Notch-Regulated Dendritic Cells Restrain Inflammation-Associated Colorectal Carcinogenesis

Inflammation increases risk for cancer (1). Colorectal cancer is a multifactorial disease that progresses from transformed epithelial cells in a complex microenvironment composed of immune cells, cytokines, and gut bacteria. The immune system can affect all aspects of colorectal cancer development by suppressing tumor initiation and progression, as well as promoting proliferation and metastasis. The importance of the antitumor immune response is underscored by the rapid development of immune-based cancer therapies over the last 30 years (2). Although the effect of cancer immunotherapy and checkpoint blockade has centered largely around cytotoxic CD8⁺ T cells, emerging work has begun to reveal a critical role for dendritic cells (DC) in promoting T-cell antitumor immunity (3–8). However, exactly how DCs in a complex tumor immune microenvironment impact inflammation-associated colorectal cancer carcinogenesis has not yet been defined.

Previously, we reported that genetic disruption of the GDP-4-keto-6-deoxy-mannose-3,5,epimerase-4-reductase (Fx) gene results in spontaneous colitis, colitis-associated dysplasia, and ultimately adenocarcinoma (9). In this animal model, loss of Fx disrupts the conversion of GDP-mannose to GDP-fucose and leads to fucosylglycan deficiency (10), a condition phenocopied by deletion of the GDP-mannose 4,6-dehydratase (GMDS) gene (11), which is seen in 6% to 13% of colorectal cancers (9, 12). Notch signaling transactivation following Notch–ligand interaction requires the posttranslation modification of Notch with O-fucose addition to the consensus Ser/Thr residue present on the core ligand-binding EGF-like repeats and subsequent modification by Fringe, an N-acetyl-glucosaminyl transferase (13–15). As a result, fucosylglycan deficiency in the gut epithelium disrupts Notch and downstream Hes1 signaling, resulting in aberrant crypt proliferation and goblet cell expansion accompanied by profound inflammation and serrated-like lesions (16, 17). Importantly, loss of Hes1 expression is commonly observed in sessile serrated adenomas/polyps and colorectal cancers in the SSA pathway (9, 18). Fx deletion also alters blood cell homeostasis by affecting Notch-dependent T-cell differentiation and myelopoiesis (13, 15). A role for hematopoietic fucosyglycan deficiency in the gut inflammation and transformation seen in Fx^−/− mice was suggested by bone marrow chimeric studies, in which wild-type (WT) bone marrow cells have a colitis-limiting and tumor-suppressing effect (9). This finding prompted our current investigation of whether fucosyglycan deficiency in the context of aberrant Notch signaling alters immune surveillance in one or more blood lineages and promotes inflammation and progression of colitis-associated carcinogenesis. Notch signaling is evolutionarily conserved to determine cell fate and has emerged as a critical regulator of lymphoid cell development and T-cell function (19). Less elucidated is the significance of Notch in innate immune cell development and function. Several groups reported that Notch2 regulates terminal differentiation of splenic and lamina propria (LP) conventional DCs (cDC; refs. 20–22). In addition, Notch activation in DCs interacts with signaling triggered by various Toll-like receptor agonists to modulate inflammatory cytokine expression (23). Furthermore, Notch signaling in DCs is critical for evoking antitumor responses to mouse melanoma cells (24). However, the role of Notch-regulated cDC immune responses in colorectal cancer and particularly in the development of inflammation-associated colorectal cancer is unknown. Here, by using Fx^−/− mice as a model of colitis-associated colorectal cancer, we identified that Notch2-controlled DCs play a critical role in limiting the development of inflammation-associated colon cancer in mice. Through database analysis, we identified a link between Notch2 and cDC1 in human colorectal cancer and showed that fucosylation-deficient colorectal cancers have decreased CCR7 expression and suppressed cDC1 signature gene expression.

The animal research was approved by Case Western Reserve University Institutional Animal Care and Use Committee (Cleveland, OH). C57Bl/6 (Ly5.2) and B6.SJL-Ptrca Pep3b/BoyJ (B6.BoyJ:Ly5.1) mice were maintained in the laboratory (9, 13). Rbpj^flox/flox (Rbpj^F/F) mice were a gift from Dr. Tasuku Honjo (Kyoto University, Kyoto, Japan). Notch1^F/F mice were a gift from Dr. Ralph Kopan (University of Cincinnati College of Medicine, Cincinnati, OH). VavCre/Notch1^F/F and VavCre/Notch2^F/F mice were generated by crossing Vav-Cre mice (#008610; The Jackson Laboratory) with Notch1^F/F and Notch2^F/F mice (#010525; The Jackson Laboratory), respectively. Mx-Cre/Rbpj^F/F, CD4-Cre/Rbpj^F/F, and Lys-Cre/Rbpj^F/F mice were generated by crossing Mx-Cre (#003556; The Jackson Laboratory), CD4-Cre (#022071; Jackson Laboratory), and Lys-Cre mice (#004781; The Jackson Laboratory) with Rbpj^F/F mice, respectively. CD11c-Cre/Notch2^F/F mice were generated by crossing CD11c-Cre mice (#008068; The Jackson Laboratory) with Notch2^F/F mice. OT-1 mice (003831) were obtained from The Jackson Laboratory. Fx^−/− mice (Ly5.2 and Ly5.1) were generated in our laboratory and maintained on fucose-supplemented (on-fucose diet) chow diet (0.5% l-fucose) from weaning, as described previously (9, 13, 25). Briefly, in experiments where fucose-deficient Fx^−/− mice were used, mice were maintained on a fucose-supplemented diet until 12 weeks, and then on standard chow for 4 weeks before use. For bone marrow transplantation, all Fx^−/− recipient mice were maintained on fucose-supplemented diet before transplantation and for 14 days after transplantation, at which time they were switched to the standard diet. In some experiments, mice received antibiotics containing amoxicillin (0.06%), clarithromycin (0.01%), metronidazole (0.02%), and omeprazole (0.0004%) for 8 weeks (9).

Bone marrow transplantation was performed in lethally irradiated (950 cG) 8- to 12-week-old mice (Ly5.1) by intravenous transfer of 2 × 10⁶ donor cells (Ly5.2; ref. 13). Mouse colitis scoring and dysplasia scoring were performed as described previously (9). Briefly, the final histologic score was the sum of the indices of active inflammation, chronic inflammation, transmural inflammation, as well as the ulceration (scale 0–3) and regeneration (scale 0–4). The index of inflammation (active, chronic, and transmural) was the product of the intensity of the inflammation (scale 0–3) and the area of involvement (scale 0–4). Dysplastic scores were determined by the percentage of epithelial cells showing cytologic dysplastic changes.

Mouse spleens were crushed on a sterile 40-μm cell strainer using the flat end of the plunger from a sterile 1 mL syringe. Dissociated cells were passed through the strainer using 1 mL of FACS buffer (Hank's Balanced Salt Solution with 0.5% BSA and 1 mmol/L EDTA). The cell suspension was subjected to red cell lysis using the Red Blood Cell Lysis Buffer (11814389001, Sigma) for 5 minutes at room temperature. cDC1 enriched cells were isolated from the dissociated splenocytes by negative selection with biotinylated B220 (RA3-6B2, #103204, BioLegend) and CD11b (M1/70, #101204, BioLegend) antibody and anti-Biotin MicroBeads (#130-090-485, Miltenyi Biotec), followed by positive selection with CD11c MicroBeads (#130-125-835, Miltenyi Biotec).

To prepare bone marrow–derived DCs (BMDC), bone marrow cells were flushed from femurs and tibias using a 21-gauge needle with FACS buffer, followed by red cell lysis and filtered through a sterile 40-μm cell strainer. Lysed bone marrow cells were suspended in RPMI (#SH30027.01, HyClone) supplemented with 10% FBS (#SH30088.03, HyClone), 1% l-glutamine (#25030081, Thermo Fisher Scientific), 1% sodium pyruvate (#11360070, Thermo Fisher Scientific), 1% MEM-NEAA (#11140050, Thermo Fisher Scientific), 55 μmol/L 2-mercaptoethanol (#21985023, Thermo Fisher Scientific), and 100 ng/mL Flt3L (#250–31L, PeproTech). Cells were plated in 6-well plates at 4 × 10⁶ cells per well and cultured at 37°C in a humidified atmosphere at 5% CO₂ for 7 days.

To prime DCs with DLL1, BMDCs were prepared as described previously and cultured first with Flt3L for 3 days. Cells were then cocultured with OP9 cells transduced with retroviruses encoding GFP or Notch ligand, DLL1, starting on day 3 and cocultured for 4 more days. OP9 and OP9-DLL1 cells were gifts from Dr. John Lowe (retired; affiliated with Case Western Reserve University, Cleveland, OH) and authenticated by genotyping and FACS analysis for GFP expression (13). These cells were authenticated and tested for Mycoplasma yearly. These cells were cultured in Minimum Essential Medium (MEM) Alpha Medium (#12561072, Thermo Fisher Scientific) supplemented with 20% FBS (#SH30088.03, HyClone). Prior to coculture with DCs, OP9 cells were treated with mitomycin C (10 mg/mL) for 2 hours. BMDCs primed with DLL1 were transferred into Fx^−/− mice via intravenous injection (1 × 10⁶/mouse) weekly for 6 weeks.

FACS analysis was performed on BD FACSAria I and BD CytoFLEX and analyzed using BD FACSDiva software version 4.1. and CytExpert software. Briefly, 0.2–1 × 10⁶ cells were incubated with 1:100 dilutions of appropriate antibodies in 0.2 mL FACS buffer on ice for 20 minutes, followed by two washes with FACS buffer (26). Antibodies used included: CD4 (RM4-5, #100508, BioLegend), CD8α (53-6.7, #100712, BioLegend), B220 (RA3-6B2, #103204, BioLegend), CD11b (M1/70, #101216, BioLegend), Gr-1 (RB6-8C5, #108404, BioLegend), MHCII (I-A/I-E; M5/114.15.2, #107606, BioLegend), TER119 (TER-119, #116204, BioLegend), CD103 (M290, #557495, BD), NK1.1 (PK136, #108704, BioLegend), ESAM (1G8, #136204, BioLegend), c-kit (2B8, #105812, BioLegend), Sca1 (D7, #108114, BioLegend), CD11c (HL3, #561241, BD), Flt3 (A2F10, #135306, BioLegend), CD24 (M1/69, #138506, BioLegend), XCR1 (ZET, #148206, BioLegend), CD115 (AFS98, #135532, BioLegend), and SIRPα (P84, #144028, BioLegend).

Transwell assay was performed using 24-well plates with 6.5 mm transwells with 5 μm pore polycarbonate membrane (#3421, Corning). Briefly, 100 μL of cDC1-enriched splenocyte suspension (2 × 10⁵/mL) in RPMI medium supplemented with 2% FBS, 1% l-glutamine, 1% sodium pyruvate, and 1% MEM-NEAA (migration medium) was added to the top chamber and 600 μL migration medium containing chemokines [100 ng/mL of CCL2, #250-10; CCL5, #250-07; CCL19, #250-17B; and CCL21, #250-013 (all from PeproTech) and XCL1, #783502, BioLegend] was added to the bottom chamber. Migrated cells at the bottom chamber were recovered 3 hours later and enumerated by flow cytometry by gating on the viable cDC1 cells.

For cross-presentation assays, endotoxin-free chicken ovalbumin (OVA; #S7951-1MG, Sigma-Aldrich) was added (10 ng/mL) to enriched cDC1 cells. CD8⁺ T cells were isolated from spleens of OT-1 mice by negative selection using biotinylated antibodies (B220, Ter119, CD11b, Gr1, CD11c, and NK1.1) and Streptavidin Microbeads (#130-048-101, Miltenyi Biotec). CD8⁺ T cells were stained with carboxyfluorescein diacetate succinimidyl ester (CFSE, 1 μmol/L; 65-0850-84, Invitrogen) in RPMI supplemented with 10% FBS at 37°C in a humidified atmosphere at 5% CO₂ for 15 minutes and washed twice with the same staining buffer. A 96-well plate was precoated with 1:1,000 anti-CD3 (145-2C11, #557306, BD) and anti-CD28 (D665, #566883, BD) in PBS for 90 minutes. A total of 10 × 10⁴ cells per well were cocultured in the precoated 96-well plate with 2 × 10⁴ CD11c⁺ cells per well in RPMI supplemented with 10% FBS for 3 days at 37°C in a humidified atmosphere at 5% CO₂. T-cell proliferation was assessed for CFSE dilution using the FACS buffer indicated by the FITC fluorescence signal on the FACSAria I.

The upstream 5,000 bp promoter sequence (−5,000 to −1 nt) of mouse Ccr7 gene was searched for the recombination signal binding protein-Jk (RBPJ) binding motif, TG(G/A)GAA, in both positive and negative complement strands by using BioEdit 7.2 software. BMDCs primed with Flt3L for 6 days as described above (see DC isolation, immunophenotyping, differentiation, and priming with Notch ligand) were transferred to 6-well plates precoated with recombinant mouse DLL1-Fc (2.5 μg/mL, #5026-DL-050, R&D Systems) and cultured for 17 hours. Chromosome immunoprecipitation (ChIP) assay was performed using an EZ-ChIP Assay Kit (#17-295, Millipore) according to the manufacturer's instructions. Briefly, 3 × 10⁶ harvested BMDCs were fixed with formaldehyde with a final concentration of 1% and incubated for 10 minutes at 37°C. Crosslinking was stopped by glycine (0.125 mol/L) for 5 minutes. Cells were resuspended in 600 μL of SDS lysis buffer with Protease Inhibitors (#20-163, Millipore) and incubated for 10 minutes on ice. Cell lysate was sonicated for 60 cycles, each with a 10-second pulse at 50% of maximal power, followed by a 20-second cooling period on ice to shear DNA to lengths of about 200 bp. Samples were subjected to immunoprecipitation using rabbit polyclonal anti-RBP-JK antibody (#25949, Abcam) overnight at 4°C, followed by adding 60 μL of Protein A Agarose (catalog no., 16–157C, Millipore) for 1 hour at 4°C. The immunoprecipitated chromatin was analyzed by qRT-PCR as described below (see Affymetrix GeneChip array and qRT-PCR analysis). The primers used are listed in Supplementary Table S1.

Total RNA was extracted from spleen cDC1 cells using the RNeasy Mini Kit (#74104, Qiagen). RNA was reverse transcribed and labeled according to the instruction of Affymetrix WT Pico protocol by using 9 ng total RNA (GeneChip WT Pico Reagent Kit, #902622, Thermo Fisher Scientific), and 5.5 μg of cDNAs were hybridized to Affymetrix Gene Chip Mouse Gene 2.0 ST Array (#902118, Thermo Fisher Scientific). GeneChips were scanned using Affymetrix GeneChip Scanner 3000. Normalized robust multi-array average values were calculated and used to calculate fold change for each gene and sorted by log₂ fold change (GSE163958). A total of 327 genes were identified with a fold change greater than 2 (log₂ FC > 1) between cDC1-enriched cells isolated from on-fucose and off-fucose mice. A heatmap of the top 50 genes with the greatest fold change was generated using Clustvis. The 327 genes were cross-referenced against the Molecular Signature Database (Broad Institute) to identify gene sets enriched in differentially expressed genes.

For qRT-PCR, total RNA was prepared from freshly isolated cDC1 cells as described above. A total of 500 ng of total RNA was reverse transcribed using the Bio-Rad iScript Select cDNA Synthesis Kit (#170-8897, Bio-Rad) in 20 μL reactions following the manufacturer's instructions. A total of 2 μL of cDNA was amplified with the Bio-Rad iQ SYBR Green Supermix (#1708880, Bio-Rad) in 10 μL PCR reactions and the primer sets specific for mouse genes in triplicates with the CFX96 Touch Real-Time PCR Detection System (Bio-Rad). The relative gene expression was calculated on the basis of ΔΔC_t method and normalized to β-actin. The sequences of the primer sets are listed in the Supplementary Table S2.

Colorectal cancer data from The Cancer Genome Atlas (TCGA) Pan-Cancer Atlas (27) were accessed using cBioPortal. The Pan-Cancer Atlas includes 592 patients with colorectal cancer who have mRNA expression data available and 120 individuals were available for survival analysis. The mRNA coexpressions were reported and graphed for genes of interest on the basis of RSEM values batch normalized from Illumina HiSeq RNASeq data. Normalized expression values were transformed into Z-scores and ranked by the mean expression value of signature genes. Spearman correlation between each expressed gene was reported along with the P value and multiple testing corrected q-value using cBioPortal. Overall survival for patients relative to genes of interest was analyzed by using the gene expression profiling interactive analysis web server by categorizing samples as “high” or “low” based on the 75% quantile of the expression values (28). Mutation profiling and copy-number variation (CNV) were assembled to classify samples into GMDS WT and mutant groups separately. Samples with either CNV score less than 0 or detected mutation on gene of interest were classified as mutant samples, while the rest were considered as WT. Data processing and statistical analysis were implemented in R scripts. R package DESeq2 (https://bioconductor.org/packages/release/bioc/html/DESeq2.html) was applied to perform differential analysis, and R package fgsea (https://bioconductor.org/packages/release/bioc/html/fgsea.html) was used for gene set enrichment analysis (GSEA).

Study of archived human colorectal cancer was approved by the Institute Review Board of the University Hospitals Case Medical Center (Cleveland, OH). A total of 90 deidentified human colorectal cancer specimens were screened for GMDS mutation using primers for exons 1–2, 2–4, and 5–7, respectively, as described previously (29). GMDS WT (n = 9) and GMDS-mutant colorectal cancers (n = 7) were included in the study.

Data are presented as means ± SD, unless otherwise stated. Differences in variables were assessed by Student t test for two groups or one-way ANOVA for at least three groups using SPSS version 20 software.

Reconstituting Fx^−/− mice with WT hematopoietic cells results in decreased intestinal inflammation and more than 50% reduction of colorectal cancer (9). We first studied whether defective Notch function in hematopoietic cells of Fx^−/− mice contributes to the progression of colitis-associated cancer development in these mice by reconstituting the hematopoietic compartment of Fx^−/− mice with Notch-deficient bone marrow cells from Mx1-Cre/Rbpj^F/F (RBP-Jk^−/−or R^−/−) mice because the RBP-Jk or RBPJ is the critical transcription factor downstream of all four mammalian Notch receptors. Bone marrow cells from Mx1-Cre/Rbpj^+/+ or Rbpj^F/F (WT) and Mx1-Cre/Rbpj^F/+ (RBP-Jk^+/− or R^+/−) were used as control donors. WT mice were used as recipient controls. All Fx^−/− recipient mice were maintained on fucose-supplemented diet before transplantation and for 14 days after transplantation to allow mice to acquire a fucosylation-replete phenotype through a fucosylation salvage pathway (25). Mice were then switched to regular diet (off-fucose diet) and studied for histology and immune cell function 2 months after diet switch; this timepoint was selected on the basis of our previous studies where Fx^−/− recipients showed defined inflammation and dysplasia histology after receiving WT bone marrow transplantation (9).

Colonic histology analysis revealed that Fx^−/− mice receiving WT cells or cells carrying a single copy of Rbpj (Rbpj^+/−) displayed variable colitis and dysplasia (8.8 and 28 were inflammation and dysplasia indexes for RBP-J^+/−, while 6.1 and 19.4 were for WT, respectively). Invasive adenocarcinoma developed in 40% of Fx^−/− mice receiving RBP-Jk^+/− cells, but only developed in 2 of 12 (16.7%) mice receiving WT cells (Table 1). In comparison, Fx^−/− mice receiving RBP-Jk^−/− cells displayed more severe inflammation and dysplasia, showing statistically significantly higher mean scores reaching 13 and 50, respectively, when compared with mice receiving WT bone marrow (Fig. 1A–D; Table 1), and 7 of 9 mice developed adenocarcinoma (Table 1). In WT recipients, neither RBP-Jk^+/− nor RBP-Jk^−/− cells caused inflammation or dysplasia. Accordingly, we found that most inflammatory cytokine expression by WT recipients was not affected by the type of bone marrow cells transplanted. In contrast, consistent with our previous report that Fx^−/− gut mucosa was sufficient to induce inflammation and dysplasia (9), most of the inflammatory cytokines expressed by the colonic epithelium were increased in Fx^−/− mice receiving control RBP-Jk^+/− cells. Inflammatory cytokine expression was further enhanced in Fx^−/− mice receiving RBP-Jk^−/− cells (Fig. 1E). These findings suggest that Notch signaling deficiency in hematopoietic cells promotes inflammation-associated carcinogenesis of the intestinal epithelium of Fx^−/− mice in a dose-dependent manner. Consistent with our previous findings that the microbiome is critical for promoting inflammation-associated carcinogenesis, antibiotic treatment completely eliminated dysplasia and significantly decreased inflammation in Fx^−/− mice receiving RBP-Jk^−/− cells (Fig. 1A and B, two rightmost columns). None of the mice treated with antibiotics developed adenocarcinoma.

To discern which Notch receptor (Notch 1–4) is required for limiting inflammation-associated transformation in colon epithelium, we reconstituted lethally irradiated Fx^−/− mice with Notch1- or Notch2-deficient bone marrow cells (Vav-Cre/Notch1^F/F or Vav-Cre/Notch2^F/F) or their controls (Vav-Cre/Notch1^F/+ or Vav-Cre/Notch2^F/+) and compared the gut pathology after transplantation. Only mice reconstituted with Notch2-deficient (Supplementary Fig. S1A–S1D), but not Notch1-deficient (Supplementary Fig. S1E–S1H), hematopoietic cells developed significant colitis and dysplasia. The inflammatory and dysplastic indexes in these mice were similar to those recorded for Fx^−/− mice reconstituted with the RBP-Jk^−/− cells (Supplementary Fig. S1A and S1B). Accordingly, inflammatory cytokine expression was further enhanced in Fx^−/− mice receiving Notch2-deficient bone marrow cells (Supplementary Fig. S1I). Half of the mice receiving Notch2-deficient bone marrow cells developed colorectal cancer, while none of the mice receiving Notch1-deficient cells had cancer (Table 1).

Because Notch1 is critical for T-cell differentiation and adaptive immune function (19), and RBP-Jk–dependent signaling has also been shown to promote T-cell cytotoxicity (30), we asked whether defective Notch function in T cells promotes inflammation-associated cancer development. Surprisingly, we found that loss of Notch signaling in T cells does not enhance inflammation and dysplastic progression in Fx^−/− mice (Supplementary Fig. S2). None of the recipient mice developed colorectal cancer after receiving CD4-Cre/Rbpj^F/F cells (Table 1). Therefore, lack of Notch signaling in hematopoietic cells promoted inflammation-associated transformation in this model through a Notch2-controlled and T-cell–independent mechanism. Furthermore, we excluded a role of Notch-deficient myeloid cells in this process (Supplementary Fig. S3).

Because Notch2 regulates terminal DC differentiation (20–22), we sought to determine whether the phenotype associated with Notch2-deficient bone marrow reconstitution of Fx^−/− mice can be recapitulated by Notch2-deficient DC bone marrow chimerism. We transplanted CD11c-Cre/Notch2^F/F marrow cells to WT or Fx^−/− mice. CD11c-Cre/Notch2^F/+ or WT cells were used in control transplantation. Compared with Fx^−/− mice receiving WT cells, Fx^−/− mice receiving CD11c-Cre/Notch2^F/F cells displayed significantly higher inflammation (9.6 and 7 for recipients of CD11c-Cre/Notch2^F/F and WT bone marrow, respectively; P = 0.025; Fig. 2A) and dysplasia scores (38.4 and 17.9 for recipients of CD11c-Cre/Notch2^F/F and WT bone marrow, respectively; P = 0.01; Fig. 2B). These findings were consistent with the histologic analysis (Fig. 2C and D). In addition, around 45% of mice developed cancer in the CD11c-Cre/Notch2^F/F group. These inflammation and dysplasia scores were similar to those recorded for Fx^−/− mice receiving Fx^−/− bone marrow cells. Notably, mice receiving DCs carrying a single copy of Notch2 had moderately increased levels of inflammation and dysplasia (Table 1), where 40% of mice also developed adenocarcinoma, doubling the incidence compared with Fx^−/− mice receiving WT DCs. These findings suggest that Notch2-regulated DCs attenuated inflammation-associated carcinogenesis in this colitis-associated colorectal cancer model. Like RBP-Jk–deficient hematopoietic cells, Notch2-deficient DCs promoted colitis-associated carcinogenesis in a dose-dependent manner.

We next examined how DC frequency and function are altered in Fx^−/− mice. We characterized the DCs in Fx^−/− mice by the markers expressed on cDC1 and cDC2. We found that the numbers of splenic, mesenteric lymph node (mLN), and LP DCs were decreased in Fx^−/− mice maintained on regular diet compared with those maintained on fucose-supplemented diet (Fig. 3A). In addition, cDC1 (CD11c⁺MHCII⁺CD8⁺CD11b⁻) and cDC2 (CD11c⁺MHCII⁺CD8⁻CD11b⁺; Fig. 3B) frequencies were decreased by 92% and 68%, respectively, in the spleen. The expression of XCR1 on cDC1s was markedly decreased in off-fucose Fx^−/− mice (Fig. 3C). The analysis of bone marrow DC progenitors revealed a mild reduction in the frequencies of committed precursors of cDCs (pre-cDC), but similar frequencies of the macrophage and DC precursors (MDP) and the common DC precursors (CDP; ref. 31) in off-fucose Fx^−/− mice compared with controls (Supplementary Fig. S4A and S4B). In mLN, cDC1 frequencies were also decreased by 63%, while cDC2 remained unchanged (Supplementary Fig. S4C). LP and mLN migratory CD103⁺ DCs were decreased by 34% and 44% in frequencies, respectively, in off-fucose Fx^−/− mice (Fig. 3D). Because similar levels of inflammation and dysplasia were observed in nontransplanted Fx^−/− mice and Fx^−/− mice reconstituted by CD11c-Cre/Notch2^F/F bone marrow cells, we suspected that Notch2 signaling loss in the hematopoietic compartment could similarly alter the frequencies of DC subsets. Analysis of DCs derived from CD11c-cre/Notch2^F/F donors in chimeric recipients revealed that total splenic DC numbers were decreased in both WT and Fx^−/− recipients compared with recipients receiving control (CD11c-cre/Notch2^F/+) cells. In addition, decreased cDC1s and cDC2s derived from CD11c-cre/Notch2^F/F donors in both WT and Fx^−/− recipient mice recapitulated the altered DC differentiation observed in Fx^−/− mice (Fig. 3E). However, MDPs, CDPs, and pre-cDCs were not changed in the CD11c-cre/Notch2^F/F mice, even though XCR1 expression was decreased in CD8⁺ DCs (Supplementary Fig. S5A–S5E).

Although decreased numbers of cDC1 would limit their function, the impaired Notch2 signaling in DC subsets may directly impair antitumor activity. To investigate this possibility, we assessed the migration of DCs toward chemokine ligands in transwell migration assays. We found that cDC1-enriched cells isolated from the spleens of off-fucose Fx^−/− mice (off-fucose DC) displayed decreased migration toward CCL19 (Fig. 4A) and CCL21 (Fig. 4B) compared with cDC1-enriched cells from on-fucose Fx^−/− mice (on-fucose DC). Consistent with the altered expression of XCR1 in cDC1 of Fx^−/− mice, these cells showed a decrease in migration toward XCL1 (Fig. 4C). In comparison, both off-fucose and on-fucose DCs showed minimal migration toward CCL2 and CCL5 (Fig. 4B). Consistently, expression of CCR7, which is the receptor for CCL19 and CCL21, was decreased in DCs from off-fucose Fx^−/− mice, while expression of CCR2 and CCR5, which are the receptors for CCL2 and CCL5, respectively, remained unchanged (Fig. 4D). We then analyzed the ability of Fx^−/− DCs to cross-prime T cells. Using a coculture assay with CFSE-labeled CD8⁺ T cells from OVA-specific OT-I mice and OVA-pulsed cDC1 cells, we found decreased cross-priming of CD8⁺ T cells by freshly isolated cDC1 (Fig. 4E) from off-fucose Fx^−/− mice. We asked whether the decreased cross-priming is associated with aberrant IL12 expression. Although IL12 expression did not differ significantly between off-fucose DCs and control DCs, its expression was much lower in off-fucose DCs after lipopolysaccharide (LPS) stimulation (Fig. 4F). To determine whether impaired migration and cross-priming of off-fucose DCs from Fx^−/− mice are caused by the suppressed Notch2 signaling, we examined the migration and T-cell priming of CD11c-cre/Notch2^F/F DCs. We found similar reduction in migration of Notch2-deficient cDC1s and suppression of cross-priming of CD8⁺ T cells by Notch2-deficient DCs (Fig. 4G–K). Expression of CCR7 and IL12 was also decreased in Notch2-deficient DCs (Fig. 4J and L).

Notch regulates gene expression by forming a transcriptional complex with the DNA binding protein, RBPJ/CSL. To investigate whether Notch2 directly regulates CCR7 expression, we searched the promoter region of CCR7 and found several potential RBPJ/CSL binding motifs (TGGGAA). ChIP analysis of BMDCs stimulated by DLL1 showed RBPJ binds strongly with two RBPJ/CSL sites (∼2.8 and 2 kb) upstream of the CCR7 promoter. In comparison, none of the sites were bound by RBPJ in Notch2-deficient or Fx^−/− cells (Fig. 4M). These findings indicate that CCR7 is directly regulated by Notch2 signaling and its suppressed expression contributes to the aberrant migration of Fx^−/− and Notch2-deficient DCs. To gain a comprehensive understanding of how fucose deficiency impacts DC maturation and function, we performed gene expression profiling of cDC1 cells from Fx^−/− mice maintained with on-fucose or off-fucose diet. A total of 275 genes were upregulated and 52 genes were downregulated in response to fucose treatment (Supplementary Fig. S6A). Analysis of the top differentially regulated genes revealed that these genes were enriched in pathways such as immune-effector processes, myeloid leukocyte activation, exocytosis, and secretion. In addition, some genes regulate molecular functions, such as signaling receptor binding and endopeptidase activity (Supplementary Fig. S6B). The divergent gene expression implicated in altered immune-effector functions was validated by qRT-PCR, which showed upregulation of genes encoding the myeloid inflammatory proteins and the leukocyte immunoglobulin-like receptor subfamily B member 4 (Lilrb4) by fucose-deficient cDC1s. Importantly, expression of genes encoding chemokine receptors (Ccr7), migration (Fscn1 and Hspa8), antigen processing (H2-m2), and immune defense molecules, such as guanylate-binding proteins (Gbp9 and Gbp10), was downregulated in fucose-deficient cDC1s (Supplementary Fig. S6C).

To establish the relevance of Notch2 regulation of cDC1 in human colorectal cancer, we stratified patients in the colorectal cancer dataset of TCGA by expression of cDC1-associated genes (CCR7, XCR1, FLT3, CLEC9A, and THBD; refs. 32, 33) and found that a higher cDC1 signature in tumors was positively associated with survival (Fig. 5A). The correlation between CCR7 and representative cDC1 signature genes (BATF3, FLT3, CLEC9A, XCR1, and THBD) indicates that CCR7 serves as a marker of tumor-infiltrating cDC1s for human colorectal cancer (Supplementary Fig. S7A). We then compared the cDC1 signature gene expression and the expression NOTCH2 and DLL1, which encodes the primary Notch ligand that drives Notch2-dependent cDC differentiation (34), and found they have a significant positive correlation (Fig. 5B). Consistent with our finding that Notch2 regulates cDC1 function, we found NOTCH2 gene expression in colorectal cancer correlates with the cDC1 gene THBD (Fig. 5C) and FLT3, as well as XCR1 (Supplementary Fig. S7B). Furthermore, we found that expression of the cDC1 signature genes (BATF3, FLT3, CLEC9A, XCR1, and THBD) correlated with the CD8⁺ T-cell genes CD8A and CD3E in human colorectal cancer (Fig. 5D).

We then determined the impact of fucosylation deficiency in human colorectal cancer by performing GSEA after stratifying colorectal cancer patient data in TCGA on the basis of GMDS mutation. Supporting our findings in Fx^−/− mice, we found a reduction of CCR7 and CCL19 expression in human GMDS-mutated colorectal cancers compared with colorectal cancers with no GMDS mutation (Fig. 5E). In addition, GMDS-mutant colorectal cancers showed downregulation of FLT3 and genes encoding proteins involved in MHC class I antigen processing and presentation (TREML4) and in cytoskeletal assembly (LDB3; Supplementary Fig. S7C), but upregulation of genes encoding proteins involved in releasing and responding to inflammatory cytokines (TNFAIP2, LILAR2, and NUCB2) and the inhibition of cell migration (BST2; Supplementary Fig. S7D). Furthermore, we found that reduced GMDS expression is associated with poor disease-free survival (Supplementary Fig. S8). Finally, we verified that representative cDC1 signature genes (CCR7, XCR1, THBD, IRF8, and FLT3L) and chemokine ligands (XCL1, CXCL10, and CCL4) were suppressed in our cohort of GMDS-mutant colorectal cancers compared with GMDS WT colorectal cancers (Fig. 5F; refs. 5, 7, 35). In summary, these analyses support the hypothesis that Notch2 signaling regulates the tumor-associated cDC1 population and that fucosylation-deficient colorectal cancers have similar suppression of cDC1 signature genes as observed in our mouse model.

Finally, to demonstrate that Notch signaling in DCs restrains colitis-associated transformation, we adoptively transferred WT or Notch-primed DCs to Fx^−/− mice. Unprimed or Notch-primed DCs were both derived from bone marrow progenitors with Flt3L. Expanded DCs were transferred into Fx^−/− mice weekly for 6 weeks (Fig. 6A). The inflammation and dysplastic indices of Fx^−/− mice receiving unprimed DCs remained unchanged when compared with control PBS-treated mice (15.8 and 44 were inflammation and dysplasia indexes for DC-treated mice, while 15.6 and 46 were for PBS-treated mice, respectively). However, we found that mice receiving Notch-primed DCs had significantly decreased inflammation and dysplasia (Fig. 6B and C). The incidence of colorectal cancer in mice receiving Notch-primed DCs also decreased to 20% compared with 38% in the control group (P < 0.05). The histologic improvement was accompanied with decreased expression of inflammatory cytokines, including IL1β, IL6, Cox2, and TNFα, although expression of IFNγ was increased by around 2-fold, suggesting attenuated inflammation and improved cytotoxicity in response to DC infusion (Fig. 6D). In summary, these findings indicate that ex vivo Notch-primed DCs were able to restrain colitis-associated colorectal cancer development.

We have uncovered a novel mechanism implicating defective Notch-dependent DC function in promoting inflammation-associated dysplasia and progression to colon cancer transformation. We found more progression of inflammation-associated dysplasia to colorectal cancer in Fx^−/− mice reconstituted with Notch2-defective DCs, while adoptive transfer of Notch-primed DCs attenuated transformation. In addition, we revealed a direct correlation between Notch2 signaling and infiltrating cDC1s and the association of the suppressed cDC1 signature with a poor prognosis in human colorectal cancer. Our findings thus reveal a critical role for Notch2-dependent cDC1s in restraining inflammation-associated transformation in our mouse model and tumor progression in human colorectal cancers.

Our previous studies show that in our model of colitis-associated colorectal cancer, the intestinal epithelium of Fx^−/− mice is chronically inflamed and displays aberrant proliferation and progression of dysplasia to adenocarcinoma in a defined temporal and histopathologic sequence (9). Here, we revealed that fucose deficiency impairs DC function in a manner that is regulated by Notch2 signaling. The differentiation of cDCs depends on a few key transcription factors (for cDC1s, it is IRF8, BATF3, and ID2 and for cDC2s, it is IRF4 and ZEB2; refs. 36–38). It is also known that Notch2 is involved in the differentiation of terminal cDC1, CD11b⁺ESAM^hi cDC2, and CD103⁺CD11b⁺ intestinal DCs (20–22). In Fx^−/− mice, Fx locus deletion impairs Notch signaling by abolishing fucosylation and Fringe-mediated modification of Notch ligand binding EGF-like repeats, affecting the development of multiple hematopoietic lineages through dysregulated Notch1 or Notch2 signaling (13, 15). Here, we showed that decreased numbers of cDC1 and cDC2 cells in Fx^−/− mice phenocopied altered cDC differentiation in mice with Notch2 deleted in CD11c cells (CD11c-Cre/Notch2^F/F). This indicates that lack of fucosylation impairs Notch2-dependent signaling in final DC differentiation, as we found that the cDC precursors, including MDP and CDP, frequencies were not altered. Furthermore, the functional defects of Fx-null cDC1s were shared by Notch2-null cDC1s. Defective antitumor activity in the absence of Notch2 signaling was illustrated in vivo by bone marrow chimeric studies showing that gut inflammation and dysplasia were worsened by Notch signaling–deficient (Mx1-Cre/RBPJk^F/F) total bone marrow cells or Notch2-deficient (Vav-Cre/Notch2^F/F) cells. The crippled antitransforming activity of Notch2-deficient cells was attributed to DCs, but not T cells or myeloid lineage cells, in a dosage-dependent manner. Migration and cross-presentation to cytotoxic T cells were major functional defects identified in Notch2-deficient and Fx^−/− DCs, consistent with qRT-PCR and array-based transcriptome analysis showing downregulation of migration-regulating genes and antigen processing genes. Particularly, we found that Notch signaling directly regulates CCR7 expression and that downregulated CCR7 is accompanied with decreased migration toward CCL19 and CCL21 in both Fx-null and Notch2-null DCs. This is consistent with reports that CCR7 is critical for cDC1s to present tumor antigens and that Notch priming promotes cDC1 development and upregulates CCR7 (39–41). Although terminal differentiation of both cDC1 and cDC2 was affected in Fx^−/− mice, numbers of cDC1s were more prominently decreased in the draining lymph nodes, while the numbers of migratory CD103⁺ DCs were decreased in both mLN and LP. This further supports that defective DC migration is a prominent feature in this animal model. CD103⁺ DCs in LP and mLN are uniquely capable of generating gut-tropic CD8⁺ effector T cells (42). Indeed, we observed decreased cross-presentation to CD8⁺ T cells by Fx^−/− and Notch2-deficient cDC1s, but unchanged cDC2 cross-presentation to CD4⁺ T cells. Supporting this notion, human colorectal cancers with GMDS mutation not only had decreased CCR7 expression and cDC1 signature gene expression, but also show decreased expression of cDC1-recruiting CCL4 (7) and decreased XCL1, which is expressed by natural killer (NK) cells to recruit XCR1⁺ cDC1s to form cDC1/NK clusters within tumor tissues (43). It is possible that cDC1 development is compromised in the fucosylglycan-deficient environment of GMDS-mutant tumors, and/or cDC1 recruitment is suppressed in GMDS-mutant tumors by the altered cytokine milieu. Others have reported that GMDS-mutant colorectal cancers escape NK-cell–mediated surveillance by TRAIL-induced apoptosis. The underlying mechanism remains elusive, but has been found to be independent of O-glycosylation (29). Considering that Fx^−/− and Notch2-deficient cDC1s show reduced XCR1 expression, defective cDC1 recruitment through XCL1 may further compromise cDC1 migration and represent another mechanism of escaped tumor surveillance associated with fucosylation deficiency.

We extended our findings beyond fucosylation-deficient tumors by showing a direct link between infiltrating cDC1s and Notch2 signaling in human colorectal cancer. Furthermore, consistent with the finding that cDC1s recruit CD8⁺ T cells into the tumor microenvironment, we found that the cDC1 gene expression signature correlated with CD8A and CD3E expression in human colorectal cancer (5). Decreased intratumoral cDC1s and circulating cDC1s in patients with colorectal cancer relate to disease stage (44, 45). Our TCGA analysis extends these findings by revealing that lower cDC1 biomarker expression is associated with worse survival among patients with colorectal cancer. Unlike the Fx^−/− mice, human cancer cell extrinsic mechanisms of suppressed cDC1 signature associated with inhibited Notch activity may involve fucosylation-independent mechanisms. The correlation between the Notch2/DLL1 gene signature and the cDC1 gene signature suggests that downregulation of DLL1 in the tumor microenvironment could potentially account for the suppressed Notch2-dependent cDC1 homeostasis. Chronic inflammation regulates specific inflammatory conditions that can control activation of the Notch pathway in DCs. For example, DLL1 is downregulated, while JAG1 is upregulated in the intestinal epithelium of off-fucose Fx^−/− mice. JAG1 is upregulated in other cell types under various chronic inflammation conditions (46, 47). Both Notch ligands, JAG1 and DLL1, can activate Notch signaling, but show opposite effects on DC differentiation (34). While DLL1 induces DC differentiation, JAG1 stimulation limits full DC differentiation by inducing an altered Hes1 activation. How inflammation alters Notch ligand expression and induces aberrant Notch signaling to suppress DC antitumor immunity needs to be explored.

Chronic inflammation induces epigenetic modification and DNA modification in intestinal epithelial cells, promoting dysplasia and contributing to initiation and progression of colorectal cancer. During this process, innate immune cells and adaptive immune cells are recruited, which generate an environment with proinflammatory cytokines and highly genotoxic oxygen/nitrogen reactive species (48). Indeed, upregulation of an array of proinflammatory cytokines was a prominent feature of colonic epithelium of the Fx^−/− mice and upregulation of cytokines was further enhanced when mice were reconstituted with Notch2 signaling–deficient whole-bone marrow cells or DCs. In contrast, adoptive transfer of Notch-primed DCs decreased expression levels of inflammatory cytokines, including IL1β, IL6, and TNFα. Thus, Notch2-dependent DCs appear to limit protumorigenic inflammation in this model. It is interesting to note that Notch-deficient T cells appear to have a mild protective effect and decrease inflammation-associated dysplasia progression. However, the role of Notch signaling dictated by different Notch isoforms and the mechanism by which they coordinate immune responses in DCs and other immune cells to promote an inflammation-limiting and antitumor microenvironment remains to be further defined. Enhancing DC numbers and function within the precancerous and early-cancer lesions is paramount for cancer prevention (49). For immunotherapy, cDC1s are of particular interest in cellular vaccination strategies as they traffic tumor antigens to the draining lymph nodes and cross-present cell-associated neoantigens to cytotoxic T cells (4, 40). Our findings reveal a mechanism whereby Notch signaling plays a critical role in the regulation of DC migration and antigen presentation. These findings suggest a strategy of enhancing DC antitumor activity by stimulating Notch signaling to eliminate transformed cells and to prevent tumor progression. This approach potentially promotes DC migration and increases antigen delivery without patient-specific antigen targeting. Future work is required to identify molecules and pathways that are regulated by Notch or Notch-suppressing factors in the tumor microenvironment, which are potentially important adjuvants to enhance the efficiency of antigen presentation by DCs and the stimulatory capacity of these cells in the context of antitumor activity.

I. Maillard reports grants from National Institute of Allergy and Infectious Diseases during the conduct of the study. No disclosures were reported by the other authors.

L. Wang: Data curation, formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. S. Yu: Resources, data curation, formal analysis, supervision, validation, investigation, methodology. E.R. Chan: Resources, data curation, software, formal analysis, methodology. K.-Y. Chen: Resources, data curation, software, formal analysis, validation, methodology. C. Liu: Data curation, formal analysis, investigation. D. Che: Data curation, formal analysis, validation, investigation, methodology. A. Awadallah: Resources, data curation, validation, investigation. J. Myers: Resources, data curation, methodology. D. Askew: Resources, data curation, investigation, methodology. A.Y. Huang: Conceptualization, resources, data curation, supervision, validation, investigation. I. Maillard: Conceptualization, resources, data curation, formal analysis, supervision. D. Huang: Resources, data curation, formal analysis, validation, investigation. W. Xin: Conceptualization, resources, data curation, formal analysis, supervision, validation, investigation, methodology. L. Zhou: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing.

The authors thank Dr. Kenneth Murphy and Ivaylo Ivanov for providing bone marrow cells from CD11c-Cre/Notch2^F/F mice for part of this study. They thank Ms. Alison W. Xin for editing this article. This work was supported, in part, by research funding from NCI CA222064, NIH HL103827, Case GI SPORE Research Development Award, and National Institute of Diabetes and Digestive and Kidney Diseases Digestive Diseases Research Core Centers (NIDDK DDRCC) Pilot/Feasibility Award (to L. Zhou) and by the Department of Pathology Case Western Reserve University faculty start-up fund (to W. Xin and L. Zhou).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.