatment in the three models (Fig. 5B). In the LLC model, the poly I:C/anti–PD-1 combination showed a therapeutic effect only in the presence of C5aR1 blockade. In the 393P model, complete responses were observed in six of eight mice in the poly I:C/anti–PD-1 group, and in all mice in the poly I:C/anti–PD-1/C5aR1-Inh group. One mouse treated with the triple combination was cured in the B16.F10 model, and statistical differences between the double and the triple combination were significant at the end of the experiment (day 20; P = 0.017). Cured mice from the different models were resistant to tumor rechallenge (Supplementary Fig. S12A), suggesting the development of an efficient antitumor long-term memory response. We also performed an experiment in which C5aR1-expressing moBMDCs were inoculated into the LLC tumors during treatment. Inoculated DCs were able to partially reverse the therapeutic effect of C5aR1 blockade (Supplementary Fig. S12B).

To evaluate the contribution of cDC1s to the antitumor activity of the combination treatment, we used an Xcr1-DTR-Venus mouse model, which allows selective depletion of cDC1s upon administration of diphtheria toxin (Supplementary Fig. S12C). cDC1 depletion abolished the antitumor activity of the triple combination treatment (poly I:C, anti–PD-1, and C5aR1 blockade) in LLC tumor–bearing mice (Fig. 5C). CD8⁺ T cells, CD4⁺ T cells, and NK cells were also selectively depleted in LLC tumor–bearing mice using specific monoclonal antibodies (Supplementary Fig. S12D). CD8⁺ depletion, but not CD4⁺ or NK depletion, abolished the antitumor activity of the treatment (Fig. 5D). These data demonstrate that both cDC1s and CD8⁺ T cells are required for the antitumor activity of the poly I:C/anti–PD-1/C5aR1-Inh combination treatment. The efficacy of the triple combination also was significantly reduced when mice were cotreated with FTY720 (Fig. 5E), a lipid mediator that reduces lymphocyte egress from LNs (Supplementary Fig. S12E), supporting the necessity of lymphocyte trafficking for the antitumor effect of the triple combination treatment.

Next, we analyzed the effects of the triple combination treatment on the immune infiltrate of the TME. This analysis was performed by flow cytometry on day 16 after LLC implantation, when the differences in tumor growth became apparent (Supplementary Fig. S13A). The triple combination treatment increased the number of PMN-MDSCs, monocytic MDSCs (M-MDSCs), and macrophages in the TME, although in the case of DCs, differences did not reach statistical significance (Fig. 6A). Among subpopulations, M1-like macrophages and moDCs were significantly increased, whereas a near-significant increase was observed for cDC1s (Supplementary Fig. S13B). PD-L1 expression increased in all cell populations of the myeloid compartment, except for DCs, where PD-L1 upregulation was observed only in cDC2s and moDCs (Supplementary Fig. S13C and S13D). CD8⁺ T cells and the CD8⁺/Treg ratio were significantly increased in tumors treated with the triple combination (Fig. 6B). By multiplex immunofluorescence, we observed that the triple combination treatment was associated with the accumulation of CD11b⁺ cells in the tumor periphery, whereas CD8⁺ and Tregs infiltrated the tumors (Fig. 6C). No changes in other immune subpopulations (namely CD4⁺ T cells, Tregs, B cells, NK cells, or M2-like macrophages) were detected in the TME of tumors treated with the triple combination (Supplementary Fig. S13E).

A significant increase of tumor-specific CD8⁺ T cells was found at day 15 in OVA-expressing LLC tumors of mice treated with the triple combination [Fig. 6D (left)]. In contrast, in tumor-draining LNs, an increase in the number of OVA-specific CD8⁺ T cells was detected in the double combination treatment but not in the triple combination treatment [Fig. 6D (right)], suggesting that the triple combination treatment may enhance the egress of CD8⁺ T cells from tumor-draining LNs. Accordingly, inhibition of lymphocyte egress with FTY720 resulted in an increase in the number of OVA-specific CD8⁺ T cells in the tumor-draining LNs of triple combination-treated mice (Fig. 6E). Finally, no changes in total DCs in tumor-draining LNs from LLC tumor–bearing mice were observed, but there was a significant reduction in the proportion of cDC2s, accompanied by an increase in the proportion of cDC1s (Fig. 6F). Taken together, these analyses demonstrate that C5aR1 blockade enhances the antitumor activity of poly I:C/anti–PD-1 combination treatment. This effect may be regulated by the ability of C5aR1 blockade to both reduce the recruitment of cDC2s and facilitate the egress of CD8⁺ T cells in tumor-draining LNs.

We analyzed the transcriptional changes induced by C5aR1 blockade in tumors treated with poly I:C and anti–PD-1. RNA-seq analysis was performed on day 17 after subcutaneous implantation of LLC cells (Supplementary Fig. S14A). A differential expression analysis revealed a distinct transcriptomic profile of tumors treated with vehicle, poly I:C/anti–PD-1, or poly I:C/anti–PD-1/C5aR1-Inh (Supplementary Fig. S14B and S14C). Gene Ontology: Biological Process (GO:BP) gene sets analysis revealed several enriched metabolic pathways in tumors treated with poly I:C/anti–PD-1 (Supplementary Fig. S14D), whereas processes related to immune responses were highly enriched in tumors treated with the triple combination (Supplementary Fig. S14E). The tumors treated with the triple combination showed enrichment in pathways related to myeloid and lymphoid activation, whereas there was a reduction in the expression of genes associated with protein metabolism (Fig. 7A). Pathways related to DC functions were enriched, such as “phagocytosis,” “positive regulation of phagocytosis,” and “antigen processing and presentation” (adjusted P values: 1.4 × 10⁻¹⁵, 7.9 × 10⁻¹⁰, and 2.9 × 10⁻¹⁰, respectively; Fig. 7A and B). A more detailed analysis of a set of genes involved in antigen processing and presentation (43) revealed a pronounced increase in the expression of most of the genes associated with this process upon C5aR1 inhibition (Fig. 7C).

The mRNA expression of a number of cytokines and immunomodulators, including those previously identified as upregulated in human tumor-infiltrating C5aR1⁺ DCs (Fig. 1F), was increased in tumors treated with the poly I:C/anti–PD-1 double combination compared with vehicle (Fig. 7D). When the C5aR1 inhibitor was added to the treatment, some cytokines remained unchanged, whereas others increased or decreased. Immunosuppressive cytokines or immunomodulators, such as IL6, CXCL1 (the functional homolog of human IL8), or PGE2 (Ptges2), were decreased after C5aR1 blockade. In the case of IL10, C5aR1 blockade seemed to increase its levels (Fig. 7D). The expression of IL10 was evaluated specifically in DCs using IL10-GFP-reporter mice (Vert-X), in which IL10-expressing cells can be quantified as GFP⁺ cells (44, 45). IL10 expression was reduced in cDC2s upon C5aR1 blockade but not in the other DC subtypes (Fig. 7E; Supplementary Fig. S15).

Poly I:C activity is characterized by the induction of type I IFNs and the expression of IFN-stimulated genes (46). Accordingly, the GO processes “type I IFN production,” “response to IFNα,” and “response to IFNβ” were significantly enriched in tumors treated with poly I:C/anti–PD-1 versus vehicle (adjusted P values: 1.3 × 10⁻⁴, 2.0 × 10⁻³, and 1.1 × 10⁻², respectively). Moreover, the triple combination treatment increased the mRNA levels of IFNγ (Fig. 7D), as well as of IL12 and IL18, two cytokines that induce the expression of IFNγ and stimulate antitumor immune responses (47, 48). In addition, GO processes related to IFNγ were enriched in tumors treated with the triple combination compared with the double treatment (adjusted P values for “IFNγ production” and “response to IFNγ”: 2.5 × 10⁻⁶ and 7.4 × 10⁻⁵, respectively). The expression of IFNα and IFNβ response genes increased significantly in poly I:C/anti–PD-1 treated tumors but did not increase further after the addition of the C5aR1 inhibitor (Fig. 7F). In contrast, C5aR1 blockade resulted in a significant increase in IFNγ response (Fig. 7F). To assess the changes in cytotoxic activity of the TME, a cytotoxicity score was calculated using the mRNA expression of perforin (Prf1), granzymes (Gzma, Gzmb, Gzmk, and Gzmm), and NK-cell granule protein 7 (Nkg7). Poly I:C/anti–PD-1/C5aR1-Inh combination treatment significantly increased the cytotoxic properties of the TME (Fig. 7G). In conclusion, our study suggests that treatment of tumors with C5aR1 inhibition, poly I:C, and anti–PD-1 induces a switch in DC subsets that results in higher antigen-presenting capacity, upregulation of IFNγ signaling, and downregulation of immunosuppressive cytokines, leading to enhanced cytotoxic activity (Fig. 7H).

Activation of C5aR1 in myeloid cells is recognized as one of the major mechanisms by which complement exerts its influence on cancer immunity (10, 11). However, the influence of C5a in cancer-associated DCs remained unknown. Here, we describe for what we believe to be the first time an upregulation of C5aR1 in tumor-infiltrating tolerogenic DCs, which may mediate the trafficking of these cells from primary tumors to LNs, thereby inhibiting DC-mediated T-cell priming and activation [Fig. 7H (left)]. We also propose that targeting the C5a/C5aR1 axis enhances the antitumor efficacy of DC-driven therapies [Fig. 7H (right)].

Based on transcriptomic profiling of both mouse and human DCs from lungs and LNs of healthy donors, the expression of C5aR1 has been proposed as a marker to distinguish monocytes/macrophages from DCs because of the low expression of this receptor in the latter (49). We now report the upregulation of C5aR1 expression in tumor-infiltrating DCs by cytokines commonly found in the TME.

The role of C5aR1-expressing DCs has been described in healthy mice and in models of inflammation and immune-related diseases. C5aR1 engagement in both mouse and human DCs derived from spleen or bone marrow increases antigen presentation, expression of costimulatory molecules, viability of naïve T cells, and induction of Th1 polarization (50, 51). However, environmental cues influence DC maturation, distribution, and function. C5a can inhibit the production of proinflammatory cytokines induced by various TLR ligands (52). The effect of C5aR1 activation on DC biology is also modified by pathologic conditions such as cholangitis (53) or allergic asthma (54). We found that IL6 and IL10, two widely recognized negative regulators of antitumor DC functions (4), induce the expression of C5aR1 in DCs. This observation is consistent with a study in which BMDCs treated with conditioned medium from LLC cells showed upregulation of complement molecules, including C5aR1 (55). Our study also reveals the role of C5aR1 in the migration of tolerogenic C5aR1-expressing DCs.

Migration of tumor-infiltrating DCs to LNs is a critical step in the generation of effective antitumor immunity. Studies on C5a-mediated migration of DCs are limited, and none have been performed in the tumor context. In our in silico study of lung adenocarcinoma, we found an upregulation of the expression of migration-related genes in tumor-infiltrating C5aR1-expressing DCs. Furthermore, we show that blockade of the C5a/C5aR1 axis limits the migration of mouse cDC2s to LNs. In our experimental setting, we did not detect migratory moDCs in LNs, so the effect of C5aR1 inhibition on the migration of C5aR1-expressing DCs could only be observed in cDC2s. Accumulation of C5aR1-expressing DCs in LNs may impair antitumor T-cell responses, as tolerogenic DCs can mediate Treg accumulation and proliferation (56). An interesting observation was that despite the reduction of migrating cDC2s in the LNs after C5aR1 blockade, we did not observe an accumulation of cDC2s in the tumors. At this point, we can only speculate about the reasons for this. The number of migrating cDC2s may not be high enough to significantly affect the fraction of cDC2s in the tumors. Alternatively, C5a blockade may have additional uncharacterized effects on intratumoral cDC2s, influencing their differentiation or even their migration to anatomic sites other than LNs. In particular, it would be interesting to investigate whether C5a is also involved in the recruitment of these tolerogenic DCs to distant metastatic sites, thereby contributing to the dissemination of cancer cells.

Our research supports the concept that blocking C5aR1 would enhance the antitumor efficacy of DC-dependent cancer immunotherapies. In the LLC model, poly I:C and PD-1/PD-L1 blockade induced the secretion of proinflammatory cytokines and type I interferons (i.e., IFNα and IFNβ) but failed to induce the secretion of IFNγ, a major driver of antitumor cytotoxic T-cell responses. IFNγ induction was achieved when C5aR1 blockade was added to the treatment. Accordingly, the triple combination treatment resulted in significant antitumor activity associated with increased antigen presentation and cytotoxic capacity, as well as the accumulation of cross-presenting cDC1s and CD8⁺ T cells in the TME. Nevertheless, the distinct antitumor effect of C5aR1-expressing DCs after C5aR1 inhibition cannot be distinguished from the effect mediated by other C5aR1-expressing cells present in the TME, such as macrophages or MDSCs. Further studies are warranted to determine the specific role of C5aR1 blockade on C5aR1-expressing DCs in vivo.

In conclusion, this study provides insights into the involvement of complement in the biology of DCs in the context of cancer. Based on these findings, we propose an immunotherapy combination aimed at enhancing the efficacy of DC-based therapies by targeting the tolerogenic mechanisms exerted by the C5/C5aR1 axis. Clinical application of this strategy could substantially enhance anticancer immune responses to immune checkpoint inhibitors in solid tumors.

T.M. Woodruff reports grants from the National Health and Medical Research Council of Australia during the conduct of the study. No disclosures were reported by the other authors.

Y. Senent: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. A. Remírez: Methodology. D. Repáraz: Methodology, writing–review and editing. D. Llopiz: Methodology, writing–review and editing. D.P. Celias: Methodology. C. Sainz: Methodology. R. Entrialgo-Cadierno: Methodology. L. Suarez: Methodology, writing–review and editing. A. Rouzaut: Methodology, writing–review and editing. D. Alignani: Methodology, writing–review and editing. B. Tavira: Investigation, methodology, writing–review and editing. J.D. Lambris: Investigation, methodology, writing–review and editing. T.M. Woodruff: Methodology, writing–review and editing. C.E. de Andrea: Methodology, writing–review and editing. B. Ruffell: Methodology, writing–review and editing. P. Sarobe: Investigation, methodology, writing–review and editing. D. Ajona: Conceptualization, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. R. Pio: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

We thank Drs. Laura Guembe (Morphology platform), Elisabet Guruzeaga (Bioinformatics platform), Ibon Tamayo (Bioinformatics platform), Sandra Hervás (Program in Immunology and Immunotherapy), and Sara Labiano (Program in Solid Tumors), all at Cima Universidad de Navarra, for their help and advice. This work was supported by Foundation for Applied Medical Research, Centro de Investigación Biomédica en Red Cáncer, Fundación Científica de la Asociación Española Contra el Cáncer (IDEAS211016AJON), Departamento de Salud-Gobierno de Navarra cofunded at 50% by Fondo Europeo de Desarrollo Regional 2014-2020 (51-2021), Fundación Ramón Areces, Fondo de Investigación Sanitaria-Fondo Europeo de Desarrollo Regional “Una manera de hacer Europa” (PI20/00419; PI23/00573; PI20/00260), and the National Health and Medical Research Council of Australia (2009957). Y. Senent was funded by a predoctoral fellowship from the Spanish Ministry of Science and Innovation (FPU18/02638) and two mobility fellowships from the University of Navarra and government of Navarra. R. Entrialgo-Cadierno was funded by a donation from Maria Eugenia Burgos de la Iglesia’s family.