Tn Antigen Expression Contributes to an Immune Suppressive Microenvironment and Drives Tumor Growth in Colorectal Cancer

Expression of the tumor-associated glycan Tn antigen (αGalNAc-Ser/Thr) has been correlated to poor prognosis and metastasis in multiple cancer types. However, the exact mechanisms exerted by Tn antigen to support tumor growth are still lacking. One emerging hallmark of cancer is evasion of immune destruction. Although tumor cells often exploit the glycosylation machinery to interact with the immune system, the contribution of Tn antigen to an immunosuppressive tumor microenvironment has scarcely been studied. Here, we explored how Tn antigen influences the tumor immune cell composition in a colorectal cancer (CRC) mouse model. CRISPR/Cas9-mediated knock out of the C1galt1c1 gene resulted in elevated Tn antigen levels on the cell surface of the CRC cell line MC38 (MC38-Tnhigh). RNA sequencing and subsequent GO term enrichment analysis of our Tnhigh glycovariant not only revealed differences in MAPK signaling and cell migration, but also in antigen processing and presentation as well as in cytotoxic T cell responses. Indeed, MC38-Tnhigh tumors displayed increased tumor growth in vivo, which was correlated with an altered tumor immune cell infiltration, characterized by reduced levels of cytotoxic CD8+ T cells and enhanced accumulation of myeloid-derived suppressor cells. Interestingly, no systemic differences in T cell subsets were observed. Together, our data demonstrate for the first time that Tn antigen expression in the CRC tumor microenvironment affects the tumor-associated immune cell repertoire.


INTRODUCTION
Tumor cells are frequently characterized by an aberrant glycosylation profile, hence, current research is focused on how tumor-associated glycan structures support tumor progression. Glycosylation is a post-translational modification of proteins and lipids, which is not templatedriven, but instead, is influenced by the metabolic state of the cell and the availability of the sugar donors. Glycan structures are known to drive diverse biological functions, among others, protein folding, cell-cell and cell-matrix adhesion and cell signaling (1). Moreover, the ability of glycan structures to modulate immune responses has led to the hypothesis that tumor-associated glycan structures are responsible for skewing the tumor microenvironment toward an immune suppressive phenotype (2). However, our knowledge on how individual tumor-associated glycan structures actually affect anti-tumor immunity is still quite limited.
Compared to their non-malignant counterparts, tumor cells express much higher levels of the Tn antigen glycan structure (3,4). Expression of Tn antigen is a prognostic factor for overall and relapse-free survival in lung adenocarcinoma (5) and correlates with metastatic potential in colorectal cancer (6). Tn antigen is made up of one N-acetylgalactosamine (GalNAc) monosaccharide and represents the initiation of mucin-type O-glycosylation. In humans, a repertoire of twenty GalNAc-transferases (GalNAcTs) is responsible for the initiation of O-glycosylation (7) and thus Tn antigen synthesis. The expression of GalNAcTs is dynamic and these enzymes have been shown to relocate from the Golgi to the ER in certain tumor types (8). Although contradictory findings have been published (9), this relocation is thought to enable a prolonged mode of action, leading to overexpression of Tn antigen, thereby promoting cancer cell invasiveness. Since epithelial cells produce high levels of mucin proteins that are heavily O-glycosylated, high expression of Tn antigen predominantly occurs in epithelial cancer types, including colorectal cancer (CRC), where 86% of primary and metastatic human CRC tissues express the Tn epitope (10).
Tn antigen overexpression has been shown to directly induce oncogenic features including enhanced cell proliferation, decreased apoptosis, increased adhesion and migratory capacities (11)(12)(13). Since immune cells highly express receptors recognizing glycan structures (14), inhibition of anti-tumor immunity might be an additional mechanism by which Tn antigen supports tumor growth. However, whether cancer cells exploit Tn antigen in vivo to evade immune attack has never been thoroughly investigated.
In the present study we assessed the impact of Tn antigen on in vivo tumor growth and the immune cell composition present at the tumor site using CRISPR/Cas9 glycoengineered mouse colorectal cancer MC38 cells. We report that overexpression of Tn antigen drives tumor growth in CRC, which coincided with reduced tumor immune cell infiltration, increased myeloid-derived suppressor cells and decreased CD8 + T cell infiltration. Together, these data suggest that Tn antigen may promote an immune suppressive tumor microenvironment, which could contribute to tumor immune evasion and thus tumor progression.

Generation of the MC38-Tn high Cell Line
MC38 cells were cultured in DMEM supplemented with 10% heat inactivated fetal calf serum (FCS, Biowest), 1% penicillin and 1% streptomycin. MC38 cells were transfected with CRISPR/Cas9 constructs either targeting the C1galt1c1 gene (MC38-Tn high ) or an empty CRISPR/Cas9 construct (MC38-MOCK). For transfection, Lipofectamine LTX with PLUS TM reagent (ThermoFisher Scientific) was used and applied according to the manufacturer's protocol. Transfected MC38 cells were selected in bulk based on their Tn antigen cell surface profile as described below. Transfected MC38 cells were incubated with 5 µg/mL of the biotinylated α-N-acetylgalactosaminespecific lectin Helix Pomotia agglutinin (HPA, Sigma) for 1 h on ice, washed with medium and subsequently incubated with streptavidin-PE (Jackson ImmunoResearch) again for 1 h on ice. Cells were washed with medium and sorted in bulk on HPA high binding cells. The sorting procedure was performed twice to obtain the final MC38-Tn high cell line.

Surveyor Assay
To obtain genomic DNA, fresh cells were harvested and DNA was isolated with the Quick-DNA TM kit (Zymo research) according to manufacturer's instructions. The C1galt1c1 gene was amplified with qPCR (forward primer: CTGGCGGTCTGCCTGAAATA, reverse primer: TGTACAAGCAGACTTCAATG). The qPCR products (416 bp) were hybridized and treated with Surveyor Nuclease (Surveyor Mutation Detection Kit, Integrated DNA Technologies), which recognizes and cleaves any DNA mismatches. To visualize the mutation, the Surveyor Nuclease-treated products were separated by DNA agarose gel electrophoresis.

T Synthase Assay
Cell lysates were obtained using 0.5% Triton X-100 in TSM (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 2 mM MgCl 2 , 1 mM CaCl 2 ) and subsequently used for the T synthase assay as described by Ju and Cummings (16). Shortly, T synthase present in cell lysates utilizes the commercially available acceptor-substrate GalNAcα-4MU (Sigma Aldrich) and the donor-substrate UDP-Gal (Sigma Aldrich) to generate Galβ1-3GalNAcα-4MU structure. This product is hydrolyzed by O-glycosidase (New England Biolabs), leading to free 4-MU that is highly fluorescent and can be measured at an excitation of 355 nm and emission of 460 nm. The protein concentration in the cell lysate was determined with Pierce TM BCA protein assay kit (ThermoFisher Scientific) to determine T synthase enzyme activity per microgram protein.

Determination of the Cellular Glycosylation Profile
MC38 cells were incubated with 5 µg/mL of the HPA or peanut agglutinin (PNA, Vector Laboratories) lectin, anti-Tn antigen antibodies [5F4, kindly provided by H. Clausen and H. Wandall (17)] or 10 µg/mL of mouse MGL-2-Fc for 30 min at 37 • C. Cells were washed and incubated with streptavidin-APC (BD Biosciences), goat anti-mouse IgM-A488 (ThermoFisher Scientific) or goat anti-human IgG-FITC (Jackson ImmunoResearch) respectively for 30 min at 37 • C. Cells were washed and acquired on the Beckman Coulter Cyan flow cytometer. Data was analyzed with FlowJo v10.

CellTiter-Blue R Cell Viability Assay
The CellTiter-Blue R Cell Viability assay (Promega) was used according to the manufacture's protocol to measure the metabolic activity of MC38 cells. Briefly, 30.000 MC38 cells were cultured in a 1:6 dilution of the CellTiter-Blue R Reagent. The metabolic activity was measured during the first 24 h of culture using a FLUOstar Galaxy (MTX Lab systems) with an excitation and emission of 560 and 590 nm, respectively.

mRNA Library Preparation
The mRNA library was prepared as described previously (18). MC38 cell lines (passage 4-6) from three independent passages were harvested at three independent time points with 1% Trypsin EDTA and washed with PBS. Total RNA was extracted with a standard TRIzol isolation protocol (Thermo Fisher, 15596018). Quantity and purity were tested using the Nanodrop-2000 spectrophotometer (Nanodrop Technologies, United States). The library was synthesized using the TruSeq R Stranded mRNA Sample preparation kit (Illumina, RS-122-9004), according to manufacturer's LS protocol. The product quality during library generation was analyzed on the Agilent 2100 Bioanalyzer using the DNA 7500 chip (Agilent Technologies, 5067-1506).
The R package edgeR (v3.18.1) (24, 25) was used for library size adjustment, trimmed mean of M-values (TMM) normalization and differential expression analysis. Multidimensional scaling (MDS) plots were used to visualize sample distribution among MC38 cell lines. For differential expression analysis, the negative binomial dispersion was shrunken toward the common dispersion. EdgeR's exact test for two-group comparison was used for computing p-values. Statistical differences in mRNA expression were identified using the following pairwise comparisons: (1) MC38-WT vs MC38-MOCK cells and (2) MC38-Tn high vs MC38 MOCK cells. Here, per comparison, genes with more than 4 zeros across the 6 samples were discarded a priori. Significance was assessed using Benjamini-Hochberg false discovery rate (FDR) <0.05. Sequencing data is publicly available at the Sequence Read Archive (SRA) Gene Expression Omnibus thought GSO Series accession number GSE143700.

Comparative Analysis of Gene Sets
Comparative analysis was performed among the differentially expressed genes (DEGs) in the (1)

Gene Ontology Term Enrichment and Pathway Analysis
Gene Ontology (GO) term enrichment analysis was performed on the 1348 DEGs in the MC38-Tn high cells using Cytoscape v3.6.0 4 (26), and the ClueGO plugin v2.5.0 5 (27). Significantly enriched GO terms (Benjamini-Hochberg correction, False discovery rate (FDR) <0.05) were determined with the ontology source GO_BiologicalProcess-EBI-Quick-GO-GOA, and were subsequently visualized using view style Groups, GO level 6-13, and a kappa score threshold of 0.4 (Supplementary Table 4). Normalized counts of GO term-associated genes were visualized in heatmaps using Morpheus 6 . Genes assigned to GO terms on antigen presentation and T cell activation (GO level 3-8, FDR < 0.05, Supplementary Table 5) were included in the heatmap depicted in Figure 2D (Bcl10, Erap1, Kcnn4, Pawr, Prkd2, Ptprj, Tap1, Tapbp, Tapbpl, Usp9x). Hierarchically clustering was applied on cell line and log2 normalized counts for the DEGs using an one minus pearson correlation.

In vivo Tumor Experiments
C57BL/6 mice were used at 8-12 week of age and bred at the animal facilities of Amsterdam UMC/VU University Medical Center (VUmc). An equal distribution of female and male mouse was used for all in vivo experiments. Experiments were performed in accordance with national and international guidelines and regulations. Each mouse received 2 × 10 5 tumor cells in 100 µl PBS, which was subcutaneously injected in the flanks. Tumor measurements were performed three times per week in a double blinded manner. The total tumor volume was calculated according the formula 4/3 × π × abc (a = the radius of the width, b = the radius of the length and c = the average radius of width and length). Mice were sacrificed at day 13 or when the tumor reached a size of 2,000 mm 3 . The tumor, tumor draining lymph nodes and spleens were isolated and used for further experiments.

Tumor and Organ Dissociation
First, tumors were finely minced and enzymatically digested for 25 min at 37 • C in RPMI containing 1 mg/mL Collagenase type 4 (Worthington), 30 units/mL DNase I type II (Sigma-Aldrich) and 100 µg/mL hyaluronidase type V (Sigma-Aldrich). Lymph nodes and spleens were finely minced and digested for 10 min at 37 • C with 2 U/mL Liberase TM (Roche) containing 30 units/mL DNase I type II. Subsequently, cell suspensions were passed through a 70 µm cell strainer and washed once with RPMI supplemented with 10% FCS, 1% penicillin, 1% streptomycin and 1% glutamax. Spleen digestions were subjected to ACK lysis and, together with the lymph node digestions, washed twice before use in subsequent experiments. To enrich for the lymphocytes and to remove dead cell debris from tumor digestions, the tumor digestion mixture was loaded on a Ficoll gradient. The interface was collected, washed twice and used for subsequent flow cytometric analysis.

Flow Cytometric Analysis
To block Fc receptors, cells were pretreated with 2.4G2 (anti-CD16/32) for 10 min at RT. Cell viability was measured using a fixable viability dye (FVD, Zombie NIR, Biolegend). Cells were stained for 20 min at RT using the following cell surface markers: To stain for Foxp3, cells were fixed and permeabilized (Foxp3 Transcription Factor Staining buffer set, eBioscience) and incubated with anti-Foxp3-PE antibody (150D, Biolegend) for 20 min at RT. Additionally, tumor-infiltrating lymphocytes were restimulated in vitro with 100 ng/mL PMA and 500 ng/mL ionomycim and Brefeldin A was added to block cytokine secretion. After 5 h, cells were incubated with 2.4G2 and fixable viability dye (FVD, Zombie NIR) as described above and subsequently fixed and permeabilized using BD Cytofix/Cytoperm TM (BD Biosciences) according to the manufacturer's protocol and stained for 30 min at 4 • C with anti-IFNγ-APC (eBiosciences). All samples were acquired on BD LSRFortessa TM X-20. Data was analyzed with FlowJo v10.

Loss of C1galt1c1 Increases Tn Antigen Expression
In this study we aimed to dissect the effect of high Tn antigen expression on colorectal cancer (CRC) tumor growth and the immune cell composition at the tumor site. Therefore, we first analyzed the steady state levels of Tn antigen on the mouse CRC cell line MC38 (MC38 wild type, MC38-WT), using the GalNAc-specific snail lectin Helix Pomatia agglutinin (HPA), and observed that our MC38-WT cells displayed only intermediate Tn antigen levels ( Figure 1A). Normally, the T-synthase enzyme and its chaperone COSMC further elongate the Tn antigen with galactose, thus forming the T antigen (Galβ1-3GalNAcα-Ser/Thr) epitope (28) [see Figure 1B for a schematic representation of the O-glycosylation pathway, adapted from Cornelissen and Van Vliet (29)]. Indeed, CRISPR/Cas9-mediated knock out of the Cosmc gene (C1galt1c1) abolished T-synthase enzymatic activity ( Figure 1C). Furthermore, C1galt1c1 gene mutations were validated using the Surveyor assay (Supplementary Figure 1A). Consistently, Tn antigen levels were elevated after C1galt1c1 gene knock out (MC38-Tn high ), as confirmed by the increased staining of HPA ( Figure 1D) and an anti-Tn antibody (5F4, Figure 1E), as well as the stronger binding of the mouse macrophage galactose-type lectin 2 (mMGL-2-Fc, Figure 1F), known for its specificity for the Tn antigen (30). Indeed, expression of T antigen, measured using the T antigen-recognizing lectin peanut agglutinin (PNA), was strongly reduced in the MC38-Tn high cells (Supplementary Figure 1B). Moreover, mouse MGL-1 (mMGL-1-Fc), recognizing the Lewis X and A structures (30), only showed negligible binding to MC38-MOCK and MC38-Tn high cells (Supplementary Figure 1C). Mutating the C1galt1c1 gene did not alter the viability or proliferation of MC38-Tn high cells (Supplementary Figure 1D). Thus, CRISPR/Cas9-mediated gene knock out of C1galt1c1 results in elevated expression of the Tn antigen epitope in MC38 without affecting its intrinsic proliferative capacity.

Knockout of C1galt1c1 in Tumor Cells Is Associated With a Decrease in Antigen Presentation and Cytotoxic T Cell Activation
To investigate whether increased expression of truncated O-glycans leads to transcriptional changes in the glycoengineered MC38 cells, we performed RNA sequencing (RNAseq) analysis in MC38-WT, MC38-MOCK and MC38-Tn high cells. Comparative analysis revealed that the expression of 1,348 genes was specifically affected in MC38-Tn high cells, of which 707 genes were increased in expression and 641 genes were downregulated    ( Figures 2A,B and Supplementary Tables 1-3). The effect of increased expression of truncated O-glycans by tumor cells on specific biological processes was assessed by subjecting the 1348 DEGs in MC38-Tn high cells to GO term enrichment analysis ( Figure 2C and Supplementary Table 4). We found evidence for effects of truncated O-glycans on MAPK signaling (GO group 103), cell migration, and blood vessel development (GO group 70, 82, 97) ( Figure 2C, Supplementary Figures 2, 3 and Table 4), which corroborates previous reports (10,31,32). GO term enrichment analysis, furthermore, suggested an effect of truncated O-glycan expression on immunerelated genes, involved in antigen processing and presentation, defense against viral infections, αβ T cell responses, and cytotoxic T cell responses (GO group 72, 85, 91, 93) ( Figure 2C, Table 1 and Supplementary Table 4). Most of the genes involved in antigen presentation and T cell activation were suppressed in MC38-Tn high cells ( Figure 2D and Supplementary Tables 3-5), including the expression of well-known genes such as B2m, Tap1, Tap2, Tapbp, H2-T23, H2-D1, as well as the expression of important transcription factors, such as Irf1, Batf, Foxp1, Eomes, Stat3. Stat6 was one the prominent genes upregulated in our MC38Tn high cells ( Figure 2D). A reduction in the MHC loading machinery and MHC class I molecules in tumor cells is known to restrict immune cell infiltration and to hamper anti-tumor immunity through the blockade of CD8 + T cell responses (33). STAT molecules play diverse pro-and anti-tumorigenic roles in cancer (34), however, in CRC high STAT3 expression is positively correlated to a better overall survival (35). Active STAT6 signaling seems to promote Thelper 2 cytokine profiles in CRC and to confer a pro-metastatic and anti-apoptotic tumor phenotype (36). Based on our RNAseq analysis, we hypothesized that truncated O-glycans and Tn antigen may have an immunomodulatory role in CRC.

Tn Antigen Drives Tumor Growth Specifically at Later Stages of Tumor Development
We first explored the impact of high Tn antigen expression on tumor growth through subcutaneous injection of the MC38-MOCK and MC38-Tn high cell lines in C57Bl/6 mice. Interestingly, and in contrast to a sialic acid knockout of MC38 (37), the in vivo growth rates were similar between MC38-MOCK and MC38-Tn high tumors at day 13 of tumor development (Figures 3A,B). However, from day 23 onward MC38-Tn high tumors started to display significantly faster growth compared to the MC38-MOCK tumors (Figures 3C,D). No differences in tumor growth were observed between male and female mice (data not shown). Thus, high Tn antigen expression seems to drive tumor growth particularly at later stages of colorectal cancer development.

MC38-Tn high Tumors Are Characterized by Lower Immune Cell Infiltration and Higher Levels of Myeloid-Derived Suppressor Cells
Next, we examined the influence of tumor-associated Tn antigen on the immune cell composition within the tumor microenvironment. We inoculated mice with our MC38-MOCK and MC38-Tn high and harvested the tumors when they reached a size of 2000 mm 3 . Strikingly, compared to MC38-MOCK tumors, MC38-Tn high tumors displayed reduced infiltration of viable immune cells (FVD − CD45 + ) ( Figure 4A). As carbohydrate binding lectin receptors are mainly expressed by antigen presenting cells (38), we first analyzed the myeloid immune cell compartment, including professional antigen presenting cells (APCs), such as dendritic cells (DCs) and macrophages, and myeloid-derived suppressor cells (MDSCs). APCs are known to internalize and present antigens to T cells, while MDSCs contribute to an immune suppressive tumor microenvironment (39). APC (MHC-II + CD11c + ) frequencies were equal between MC38-Tn high and MC38-MOCK tumors (Figure 4B), but were reduced in the tumor draining lymph node (TDLN, Figure 4D) and the spleen ( Figure 4E). Interestingly, the reduced DC frequencies in the TDLN were confined to the lymph noderesident DC (CD11c ++ MHCII + ) and not attributable to the migratory DC subset CD11c + MHCII ++ ) ( Figure 4F). Moreover, MDSC frequencies (CD11b + Gr-1 + ) (40) were significantly higher in MC38-Tn high tumors compared to MC38-MOCK tumors ( Figure 4C).

MC38-Tn high Tumors Display Reduced Cytotoxic CD8 + T Cell Frequencies
Resident DC in the TDLN are known to participate in the priming of tumor-specific CD8 + T cells through the capture  of antigens from migratory DCs (41), suggesting that also adaptive T cell immunity could be affected in the MC38-Tn high tumors. Moreover, high CD8 + T cell frequencies within the tumor microenvironment strongly correlate with a good prognosis in many cancer types (42), including CRC, supporting the substantial role of CD8 + T cells in combating cancer. Thus, we next analyzed the lymphoid compartment in our MC38-MOCK and MC38-Tn high tumors. Indeed, compared to MC38-MOCK tumors, CD8 + T cell frequencies were decreased in MC38-Tn high tumors ( Figure 5A). Moreover, CD4 + T cell frequencies tended to be lower in MC38-Tn high tumors, however, due to high variation within both groups, the differences observed were not significant ( Figure 5B). Although reduced APC frequencies were found in TDLN of mice bearing MC38-Tn high tumors (Figure 4D), T cell frequencies in TDLN were indistinguishable in the MC38-MOCK and MC38-Tn high groups (Supplementary Figure 4).
MDSCs are known to promote regulatory T cells expansion within the tumor microenvironment (43,44), yet no increase in regulatory T cell frequencies was observed in MC38-Tn high tumors ( Figure 5C) or tumor-draining lymph nodes and spleen (Supplementary Figure 4). We next analyzed the expression of immune checkpoint markers on the tumor-infiltrating CD8 + T cell population. Although equal percentages of PD-1 + CD8 + T cells were found in MC38-MOCK and MC38-Tn high tumors, the percentage of TIM-3 + CD8 + T cells was significantly lower in the MC38-Tn high tumors (Supplementary Figure 5A). Based on these data we cannot ascertain whether the CD8 + T cells in the MC38-Tn high tumors are less exhausted or less activated, especially as the percentage of IFNγ-producing cells was not different upon in vitro PMA/Ionmycin restimulation of tumor-infiltrating lymphocytes from MC38-MOCK and MC38-Tn high tumors (Supplementary Figure 5B).
In conclusion, MC38-Tn high tumors displayed increased tumor growth at later stages of tumor development, which was characterized by enhanced levels of MDSCs and decreased levels of CD8 + T cell infiltration. Together these results suggest that MC38-Tn high tumors are able to modulate the immune landscape within the tumor microenvironment, which might contribute to the enhanced tumor growth in vivo.

DISCUSSION
An aberrant glycosylation profile, including high Tn antigen expression, is a key feature of epithelial cancer types. Tumorassociated Tn antigen is found in 81-95% of CRC cases, independent of the histological subtype and differentiation status of the tumor, while it is not expressed by the healthy gut mucosa (10,45,46). Its expression has been correlated to bad prognosis and increased metastasis (5,6), however, the effect of Tn antigen on the anti-tumor immune response has not been investigated in detail. In this study, we explored in a colorectal cancer mouse model how Tn antigen influences tumor growth and the immune cell composition in the tumor microenvironment. We show that knockout of the C1galt1c1/Cosmc gene alters the cellintrinsic properties of tumor cells, affecting genes involved in MAPK signaling, cell migration and angiogenesis and immune regulation (Figure 2). Our results corroborate the work by Radhakrishnan et al., who also observed a downregulation of immune-related genes in COSMC knockouts of human keratinocytes (12). Finally, we provide evidence that enhanced Tn antigen expression coincided with enhanced tumor growth, especially at later stages of tumor development (Figure 3) and an alteration of the immune landscape within the tumor microenvironment (Figures 4, 5).
Tn antigen is the initial step of the O-glycosylation pathway and it can be further elongated in three distinct ways; through the addition of α2-6 sialic acid forming sialyl-Tn (sTn) antigen, with a galactose (forming the T antigen, Galβ1-3GalNacα-Ser/Thr) or a N-acetylglucosamine (core 3). Core 3 O-glycan structures are essential for intestinal mucus barrier function (47) and are responsible for colonic disease resistance (48). Also, the glycosyltransferase responsible for core 3 synthesis is generally downregulated during CRC progression, which actually accelerates CRC development (48,49). Together, this implies an antagonizing role of core 3 in tumor growth, hence it is not likely that core 3 is responsible for the observed increase in MC38-Tn high tumor growth. Sialyl-Tn (sTn) antigen has been described to support tumor progression as well (50) and could thus be accountable for the enhanced tumor growth of MC38-Tn high cells in vivo. However, flow cytometric analysis of α2-6 sialic acids or sTn expression, ruled out the presence of α2-6 sialic acids in any of the cell lines used (data not shown), indicating that de novo Tn antigen expression was not sialylated in our MC38-Tn high cells. Nevertheless, whether the effects we observed are due to an increase in Tn antigen or to a loss of the T antigen structure or longer O-glycans remain to be defined. In literature, conflicting reports exist on the influence of Tn or T antigen on tumor development. Overexpression of Tn antigen has been shown to promote tumor growth in CRC (11,51) and pancreatic cancer (13). Moreover, in some models the mere abrogation of T-synthase/COSMC is sufficient to induce oncogenic features and to augment tumorigenesis of otherwise healthy cells (12,52). In contrast, enhanced T antigen expression also correlates to bad prognosis and oncogenesis in, for instance, breast (31,53,54) and head and neck cancer (55). Strikingly, although Du et al. observed reduced breast cancer growth upon COSMC disruption, they also noticed an inhibition of the MAPK pathway similar to our findings (31) (Supplementary Figure 3). Thus, both Tn antigen and T antigen affect tumor progression and discrepancies might be explained by the tumor (sub)type or cell line investigated. Thus, whether Tn and T antigen work synergistically or whether one dominates over the other, remains to be investigated and is most likely tumor type-dependent.
The prominent role of anti-tumor immunity in CRC is clearly illustrated by the "Immunoscore, " a validated prognostic classification of CRC tumors according to their intratumoral density of memory and cytotoxic CD8 + T cells, superior to the classical TNM staging (56). Indeed, we also observed a decreased infiltration of CD8 + T cells in our MC38-Tn high tumors, highlighting the critical role that these cells play in CRC progression. Although very little is known on the in vivo immune response to Tn antigen-positive tumors, in vitro work has already provided some clues. We postulate that the Tn antigen-dependent immune effects might be mediated through the interaction with lectin receptors expressed on immune cells. One prime candidate in this respect would be the Macrophage Galactose-type Lectin (MGL, CD301). MGL is expressed by (tolerogenic) dendritic cells (DCs) and macrophages (57) and has a unique preference for tumor-associated Tn and sTn antigen (58-60), through which MGL is able to distinguish between tumor-associated carbohydrate antigens and healthy tissue (61). Upon ligand binding, MGL signaling converges with Tolllike receptor-induced pathways, which increases the secretion of the anti-inflammatory cytokine IL-10 by human DCs and promotes the differentiation of human Tr1 cells by these DCs (62,63). Moreover, MGL-mediated engagement of CD45 on effector T cells inhibits T cell cytokine secretion and can even induce T cell death (64), suggesting that tumor-associated Tn antigen might dampen adaptive immune responses via the MGL receptor at multiple levels to support tumor growth. Indeed, MGL ligand expression has been associated with lower survival in late stage CRC (65) and poor survival and distant metastasis in cervical squamous cell and adenosquamous carcinoma (66). Nevertheless, we cannot rule out that T and sialylated T antigen may also have immunomodulatory properties within the tumor microenvironment. COSMC KO of human breast cancer cells actually sensitized tumor cells to natural killer or cytotoxic CD8 + T cell killing (67), while sialyl-T was shown to promote the differentiation of tumor-associated macrophages through the engagement of inhibitory sialic acid-binding receptor Siglec-9, respectively (68). Clearly, more work is needed to elucidate the individual roles of truncated tumor-associated O-glycans on the anti-tumor immune response.
Interestingly, the association with MGL ligand expression was solely observed in stage III CRC and not in stage II CRC patients, which fits with our findings that Tn antigen augments tumor growth mainly from day 23 onward. Our data thus also implies that Tn antigen supports tumor progression and dampens anti-tumor immunity specifically during later stages of tumor development. Nevertheless, CRC is a very heterogeneous disease, comprising of four consensus molecular subtypes with different biological characteristics (69) and as yet, it is unclear how aberrant tumor glycosylation receptor impacts each CRC subtype. Certain glyco-gene signatures may be able identify patients groups with poor prognosis (70); however, the high incidence of Tn antigen (up to 80% in CRC) suggests that glycosylation-based immune modulation might be subtypeindependent and should thus be regarded as a novel immune checkpoint across the whole CRC spectrum (14).

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: Sequence Read Archive (SRA) Gene Expression Omnibus (accession number GSE143700).

ETHICS STATEMENT
The animal study was reviewed and approved by VU University Animal Welfare Commission, VU University, Amsterdam, Netherlands.

AUTHOR CONTRIBUTIONS
LC designed and performed experiments, analyzed data, and wrote the manuscript. AB, AZ, and JH performed the experiments. AZ analyzed RNA sequencing data. LK performed subcutaneous cell injections and mouse experiments. TO'T performed cell sorting experiments. YK and SV conceived and coordinated the study. SV designed and supervised the study and wrote the manuscript. All authors contributed to reviewing and/or revising the manuscript.