Targeting the dendritic cell-secreted immunoregulatory cytokine CCL22 alleviates radioresistance

Mice, Cell Lines, and Reagents

Six- to eight-week old C57BL/6J WT mice were purchased from Envigo. Zbtb46^cre (B6.Cg-Zbtb46^{tm3.1(cre)Mnz}/J, RRID:IMSR_JAX:028538), Rosa^YFP (B6.129X1-Gt(ROSA)26Sor^tm1(EYFP)Cos/J, RRID:IMSR_JAX:006148), and OT-I CD8⁺ T cell receptor (TCR)-Tg (C57BL/6-Tg(TcraTcrb)1100Mjb/J, RRID:IMSR_JAX:003831) mice were purchased from Jackson Labs. Batf3^−/− mice were graciously donated from the Kline Lab at the University of Chicago. Mice were maintained under specific pathogen-free conditions and utilized according to a protocol approved by the Institutional Animal Care and Use Committee of the University of Chicago. The MC38 (RRID:CVCL_B288) tumor cell line was provided by Dr. Yang-xin Fu. B16F10 (RRID:CVCL_0159) and B16F1 (RRID:CVCL_0158) cell lines were acquired from American Type Culture Collection (ATCC). All cell lines were authenticated and tested free of murine pathogens. B16-OVA, B16-EGFR, B16-EGFR-OVA and MC38-EGFR were previously described. (48) Cell lines were authenticated and tested for mouse pathogens by IDEXX (Westbrook, Maine). All cells were cultured in 5% CO₂ and maintained in vitro in Dulbecco’s modified Eagle’s medium supplemented with 10% heat-inactivated fetal bovine serum (Sigma-Aldrich), 100 U/mL penicillin, 100 μg/mL streptomycin and 0.1 mmol/L Minimum Essential Medium nonessential amino acids. αCCL22 (MDC monoclonal antibody) was purchased from Invitrogen (158132) (MA5-23780, RRID:AB_2609986). αCD8 was purchased from BioXcell (BE0004-1, RRID:AB_1107671). The SiteClick Antibody Labeling Kit (R-PE, S10467) was purchased from ThermoFisher. Anti-mouse CTLA-4 (CD152) clone 9H10 was purchased from BioXcell (BE0131, RRID:AB_10950184).

Pacific blue-anti-CD45 (103126, RRID:AB_493535), pacific blue-anti-CD11b (101224, RRID:AB_755986), Pacific-blue-anti-CD80 (104724, RRID:AB_2075999), BV421-anti-I-A/I-E (107631, RRID:AB_10900075), BV510-anti-CD19 (115545, RRID:AB_2562136), BV510-anti-CD86 (105039, RRID:AB_2562370), BV570-anti-CD4 (100541, RRID:AB_10897943), BV570-anti-NK1.1 (108733, RRID:AB_10896952), BV605-anti-CCR7 (120125, RRID:AB_2715777), BV605-anti-CD3 (100351, RRID:AB_2565842), BV605-anti-CD11c (117333, RRID:AB_11204262), BV650-anti-XCR1 (148220, RRID:AB_2566410), BV711-anti-CD64 (139311, RRID:AB_2563846), BV750-anti-CD3 (100249, RRID:AB_2734148), BV785-anti-Ly6C (128041, RRID:AB_2565852), fluorescein isothiocyanate (FITC)-anti-CD45 (103108, RRID:AB_312973), FITC-anti-CD25 (101907, RRID:AB_961210), FITC-anti-Ly6C (128006, RRID:AB_1186135), Alexa Fluor 488-anti-CD8a (100723, RRID:AB_389304), AF488-anti-FceRIa (MAR-1, 134330, RRID:AB_2687239), PerCP-Cy5.5-anti-Bcl6 (358508, RRID:AB_2566191), PerCP/Cy5.5-Ly6G (127616, RRID:AB_1877271), PerCP/Cy5.5-anti-F4/80 (123128, RRID:AB_893484), PE-anti-H2Kb-SIINFEKL (141603, RRID:AB_10897938), PE-anti-CD11b (101208, RRID:AB_312791), PE-anti-XCR1 (148203, RRID:AB_2563842), PE-Dazzle594-anti-CD103 (121430, RRID:AB_2566493), PE/Cy5-anti-CD86 (105016, RRID:AB_493602), PE/Cy5-anti-CD11c (117316, RRID:AB_493566), PE/Cy7-anti-CD11c (117318, RRID:AB_493568), PE/Cy7-anti-Ly6C (128017, RRID:AB_1732093), PE/Fire810-anti-MHCII (107667, RRID:AB_2894690), APC-anti-CD11c (117310, RRID:AB_313779), APC-anti-PD-L1 (124312, RRID:AB_10612741), APC-anti-CD4 (100412, RRID:AB_312697), AF700-anti-CD206 (141734, RRID:AB_2629637), AF700-anti-CD11b (101222, RRID:AB_493705), APC/Cy7-anti-CD172a (144017, RRID:AB_2629557), BV510-anti-CD24 (101831, RRID:AB_2563894), and Zombie NIR Fixable dye (423105) were purchased from BioLegend. BUV395-anti-Ly6G (563978, RRID:AB_2716852), BUV496-anti-CD24 (612953, RRID:AB_2870229), BUV563-anti-NK1.1 (741233, RRID:AB_2870785), BUV737-anti-CD8a (564297, RRID:AB_2722580), BUV805-anti-F4/80 (749282, RRID:AB_2873657), BV421-anti-CD40 (562846, RRID:AB_2734767), BV421-anti-FcεR1α (MAR-1, 750852, RRID:AB_2874960) and BV605-anti-I-A/I-E (MHCII, 563413, RRID:AB_2738190) were purchased from BD Biosciences. PerCP-eFluor710-anti-PD-1H (46-5919-82, RRID:AB_2573811), PE-anti-FOXP3 (12-5773-82, RRID:AB_465936), PE-anti-H2Kb-SIINFEKL (12-5743-82, RRID:AB_925774) were purchased from eBioscience. AF532-anti-CD45 (58-0451-82, RRID:AB_11218871), PerCP-eFluor710-anti-B220 (46-0452-82, RRID:AB_10717389), PE/Cy7-anti-Axl (25-1084-80, RRID:AB_2734851) were purchased from Invitrogen. eBioscience^™ Foxp3/Transcription Factor Staining Buffer Set was purchased from Invitrogen (00-5523-00). SIINFEKL was purchased from Invivogen (vac-sin). Mouse MDC (CCL22) ELISA kit was purchased from Abcam (ab204525). EasySep^™ Mouse CD11c Positive Selection Kit II (18780), Mouse Pan-DC Enrichment Kit II (19863), Mouse CD8⁺ T Cell Isolation Kit (19853), and Mouse CD8a Positive Selection Kit II (18953) were purchased from StemCell Technologies. Mouse IFNγ ELISPOT Set was purchased from BD Biosciences (551083). All primers were purchased from IDT.

Production of the αEGFR-IFNα Fusion Protein

The heterodimeric Fc variant KiHss-AkKh platform was described previously. (49) The Fab fragment of Cetuximab (anti-human EGFR) was fused with the knob variant Fc region in the pEE12.4 vector (Lonza), and the murine IFNa4 was fused with the hold variant Fc region via a GGG linker in the pEE6.4 vector (Lonza). The anti-EGFR-IFNa4 heterodimer was generated by transient co-transfection of two plasmids into FreeStyle293-F cells. The supernatant containing the fusion protein was purified using CaptivA^® Protein A Affinity Resin according to the manufacturer’s manual (Repligen). Heterogeneity was confirmed by SDS-PAGE and purity evaluated by SEC-HPLC.

Tumor Growth and Treatments

Tumors were established in mice by subcutaneous injection of 1 × 10⁶ tumor cells in 0.1 mL PBS into the hind flanks. Tumors were irradiated 10–11 d after inoculation with 20 Gy while the rest of the body was shielded by lead. For tumor rechallenge, cured or naïve mice were rechallenged with 1 × 10⁶ tumor cells. Tumor growth was monitored by measuring the length, width, and depth and volume calculated as one half the length x width x depth. For αCTLA4 treatment, 200 μg/0.1 mL PBS was given intraperitoneal (i.p.) on day 10 with RT and day 16. αCCL22 (4 μg/mL_tumor) was given every other day starting day 11 for 2 weeks. αEGFR-IFNα was injected intratumoral on day 10 (10–20 μg/0.02 mL PBS) and day 14 (5–10 μg/0.02 mL PBS). αCD8 200 μg/0.1 mL PBS was administered by intraperitoneal injection on the same day as RT and again 3 days later.

RNA Sequencing and Bioinformatic Analysis

Samples for scRNA-seq of CD45⁺ tumor immune cells were obtained from pooled MC38 tumors in WT mice with or without treatment with 15 Gy as described previously. (16) Samples were stained using ZombieRed^™ dye for 30 min and surface stained for mouse CD45 for 20 min. ZombieRed⁻ CD45⁺ single cells were sorted and used for library construction. Library construction was performed using Chromium Next GEM Single Cell 3’ GEM, Library & Gel Bead Kit v3.1 (Cat: 1000128) purchased from 10X Genomics according to the manufacturer’s instructions. Target cell recovery for each library was 8.000 and sequencing was performed on an Illumina HiSeq X Ten platform.

10X Genomics Cell Ranger (v6.0.1, RRID:SCR_023221) was used to process raw scRNA-seq data into FASTQ files with “cellranger mkfastq” function, align the reads to the mouse reference genome (mm10, GENCODE vM23/Ensembl 98 released on July 7, 2020 from 10X Genomics) and count the unique molecular identifier (UMI) with “cellranger count” function. Low-quality cells were discarded if (1) the number of expressed genes was smaller than 200; (2) the proportion of mitochondrial gene expression were larger than 25%. We further identified and removed potential doublets by using DoubletFinder (v2.0.3, RRID:SCR_018771) assuming 6% doublet formation rate. The processed whole gene expression matrix was then fed to Seurat (v4.0.6, RRID:SCR_016341) for downstream analyses.

Briefly, the UMI count matrix from untreated and RT samples were integrated by Seurat::IntegrateData method after normalized with ‘SCT’ method using 3,000 highly variable feature genes. Clustering analysis was performed using the first 40 principal components for constructing the shared nearest neighbor (SNN) graph with resolution as 0.7. Later, marker genes were called by Seurat::FindAllMarkers with log2 fold change >= 1. Identified clusters were annotated by scClassify (v1.2.0) (50) based on cell types hierarchies constructed from reference datasets (E-MTAB-8832, CD45+ immune cells sorted from MC38 tumor bearing C57BL/6 mice) (51) and labeled using the selected marker gene. To validate the annotation of the DC cells, we scored each cell in our dataset using the Seurat::AddModuleScore function using expression of cDC feature genes (H2-Ab1, Flt3 and Itgax) as described in Duong, et al. (35) Dendritic cells in the two clusters were extracted and further clustered into sub clusters and annotated using the same clustering and annotating method. The dendritic cell subtype score was calculated as the fraction of RNA in a cell belonging to genes in the gene list for the mregDC, DC1 and DC2 feature gene list described in Maier, et al. (36) The scRNA-seq dataset has been deposited in the Gene Expression Omnibus (GEO) under the accession number GSE206387.

For bulk RNA-seq Live CD45⁺ CD11c⁺ cells were flow sorted from tumors and mixed with Trizol for RNA extraction. The RNA library for sequencing was constructed using SMARTer^® Stranded Total RNA-Seq Kit v2 - Pico Input Mammalian (Takara Bio) and sequencing performed at the University of Chicago Genomics Facility on an Illumina NovaSEQ machine in pair-read mode with 100 bp per read. Raw sequencing and processed data is deposited within GEO under the accession number GSE262864. RNA sequencing analysis was performed using the Galaxy platform (RRID:SCR_006281). Raw data quality control was done using FastQC, and all samples were included in the analysis. In brief, trimming was performed using Trimmomatic. Sequenced reads were aligned to the mouse reference genome (mm10) using RNA STAR (RRID:SCR_004463) and quantified by featureCounts (RRID:SCR_012919). Count data was further analyzed for differential expression using DESeq2 (RRID:SCR_015687). Gene ontology enrichment analysis was performed using ShinyGo (RRID:SCR_019213) on genes with increased expression, a Log₂ fold change > 2, and p value < 0.05.

Flow Cytometry

Single cell suspensions were prepared as previously described. (52) Single cells from tumors were harvested by digestion using 1 mg/mL collagenase IV (Sigma-Aldrich) and 200 μg/mL DNaseI (Sigma-Aldrich) at 37°C for 30–60 minutes. Blocking was accomplished with anti-Fc receptor (2.4G2, Leinco C381, RRID:AB_2737484) prior to cell staining. Cells were stained according to manufacturer recommendations. Surface and live/dead staining was performed in PBS for 45 minutes followed by 2 consecutive washes and fixation according to the eBioscience Foxp3/Transcription Factor Staining Buffer Set (00-5523-00). FoxP3 staining was performed in permeabilization buffer for 45 minutes at room temperature followed by 2 consecutive washes according to the FoxP3 staining protocol (ThermoFisher, 00-5523-00). CCL22-PE was prepared using the SiteClick Antibody Labeling Kit for PE (ThermoFisher) and αCCL22 according to the manufacturer instructions. CCL22-PE staining was performed using the ThermoFisher protocol for cytoplasmic proteins. Cell suspensions were analyzed on a Fortessa (BD Biosciences), Attune (Invitrogen), or Aurora (Cytek) flow cytometer. Cell sorting was performed using a AriaIII and BC FACSAria Fusion (BD Biosciences). Data was analyzed with OMIQ software from Dotmatics.

ELISPOT Assay

For CD8⁺ T cell functional assays, CD8⁺T cells were isolated from tumors using an EasySep^™ Mouse CD8a Positive Selection Kit (STEMCELL Technologies) 7 d after treatment initiation. 2–4 × 10⁵ CD8⁺ T cells were stimulated with 1 μg/mL SIINFEKL for 48 h. ELISPOT assays were performed to detect cytokine spots of IFNγ according to the product protocol (Millipore). In brief, a 96 well HTS-IP plate was precoated with an anti-IFNγ antibody at a 1:250 dilution and incubated overnight at 4°C. Cells were removed and biotinylated anti-IFNγ antibody (2 mg/mL) at a 1:250 fold dilution was added and incubated for 1–2 h at room temperature. Avidin-horseradish peroxidase in a 1:1000 dilution was added to the plate and incubated for 1 h at room temperature. Spots were developed according to the manufacturer’s instructions (BD) and calculated using an ELISPOT plate reader.

Ex Vivo Dendritic Cell Culture

Spleens were isolated and single cell suspensions prepared from 8–12 week old Zbtb46^creRosa^YFP mice as described for flow cytometry. cDC were enriched using an pan-DC enrichment kit according to the manufacturer’s instructions (STEMCELL Technologies). Enriched cell suspensions were stained for αCD24 and αCD172a and immediately sorted based on YFP and stains as shown in the representative gating strategy. B16 tumor cells were either untreated or treated with 40 Gy radiation the day prior to plating. cDC were plated at a 1:1 ratio with tumor cells in 96-well cell culture plates in 150 μL RPMI 1640 (Gibco) media supplemented with 10% FBS, penicillin and streptomycin. Cells collected by trypsinization, incubated with eBioscience Protein Transport Inhibitor Cocktail (00-4980-03) for 6 hours in supplemented RPMI at 37°C and stained for CD45 and CCL22 according to the ThermoFisher protocol for cytoplasmic protein staining. PE staining in CD45⁺ YFP⁺ cells was quantified and analyzed using OMIQ.

Statistical Analysis

For tumor growth measurements, flow cytometry, gene expression, Kaplan-Meier analysis, and protein measurements descriptive statistics were analyzed using Prism software (version 7.0, GraphPad). Kaplan-Meier survival analysis was compared using the log-rank (Mantel-Cox) test. All other data was tested using either two-way ANOVA, Student’s t-test, or one-way ANOVA as described in the corresponding figures. R was used for analysis of public datasets. The human homolog of DC_Ccl22 was generated using the Mouse Genome Informatics Database from The Jackson Laboratory. The resulting gene list was: CACNB3, CCR7, FSCN1, TBC1D4, TMEM123, SERPINB6, CCL5, CD63, SAMSN1, RAMP3, MMP25, MARCKS, FABP5, FABP5P3, CCL22, STAT4, SERPINB9, ROGDI, SOCS2, MARCKSL1, CD200, CALM1, ISCU, RELB, BCL2A1, BASP1, TRAF1, MAP4K4, IL4I1, GYG1, PCGF5, PDLIM4, GADD45B, ANXA3, BIRC2, ARL5C, KTN1, RPS27L, GLIPR2, NET1, UAP1, ID2, ZMYND15, TXNDC17, GNB4, EPSTI1. Data from the Farren, et al. dataset was used from the Gene Expression Omnibus dataset under accession GSE129492. (53) TCGA data was acquired and analyzed in part using the Xena Platform. (54) For analysis of gene signatures in survival and correlation each cohort gene expression was divided into tertiles. Samples in the lowest and highest tertiles received a score of −1 and 1, respectively. Scores were added for each sample and the lowest quartile and highest tertile isolated at ‘Low’ and ‘High’ gene expression signature, respectively. Survival data from these two groups were then compared. Data analysis of TCGA and public datasets was performed using R. Sample sizes were determined based on previous reports and all animals were randomly divided into treatment groups based on tumor volume. The number of biological replicates for all bar graphs are represented by individual data points. All tumor growth/survival and flow cytometry experiments are representative of at least 2 independent experiments. For all tumor growth and survival studies a minimum of 5 mice per group were used. All experiments were unblind. All data presented as mean ± standard error unless otherwise noted. n.s. (not significant), *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001.

Study Approval

Mice were maintained under specific pathogen-free conditions and utilized according to protocol approved by the Institutional Animal Care and Use Committee of the University of Chicago.

Data and materials availability:

The scRNA-seq dataset has been deposited in the Gene Expression Omnibus (GEO) under the accession number GSE206387. Raw bulk RNA sequencing and processed data is deposited within GEO under the accession number GSE262864. All other data is available at request from R.R.W.

Local irradiation expands tumor mregDC

We previously performed single-cell RNA sequencing (scRNA-seq) on CD45⁺ cells isolated from murine MC38 colorectal adenocarcinoma subcutaneous tumors 4 days after treatment with 15 Gy to explore changes in suppressive myeloid populations (Supplementary Figure S1A). (16) Doses of 15–20 Gy were utilized since they have demonstrated CD8⁺ T cell-dependent antitumor immunity and are included in various stereotactic ablative regimens from clinical studies. (55–58) A timepoint of 4 days was chosen as we observe significant tumor myeloid cell infiltration and cDC activation 3–5 days after ablative regimens. (8, 15, 16, 59, 60) Twenty-four populations were identified using unbiased clustering of all CD45⁺ tumor infiltrates based on canonical immune signatures. Both unbiased assignment and mapping a cDC signature score based on expression of H2-Ab1, Flt3 and Itgax (35) revealed 2 cDC clusters annotated by signature gene expression (DC_Ccl22 and DC_Atox1, Supplementary Figure S1A–B). DC_Atox1 exhibited characteristic expression of both cDC1 and cDC2 markers. Instead, DC_Ccl22 exhibited an upregulation of markers of maturation, migration and regulation, suggesting an activated mregDC phenotype. Scoring for cDC1, cDC2 and mregDC as described in Maier, et al. revealed an enriched mregDC score in the DC_Ccl22 cluster (Supplementary Figure S1C). To examine changes in cDC lineage-based populations, we isolated cDC and identified 3 canonical populations derived from DC_Atox1: DC_Rab7b (cDC1), DC_Ly6a (cDC2), and DC_Cd7 (plasmacytoid dendritic cells, pDC), as well as a distinct separate population of mregDC (DC_Ccl22) (Figure 1A–C).

The DC_Ccl22 cluster shared a high similarity with mregDC reported in Maier, et al. (36), including an upregulation of genes involved in immune regulation (Cd274, Socs2, Aldh1a2, Cd200, and Ccl22), a type II helper T cell response (Il4i1, Ccl22), cDC maturation (Relb, Cd80, and Cd86), and cell migration (Ccr7, Myo1G, Fscn1, and Marcks) (Figure 1A, C and Supplementary Figure S1D). Among cDC clusters, DC_Ccl22 exhibited the greatest proportionate increase after RT and upregulation of immune suppression markers (Cd274, Cd200 and Socs2), suggesting that they may contribute to RT-driven immune suppression (Figure 1D and Supplementary Figure S1D). Moreover, among the top genes differentially expressed in DC_Ccl22, the known Treg-recruiting chemokine CCL22 was exclusively restricted to the DC_Ccl22 cluster (Supplementary Figure S1E). We developed a 46-gene human homolog signature using the top differentially expressed genes in DC_Ccl22 to examine an association with cancer patient outcomes. A higher Human Ccl22 Signature Score was associated with worse survival in The Cancer Genome Atlas (TCGA) Pan-Cancer Project (PANCAN, Supplementary Figure S1F).

Single cell RNA sequencing alone was unable to distinguish mregDC derived from each cDC lineage (mregDC1 and mregDC2), similar to previous studies which utilized surface marker labeling to distinguish these populations. (36) We performed spectral flow cytometry of radiosensitive MC38 and radioresistant B16 murine melanoma expressing the model tumor antigen ovalbumin (B16-OVA) to characterize changes in tumor cDC populations after treatment with 20 Gy (Figure 2 and Supplementary Figure S2–3). B16-OVA was also utilized to identify MHCI-SIINFEKL⁺ cDC populations (processed model OVA tumor antigen presented on MHCI). cDC cells were gated as Live CD45⁺ F4/80⁻ MHCII^HI CD11c^HI and either CD103⁺ (cDC1) or CD11b⁺ (cDC2) (Supplementary Figure S2A). cDC were plotted by UMAP and unsupervised clustering performed with FlowSOM. (61) In both tumor models unsupervised clustering revealed 4 tumor cDC populations (Figure 2A–B). Clusters were distinguished by expression of canonical lineage markers for cDC1 (CD24, XCR1 and CD103) and cDC2 (CD11b, CD172a), as well as maturation markers for mregDC1 and mregDC2 (lineage markers plus CCR7 and CD40) (Figure 2A, C and Supplementary Figure S2B–E). CCR7 and CD40 were characteristically upregulated in the DC_Ccl22 mregDC cluster and have been previously reported as mregDC markers. (36) Similarly, both mregDC1 and mregDC2 exhibited greater staining for MHCI-SIINFEKL, confirming mregDC as the major cDC antigen presenting population after RT (Figure 2C). Both mregDC1 and mregDC2 also had greater staining for the cell surface coinhibitory molecules PD-L1 and VISTA, a potent suppressor of the early stages of T cell activation, compared to their resting state cDC1 and cDC2 counterparts, respectively (Supplementary Figure S2F). (62–65) Herein, lineage-based resting populations are described as cDC1 and cDC2, whereas mregDC1 and mregDC2 contain both lineage markers and expression of mregDC markers.

In both MC38 and B16-OVA tumors, cDC1 and mregDC1 were significantly increased 5 days after 20 Gy (Supplementary Figure S3A–B and Figure 2D, respectively). In contrast, while a trend towards an increase in cDC2 and mregDC2 was observed in B16-OVA tumors, this was not the case in MC38. These observations were consistent with conventional gating approaches as both CCR7⁺ cDC1 and H2Kb-SIINFEKL⁺ cDC1 were significantly elevated 5 days after 20 Gy in B16-OVA tumors (Supplementary Figure S3C–D). Taken together with scRNA-seq findings, these observations show that RT increases a population of mregDC1 with increased regulatory markers and immune suppressive potential.

mregDC production of CCL22 drives Treg immune suppression after RT

The expansion of mregDC1 and their enrichment in regulatory markers suggested that they contribute to immune suppression after RT. The Treg-recruiting chemokine CCL22 was among the top genes enriched in mregDC (Log₂FC 3.7, P_adjusted = 9 × 10⁻⁷⁴; Figure 1A). Moreover, expression of Ccl22 among all CD45⁺ infiltrating tumor cells was completely restricted to the DC_Ccl22 cluster (Supplementary Figure S1E). Treg tumor infiltration and their role suppression of RT efficacy is well established. (12–14, 17) Consistent with this, expression of the Treg marker FOXP3 among pancreatic cancer patient tumor samples after RT was significantly increased in the dataset reported in Farren, et al (Supplementary Figure S4A). (53) We also observed a positive correlation between the Human DC_Ccl22 Homolog Score and both FOXP3 and CCL22 expression across the majority of TCGA solid tumor cohorts (Supplementary Figure S4B and Supplementary Figure S5, respectively).

Treg increased in both B16 and MC38 tumors 5 days after 20 Gy, with B16 exhibiting a greater overall increase (Figure 3A and Supplementary Figure S6A–B). Administration of a CCL22 neutralizing antibody (αCCL22) significantly decreased tumor Treg 5 d after 20 Gy in both models (Figure 3B and Supplementary Figure S6B). For further mechanistic studies we focused on B16 tumors given their more radioresistant phenotype and larger magnitude of Treg accumulation after RT. (66) Tumor levels of CCL22 protein were significantly elevated by day 3 after RT (Figure 3C). Previous studies suggest that mregDC1 exhibit increased Ccl22 expression in murine lung tumors compared with mregDC2. (36) To determine the contribution of cDC lineages in CCL22 production after RT we isolated splenic cDC from Zbtb46^cre Rosa^YFP mice in which cre recombinase under control of the Zbtb46 promotor expressed in cDC cleaves a loxP-flanked STOP sequence allowing expression of the fluorescent protein YFP in cDC. (67) Splenic cDC1 (YFP⁺ CD24⁺) and cDC2 (YFP⁺ CD172a⁺) were sorted, cocultured with either untreated or irradiated B16 tumor cells for 18 h, and then analyzed by flow cytometry for CCL22 expression (Figure 3D and Supplementary Figure S6C). Although cDC1 exhibited a higher baseline CCL22 staining compared with cDC2, coculture with irradiated tumor cells only induced CCL22 in cDC1.

To examine the role of cDC1 on Treg accumulation after RT we used flow cytometry to analyze Treg populations in Batf3^−/− mice, as they are deficient in cDC1 and continue to be absent from tumors after irradiation (Supplementary Figure S7A). B16 tumors in Batf3^−/− mice had significantly less Treg accumulation 5 d after 20 Gy compared to tumors in wild type (WT) mice (Figure 3E). To determine if cDC1 contribute to radioresistance we examined the tumor growth of B16F1 melanoma, a highly aggressive radioresistant cell line. (66) Despite a faster baseline tumor growth in Batf3^−/− mice, irradiated tumors responded significantly better to 20 Gy in Batf3^−/− mice compared to those in WT (Figure 3F). It is important to note that Batf3^−/− mice have additional alterations in immune cells that limit the use of this model. However, it has been previously demonstrated that Batf3^−/− mice exhibit favorable CD4⁺ T cell differentiation into Treg both in vitro and in the murine gut. (68) Taken together, these findings suggest that cDC1 is a major contributor to Treg accumulation and radioresistance through production of CCL22.

We hypothesized that perturbing the cDC1-CCL22-Treg axis may hold therapeutic potential to enhance RT efficacy. Mice bearing either B16 or MC38 tumors responded significantly better to RT in the presence of αCCL22 by both tumor growth and survival (Figure 3G–H and Supplementary Figure S7C–D, respectively). To determine if radiation-recruited Treg suppress the RT response and represent a potential therapeutic target, we treated tumor bearing mice with anti-CTLA4 clone 9H10, as this clone has been shown to deplete Treg in mice. (69–71) Concurrent treatment with αCTLA4 clone 9H10 significantly enhanced the antitumor activity of local RT and prolonged survival in mice bearing B16 and MC38 tumors (Figure 3G–H and Supplementary Figure S7C–D, respectively). These results highlight the therapeutic potential of targeting the cDC1-CCL22-Treg axis to overcome RT-driven immune suppression and enhance RT efficacy, as both CCL22 blockade and Treg depletion were independently able to augment the RT response.

Tumor targeted type I interferon reshapes tumor cDC populations

Though several clinical trials are ongoing, Treg depletion is not yet standard of care in humans due to a lack of selective targets and autoimmune adverse effects. (20, 21) IFN-I, a potent immune stimulator, has been shown to both suppress cDC CCL22 expression (72, 73) and drive cDC-mediated antitumor immunity after RT via enhanced T cell priming. (8) Indeed, the IFN-I IFNβ abrogated cDC1 CCL22 induction upon coculture with irradiated tumor cells (Figure 4A). Therefore, we hypothesized that IFN-I may relieve RT-driven cDC1-CCL22-Treg immune suppression and simultaneously enhance the priming of an antitumor CD8⁺ T cell response after RT.

Though long identified for its role in antitumor immunity, exogenous IFN-I use in cancer treatment has declined due to limited efficacy, poor pharmacokinetic properties, and dose limiting immune-mediated systemic side effects. (40–42) We previously developed an antibody-based tumor targeted IFN-I that is preferentially retained in the tumor microenvironment, improving efficacy and avoiding off-target systemic effects. (48) Based on this, we constructed a fusion protein consisting of the IFN-I IFNα4 linked to an antibody against human epidermal growth factor receptor (αEGFR), resulting in a heterodimeric αEGFR-IFNα fusion protein (Figure 4B). EGFR is a commonly overexpressed oncogenic membrane protein in various human cancers. (47) We hypothesized that αEGFR-IFNα can target IFN-I to the tumor microenvironment to increase potency, simultaneously inhibiting CCL22 production and driving an antitumor T cell response.

We performed bulk RNA-sequencing of live CD45⁺ CD11c⁺ cells isolated from B16 tumors expressing human EGFR (B16-EGFR) to investigate the effect of αEGFR-IFNα on antigen presenting myeloid cell populations after RT (Figure 4C–D and Supplementary Figure S8A–D). Mice received an intratumoral 20 μg of αEGFR-IFNα on the same day as RT and 10 μg 4 days later. Tumors were collected 3 days after the second dose (day 7 after RT). Compared to 20 Gy alone, cells isolated from tumors treated with 20 Gy and αEGFR-IFNα exhibited an upregulation in transcriptional programs associated with a response to IFN-I and positive regulation of innate immune activation (Figure 4C). In contrast, biological processes related to cell substrate adhesion, angiogenesis, cell migration and motility were significantly downregulated following the combination treatment (Supplementary Figure S8A). While the DC_Ccl22 cluster signature was enriched in tumor CD11c⁺ cells after 20 Gy alone, many of these transcripts were down regulated after treatment with both 20 Gy and αEGFR-IFNα (Figure 4D). Although tumor CD11c⁺ cells contain myeloid populations other than cDC, including monocytes and macrophages subsets, we detected a trend towards an overall decrease in Ccl22 as well as other Treg-recruiting chemokines (Ccl1, Ccl17, Ccl20 and Ccl28) in tumors treated with αEGFR-IFNα and RT (Supplementary Figure S8B–D). We measured Ccl22 mRNA by qPCR in live CD45⁺ CD11c⁺ Ly6C⁻ cells isolated from B16-EGFR tumors 3 d after 20 Gy. Expression of Ccl22 was 14-fold lower in the combination treatment group compared to irradiated tumors (Figure 4E). Similarly, CCL22⁺ cDC1 were decreased to near-undetectable levels in B16 tumors 3 days after treatment with RT and αEGFR-IFNα as measured by flow cytometry (Supplementary Figure S8E).

We examined changes in tumor cDC populations in the presence of αEGFR-IFNα after treatment with 20 Gy by flow cytometry (Figure 5). All cDC populations were significantly decreased in B16-EGFR tumors 5 d after irradiation in mice also receiving αEGFR-IFNα: cDC1, CCR7⁺ cDC1 (mregDC1), cDC2 and CCR7⁺ cDC2 (mregDC2) (Figure 5A). Although unsupervised clustering revealed 2 populations of cDC2, due to the low abundance of cDC1 and mregDC1 following αEGFR-IFNα treatment, clustering of pooled samples failed to distinguish multiple populations of cDC1 (Figure 5B–C). However, unsupervised FlowSOM clustering of all cDC revealed the expansion of a new population of cDC2 particularly enriched in tumors treated with αEGFR-IFNα (Figure 5B–D). Recent reports have detailed distinct cDC2 populations that become efficient CD8⁺ T cell activators in the presence of IFN-I. (34, 35) The new population of cDC2 that emerged after treatment with αEGFR-IFNα and radiation exhibited characteristic upregulation of markers for interferon-stimulated genes cDC2 (ISG⁺ cDC2) reported in Duong, et al., including CD80, CD86, AXL, MAR-1 and CD64 (Figure 5B). (35) Moreover, the proportion of ISG⁺ cDC2 significantly increased with the combination of 20 Gy and αEGFR-IFNα compared with irradiation alone (Figure 5D). We also observed an increase in the gene signature for interferon-stimulated cDC2 reported in Duong, et al. (35) and Bosteels, et al. (34) among bulk RNA-seq of tumor CD11c⁺ cells (Supplementary Figure S9A–B, respectively). Together, these findings demonstrate that αEGFR-IFNα reshapes tumor cDC populations after RT, decreasing CCL22⁺ cDC1 and inducing ISG⁺ cDC2.

αEGFR-IFNα overcomes RT-driven Treg immunosuppression and augments an effector T cell response to improve RT efficacy

Given the decrease in myeloid Ccl22 expression, increase in ISG⁺ cDC2, and known effects of IFN-I on cDC activation, we hypothesized that αEGFR-IFNα could simultaneously decrease CCL22 and Treg tumor accumulation, and augment an effector T cell response after RT. Indeed, both CCL22 and Treg were significantly decreased in B16-EGFR tumors 5 d after treatment with 20 Gy and αEGFR-IFNα compared to 20 Gy alone (Figure 6A–B, respectively). Moreover, the combination of αEGFR-IFNα and 20 Gy significantly increased tumor CD8⁺ T cell infiltration (Figure 6C). To determine the effect of αEGFR-IFNα on the tumor antigen-specific effector CD8⁺ T cell response we isolated CD8⁺ T cells from B16-EGFR-OVA tumors 7 d after irradiation and pulsed them with the OVA peptide ex vivo. The combination treatment significantly enhanced the effector CD8⁺ T cell response compared with any other treatment as measured by IFNγ ELISPOT assay (Figure 6D). These observations demonstrate that αEGFR-IFNα can both simultaneously alleviate RT-driven Treg immunosuppression and augment an antitumor effector T cell response.

We treated mice bearing either B16-EGFR or MC38-EGFR tumors with intratumoral αEGFR-IFNα in combination with 20 Gy radiation (Figure 6E–F and Supplementary Figure S10A–B, respectively). In both tumor models, the combination treatment of RT and αEGFR-IFNα significantly inhibited tumor growth and prolonged survival compared to any other treatment moiety alone. Moreover, 44% of mice bearing B16-EGFR and 100% of mice bearing MC38-EGFR exhibited complete tumor regression. These mice were rechallenged by a subcutaneous reinjection of tumor cells and all mice failed to reestablish tumors (Supplementary Figure S10C–D, respectively). To validate our prototype design for systemic administration of IFN-I we compared its effectiveness after IV administration with a molar equivalent dose of unmodified IFNα4 (free IFNα4, Figure 6G–H). Mice bearing B16-EGFR tumors responded significantly better to αEGFR-IFNα compared with either RT alone or in combination with free IFNα4. Among mice treated with RT and αEGFR-IFNα IV 29% exhibited complete tumor regression at 60 days (50 days after RT).

We examined expression of immunoregulatory markers and effector T cell-recruiting chemokines among CD11c⁺ RNA-sequencing data (Supplementary Figure S10E). The addition of αEGFR-IFNα tended to decrease expression of multiple immunoregulatory markers, including Ptgs2, Nos2, Ido1 and Arg1, although did increase Cd274. RT and αEGFR-IFNα also increased expression of the T cell-recruiting chemokines Cxcl10 and Cxcl11. Given this and the profound increase in CD8⁺ T cells after the combination treatment we depleted CD8⁺ T cells using αCD8 and examined the effect on the RT and αEGFR-IFNα response. CD8⁺ T cell depletion ablated RT + αEGFR-IFNα efficacy as measured both by tumor growth and survival (Supplementary Figure S10F–G). Taken together, these findings demonstrate that αEGFR-IFNα can block Treg infiltration after RT and simultaneously mediate an effector CD8⁺ T cell response to generate robust antitumor responses.