Abstract
Tumor growth is accompanied with dramatic changes in the cellular glycome, such as the aberrant expression of complex branched N-glycans. However, the role of this protumoral N-glycan in immune evasion and whether its removal contributes to enhancement of immune recognition and to unleashing an antitumor immune response remain elusive. We demonstrated that branched N-glycans are used by colorectal cancer cells to escape immune recognition, instructing the creation of immunosuppressive networks through inhibition of IFNγ. The removal of this “glycan-mask” exposed immunogenic mannose glycans that potentiated immune recognition by DC-SIGN–expressing immune cells, resulting in an effective antitumor immune response. We revealed a glycoimmune checkpoint in colorectal cancer, highlighting the therapeutic efficacy of its deglycosylation to potentiate immune recognition and, thus, improving cancer immunotherapy.
Introduction
Colorectal cancer is a global and lethal disease, remaining the third most commonly diagnosed cancer in males and the second in females worldwide (1). Despite the clinical success of cancer immunotherapies, the majority of patients with metastatic colorectal cancer fail to respond or develop resistance to therapy, which constitutes a major clinical concern (2). In fact, only 5% of the patients with metastatic colorectal cancer diagnosed with mismatch repair deficiency (MMR-d)/microsatellite instability-high (MSI-H) respond to immunotherapies (3, 4), indicating a critical gap in our understanding of how these tumors regulate immune responses (5). In this era of cancer immunotherapy, there is an urgent clinical need to understand the dynamic cross-talk between tumor and immune cells, aiming to unlock new targets for the development of more effective cancer immunotherapeutic strategies.
Cancer progression is guided by the selective pressure of the immune system on malignant cells, resulting in the selection of tumor cell variants that are more prone to survive/escape in an immunocompetent host. Some mechanisms have been described in past studies showing the mechanisms used by tumor cells to elude immunosurveillance, including the expression of immune inhibitory checkpoints. This has boosted the development of immunotherapies for CTLA-4 and PD-1/PD-L1 blockade, revolutionizing cancer treatment (6–9). Nevertheless, only a small proportion of patients with cancer ultimately benefit from immunotherapy, which supports the hypothesis that cancer cells evade immune recognition through other immune escape strategies.
All cells are covered with a dense and complex coat of glycans (the glycocalyx), which constitutes a major biological interface between cells and their environment. Glycosylation is the enzymatic process responsible for the attachment of glycans to proteins or lipids in essentially all cells. The cellular glycosylation signature is altered in cancer cells, and the aberrant expression of glycans by tumor cells regulates each pathophysiologic step of cancer development and progression, including tumor cell dissociation and invasion, signaling, angiogenesis, and metastasis (10, 11). Particularly, one of the most widely occurred tumor-associated glycan structure is the complex β1,6-GlcNAc–branched N-glycans (12). The aberrant overexpression of this structure, catalyzed by the N-acetylglucosaminyltransferase V (GnT-V) enzyme, has been consistently shown to induce promalignant, proinvasive, and prometastatic cancer phenotypes through mechanistic interference with tumor cell adhesion, and is associated with poor survival rates of patients with gastrointestinal cancer (13–17). The cellular stabilization of PD-L1 is also known to be modulated by complex N-glycosylation (18, 19). The importance of glycans' modifications of cancer proteins and cancer (neo)antigens on the regulation and instruction of the surrounding immune response is now starting to be explored (20). However, whether and how the abnormal expression of the tumor-associated complex branched N-glycans alters the way that the immune system recognizes malignant transformation and cancer progression remains unknown.
Immune cells express a variety of specific glycan–binding receptors called C-type lectins [such as dendritic cell (DC)-specific intercellular adhesion molecular-3-grabbing nonintegrin (DC-SIGN); mannose receptor (MR)], galectins, and Siglecs, which are able to recognize alterations in the glycan signature of tumor cells, instructing either immunostimulatory or immunoinhibitory pathways (21, 22). In this study, we hypothesized that the abnormal cancer glycoprofile characterized by the overexpression of complex branched N-glycans was able to impact cancer immunoediting, contributing to the creation of immunosuppressive networks associated with disease progression. From a clinical point of view, we also postulated that the removal of branching N-glycosylation on colorectal cancer cells may expose relevant immunogenic (glyco)epitopes, constituting a glycoengineered strategy to enhance immune recognition and immunosurveillance, promoting antitumor immunity. Whether and how this altered host glycoprofile perturbed the gut microbiome composition associated with the modulation of tumor immune responses in colorectal cancer were also investigated.
Materials and Methods
Mice
All mouse procedures were approved by the i3S ethics committee for animal experimentation under Portuguese regulations. Mice were housed at the animal facility of the Institute for Research and Innovation in Health, University of Porto (i3S, Porto, Portugal). C57BL/6 wild-type (WT; acquired at The Jackson Laboratory), Mgat5−/− (kindly provided by Michael Pierce, University of Georgia, Athens, GA), and Rag2−/−Il2rg−/− (kindly provided by Prof. James Di Santo, Institut Pasteur, Paris, France) mice were used for tumorigenicity assays. The ApcMin/+ (acquired at The Jackson Laboratory) and VCMsh2LoxP/LoxP (kindly provided by Prof. Winfried Edelmann, Albert Einstein College of Medicine, New York, NY) mice were crossed (first generation) with Mgat5−/− mice (ApcMin/+ Mgat5−/− and VCMsh2LoxP/LoxP Mgat5−/−) to evaluate spontaneous tumor growth and tumor immune infiltrate, comparing with ApcMin/+ and VCMsh2LoxP/LoxP Mgat5 WT background mice.
Cell culture
The MKN45 gastric cancer cell line transfected with an empty vector (MKN45 Mock) or with GnT-V cDNA vector (MKN45 T5) were used in coculture assays (23) (kindly provided by Naoyuki Taniguchi, Osaka University Medical School/Graduate School of Medicine, B1, 2–2 Yamadaoka, Suita, Osaka, Japan). These cells were cultured in RPMI1640 GlutaMAX culture medium (Gibco), supplemented with 10% FBS (Biowest) and penicillin/streptomycin (100 U/mL; Gibco), under selection pressure of G418 (500 μg/mL; InvivoGen) at 37°C in 5% CO2. All the cell lines used were tested for the presence of Mycoplasma and maintained in culture during approximately 3 passages before using in the experiments. Authentication of cell lines (genotyping) was performed before the experiments.
In in vivo assays, the murine colon adenocarcinoma cell line, MC38 (kindly provide by our collaborator and coauthor Julio C. de-Freitas, junior), was manipulated using CRISPR/Cas9 technique to knockout the Mgat5 gene (MC38 T5KO), as described below. The MC38 WT and MC38 T5KO cells were maintained in DMEM culture medium (Gibco), supplemented with 10% FBS and penicillin/streptomycin (100 U/mL) at 37°C in 5% CO2.
CRISPR/Cas9, insertion–deletion validation, and functional assays
To obtain the cell line MC38 KO to Mgat5, 3 different single-guide RNA (sgRNA) were used: AAGTTGTCCTCTCAGAAGCTGGG; TCAGAAGCTGGGCTTTTTCCTGG and CTAGCAATGTACCATTCCCTTGG (3′-5′) (all from IDT). Briefly, these sgRNA were cloned in pSpCas9(BB)-2A-GFP (#48138, AddGene) and transfected to cell line, using Lipofectamine 2000 (Invitrogen) according to manufacturer's recommended protocol. The clonal cell lines were isolated by flow cytometry sorting (BD FACSAria II, BD Biosciences) 24 hours after transfection (around 10% of the cells were positive to GFP). One positive cell per well was plated in 96-well plates, and the clonal cell line was expanded for 2–3 weeks in DMEM supplemented with 10% FBS and penicillin/streptomycin (100 U/mL) at 37°C in 5% CO2. The genomic DNA from the clonal cell lines was extract with Quick Extract reagent (Epicentre) and the Mgat5 gene was amplified by PCR using the primers forward: CAACCTGCACCATGAACTTGC and reverse: ATGTAGTTCCGTGACCGTCT. The PCR products were sequenced by Sanger method and INDELs presence was inspect by DSDecodeM online software (developed in ref. 24).
Three different clones transfected from the same sgRNA and with proved indel, were inspected for the glycoprofile to validate the functional KO of the enzyme. The cells were labeled with different lectins (1:1,000; Vector Laboratories)—Fluorescein-conjugated Phaseolus vulgaris leucoagglutinin (a lectin that recognizes the beta1,6-GlcNAc arm of branched complex-type N-glycans; L-PHA-Fluorescein), Galanthus nivalis agglutinin (a lectin that recognizes the terminal alpha-1,3 mannose residues; GNA-Fluorescein), and biotinylated Concanavalin A (a lectin that recognizes alpha-mannose residues; ConA-biotin)—for 15 minutes at 4°C, after the staining with Fixable Viability Dye (FVD)–APC-eFluor780. For the staining with biotinylated lectins, an additional incubation was performed with streptavidin-PE (BioLegend). Samples were measured on a FACSCanto II machine and the data was analyzed using FlowJo software v10.
The absence of β1,6-GlcNAc branched N-glycans was also confirmed by immunofluorescence. Samples were fixed with 4% formaldehyde and blocked with 5% BSA in PBS-Tween 0.1% for 1 hour. After incubation for 1 hour with L-PHA-Fluorescein (1:250; Vector Laboratories) or GNA- Fluorescein (1:500; Vector Laboratories) diluted in the blocking solution, nuclei were stained with DAPI and cell fluorescence was visualized using fluorescence microscopy Zeiss Axio Imager Z1 (Carl Zeiss, Germany). Analysis was performed in Axiovision 4.9 software (Carl Zeiss, Germany).
For the proliferation assay, 5 × 104 MC38 WT and T5KO clonal cells were labeled with 5-(and-6)-carboxyfluorescein diacetate succinimidyl ester (CFSE), using CellTrace CFSE Cell Proliferation Kit (Invitrogen). Cell were washed and maintained in culture at 37°C in 5% CO2. After 60 and 96 hours of culture, the fluorescence of cells was measured on a FACSCanto II machine and data analyzes using FlowJo v10 software.
Human samples
Formalin-fixed, paraffin-embedded (FFPE) tissue samples were collected in routinely medical exams or as part of patients' treatment. The cohort of clinical samples from colon carcinogenesis cascade were characterized by biopsies of normal colon (n = 5), colonic low-grade dysplasia (LGD; n = 13), and colonic high-grade dysplasia (HGD; n = 7) from Hospital Center of Porto (CHP, Porto, Portugal), and colonic dysplasia (n = 57) and colon adenocarcinoma (n = 55; stage I–III) from Portuguese Oncology Institute of Porto (IPO Porto, Porto, Portugal). The samples representing the gastric carcinogenesis cascade: normal gastric mucosa (n = 5), gastritis (n = 5), and atrophy and/or intestinal metaplasia (n = 5), were assigned by CHP, whereas gastric adenocarcinomas (n = 5) were assigned by Hospital Center of São João (CHSJ, Porto, Portugal). Whole blood from healthy donors, firstly used for peripheral blood mononuclear cells (PBMC) isolation, were kindly donated by healthy volunteers from a bank of blood donors. Buffy coats from healthy donors, used in CD14+ cells and T-cell isolation (described below), were obtained from Blood Bank of CHSJ (Porto, Portugal).
All samples were collected after written informed consents, according to directives 2004/23/EC and 2001/83/EC, which sets the standards of quality and safety for the donation, procurement, testing, processing, preservation, storage, and distribution of human samples in agreement with the World Medical Association Declaration of Helsinki for the ethical Principles for Medical Research Involving Human Subjects. The study was performed after approval by the institutional review board (IRB) of CHP and CHSJ.
Human PBMC isolation, monocyte-derived DC differentiation, naïve CD4+ T-cell isolation, and mouse bone marrow–derived DC isolation
To isolate PBMCs, 1 volume of Lymphoprep (StemCell Technologies) was used for 2 volumes of diluted whole blood (diluted 1:2 in PBS) and centrifuged for 30 minutes at 800 × g with the brake off. The interphase between the plasma/PBS and Lymphoprep, that contain PBMCs was then isolated. The cells were resuspended in RPMI supplemented with 10% FBS and penicillin/streptomycin (100 U/mL).
To isolate CD14+ monocytes, the PBMCs isolation was performed as before, but it used buffy coats (diluted 1:6 in PBS) from blood donors. From 1 × 108 PBMCs, CD14+ cells were purified using Human CD14 MACS Microbeads (Miltenyi Biotec) by magnetic separation with MS columns according to manufacturer's instructions (purity ≥ 95%). The enriched CD14+ cells were cultured in 6-well flat plates in RPMI supplemented with 10% FBS, penicillin/streptomycin (100 U/mL), and human recombinant proteins IL4 and GM-CSF (50 ng/mL; E.coli, PeproTech) at a concentration of 1 × 106 cells/mL to drive differentiation into monocyte-derived DCs (moDC). On the third day of culture, the medium was replaced, and the cells were in cultured for 3 more days, for a total of 6 days. At that time, monocytes were differentiated into immature moDCs and in half of the wells, lipopolysaccharide (LPS, 100 ng/mL; Sigma-Aldrich) was added to mature moDCs. Forty-eight hours later, the maturated moDCs were resuspended in X-VIVO 15 serum-free medium (Lonza) for posterior cocultures with cancer cells. The remanding PBMCs were frozen in freezing medium [90% FBS and 10% dimethyl sulfoxide (DMSO, PanReac AppliChem)] for later isolation of naïve CD4+ T cells.
Naïve CD4+ T cells were isolated from same donors' PBMCs used for moDCs. PBMCs were thawed in RPMI plus 10% FBS and penicillin/streptomycin (100 U/mL) and kept in culture overnight. After this period, naïve CD4+ T cells were enriched using the EasySep Human naïve CD4+ T Cell Isolation Kit II (StemCell Technologies) according to manufacturer's recommended protocol (purity ≥90%).
The bone marrow was extracted from the femur and tibias of C57BL/6 WT mice, and the isolated cells were maintained in culture in RPMI plus 10% FBS, penicillin/streptomycin (100 U/mL), and mouse recombinant proteins IL4 and GM-CSF (50 ng/mL; E. coli, PeproTech). On the third day of culture, the medium was replaced, and the cells were in cultured for 3 more days. The bone marrow–derived DCs (BMDC) were maturated with LPS (100 ng/mL) and used in cocultures with MC38 cells as described below.
IHC and lectin histochemistry
For lectin histochemistry, 3-μm sections of FFPE tissues from distinct stages of colon or gastric carcinogenesis were used. To block endogenous peroxidase, 3% H2O2 (Thermo Fisher Scientific) in methanol (Thermo Fisher Scientific) solution was used. Before staining, unspecific proteins were blocked with 10% BSA (Sigma-Aldrich). The sections were incubated for 1 hour with biotinylated lectin [13.3 μg/mL; L-PHA, GNA; Vector Laboratories]. To provide the detection, the sections were incubated with ABC HRP solution from the Vectastain ABC Kit (Vector Laboratories) followed by 3,3′-Diaminobenzidine tetrahydrochloride (0.67 mg/mL; DAB, Sigma-Aldrich) substrate.
Foxp3 and Tbet IHC was performed in sequential sections from the same samples. The slices were deparaffinized and hydrated, and antigen retrieval was performed using vapor heat during 40 minutes (in 1 mol/L citrate buffer, pH = 6). At this time, all steps before and after the staining were performed using solutions provided by the UltraVision Quanto Detection System HRP DAB kit (Thermo Fisher Scientific) according to the manufacturer's recommended protocol. The sections were incubated with primary anti-human anti-Foxp3 (6.7 μg/mL, 236A/E7, eBioscience) or anti-Tbet (5 μg/mL, eBio4b10, eBioscience) diluted in 5% BSA overnight at 4°C for colon samples or 2 hours at room temperature for gastric samples. The sections were contrasted with hematoxylin, dehydrated, and preserved with appropriated mounting medium (Entellan new, Merck Millipore) and coverslips. The results were analyzed by three independent observers, to establish an observational classification of the slides (Zeiss Axioskop 2 Microscope was used). The characterization of lectin histochemistry followed a semiquantitative classification (<25%, 25–50%, 50–75%, and >75%), whereas the IHC for Tbet and Foxp3 followed a qualitative classification (±, +, ++, +++) accordingly with the staining extension and intensity.
Real-time PCR
For all FFPE tissues of colon and gastric carcinogenesis, RNA was extracted from 10-μm sections using the RecoverAll Total Nucleic Acid Isolation Kit (Invitrogen) according to the manufacturer's recommended protocol. For colon adenocarcinomas and dysplasias, only the selected and identified lesion tissue (accordingly with the histopathologic classification performed by a pathologist) was processed. For cell lines, RNA of MKN45 or MC38 cell pellets were extracted with TRI reagent (Sigma-Aldrich) according to the manufacturer's instructions. Total RNA was quantified using the Nanodrop system, and cDNA synthesis was performed using SuperScript IV Reverse Transcriptase (Invitrogen) according to the manufacturer's recommended protocol. Real-time PCR was performed in 96-well reaction plates, and cDNA was amplified using the respective TaqMan assays: MGAT5 (Hs00159136), FOXP3 (Hs01085834), Mgat5 (Mm00455036), Mgat5b (Mm01252571), POMT1 (Hs01059558), POMT2 (Hs00203575), MGAT3 (Hs02379589), MAN2A1 (Hs01123597; all from Applied Biosystems) and 18S (Hs.PT.39a.22214856.g; IDT). Amplification data were acquired with 7500 Fast Real-time PCR system (Applied Biosystems). The mRNA expression of the genes of interest was normalized using the internal control 18S mRNA expression (ΔCt). For the MC38 cell line, RQ values were calculated by the equation 2−(ΔCt), whereas for each carcinogenesis and MKN45 cell line, RQ values were calculated by the equation 2−(ΔΔCt), where ΔΔCt represents the relative expression of ΔCt of the gene in comparison with mean of normal clinical samples or Mock cell line.
Distal small intestine sections from ApcMin/+Mgat5+/+ and ApcMin/+Mgat5−/− mice with or without lesions were digested and RNA was extracted using RNAqueous-Micro Total RNA Isolation Kit (Invitrogen) according to the manufacturer's recommended protocol. The total RNA was quantified in Nanodrop system and the cDNA synthesis was performed using SuperScript IV Reverse Transcriptase (Invitrogen) protocol.
The real-time PCR were performed in 96-well reaction plates, and cDNA was amplified using LightCycler 480 SYBR Green I Master (Roche) and respective primers (IDT): Ifng_Fw: TGGCTGTTTCTGGCTGTTACT, Ifng_Rv: GTTGCTGATGGCCTGATTGTC; Il17a_Fw: TACCTCAACCGTTCCACGTC, Il17a_Rv: TTCCCTCCGCATTGACACAG; Il10_Fw: TAACTGCACCCACTTCCCAG, Il10_Rv: AGGCTTGGCAACCCAAGTAA; Cd45/Ptprc_Fw: GGGTTGTTCTGTGCCTTGTT, Cd45/Ptprc_Rv: CTGGACGGACACAGTTAGCA; Gapdh_Fw: GAAGGTCGGTGTGAACGGAT, Gapdh_Rv: CTCGCTCCTGGAAGATGGTG. Amplification data were acquired with 7500 Fast Real-Time PCR System (Applied Biosystems). The mRNA expression of the genes was normalized using the internal control Gapdh mRNA levels (ΔCt). RQ values were calculated by the equation 2−(ΔΔCt), where ΔΔCt represents the relative expression of ΔCt of the gene in comparison with mean of the controls ApcMin/+Mgat5+/+ samples. A ratio of gene of cytokines and Cd45/Ptprc gene expression was performed to normalized to immune infiltrate.
Colorectal cancer and gastric carcinogenesis: Bioinformatics analysis
Differentially expressed genes in colorectal (n = 275) and gastric cancer (n = 408) versus adjacent normal tissue (n = 349 and n = 211, respectively) were evaluated using GEPIA2 (information and analysis details are included in http://gepia2.cancer-pku.cn/#about), a web server for large-scale expression profiling and interactive analysis that extracts data from The Cancer Genome Atlas (TCGA; https://www.cancer.gov/about-nci/organization/ccg/research/structural-genomics/tcga) and The Genotype-Tissue Expression (GTEx; https://gtexportal.org/home/index.html). Pairwise gene expression correlation analysis was performed with the TCGA data, using the GEPIA2 web server.
Cocultures
Initially, PBMCs were cocultured with MKN45 Mock or MKN45 T5 cells (1:1) during different time points, from 18–96 hours, to select the best time point (18 hours) to use in the subsequent experiments. Cocultures with PBMCs were done in 96-well U plates coated with anti-CD3 (0.1 μg/well, OKT3, eBioscience). The wells were washed, and 1 × 105 PBMCs were directly cocultured with 1 × 105 MKN45 Mock or MKN45 T5 cells in RPMI supplemented with 10% FBS, penicillin/streptomycin (100 U/mL) and anti-CD28 (0.1 μg/well, CD28.2, eBioscience) to induce T-cell activation. After 18 hours, supernatants were collected for cytokine analysis, and the cells were then labeled for CD4+ T-lymphocyte transcription factor analysis by flow cytometry, as described below. Three hours before, Brefeldin A (10 ng/mL, Sigma-Aldrich), Phorbol 12-myristate 13-acetate (PMA; 20 ng/mL, Sigma-Aldrich), and ionomycin (200 ng/mL, Sigma-Aldrich) were added to the wells designated for intracellular cytokine expression analysis by flow cytometry. Single cultures of PBMCs, MKN45 Mock, or MKN45 T5 cells were used as controls.
Similar to PBMC cocultures, immature or mature moDCs were cocultured with MKN45 Mock or MKN45 T5 (1:1) in X-VIVO 15 medium. After 18 hours, the supernatants were collected for cytokine analysis, and the moDCs were labeled to analyze their differentiation and activation by flow cytometry. The moDCs were preincubated with inhibitory anti–DC-SIGN (50 ng/mL, 120507, Invitrogen), a C-type lectin that recognizes high-mannose N-glycans and fucose residues, and anti-MR (50 ng/mL, 15–2, Invitrogen), a C-type lectin that recognizes high mannose N-linked glycans, for 1 hour, followed by the addition of MKN45 Mock or MKN45 T5 cells. In addition, transwell cocultures were also performed to determine contact dependency. The moDCs (1 × 105) were cultured in bottom of the plate, while tumor cells (1 × 105) were culture in the transwell (HTS transwell-96-well plate, 0.4-μm polycarbonate membrane, Corning). Supernatants were collected for cytokine analysis. Single cultures of moDCs, MKN45 Mock, or MKN45 T5 cells were used as controls.
For triple cocultures with tumor cells, moDCs and naïve CD4+ T cells, MKN45 Mock or MKN45 T5 (1:1:1) were cultured for 18 hours, and then of CD4+ T cells and anti-CD3 (0.1 μg/well, OKT3, eBioscience) were added in X-VIVO 15 medium. After 96 hours, the supernatants were collected, and the cells were stained against transcription factors and intracytoplasmic cytokines, as done before in PBMCs.
The BMDCs (1 × 105 cells) and MC38 WT or MC38 T5KO (1 × 105 cells) were also cocultured in X-VIVO 15 medium. After 18 hours, the supernatants were collected, and the released cytokines were assessed by ELISA as described below.
In vivo tumorigenicity assays
MC38 cells were treated with kifunensine (KF; 20 μmol/L; Sigma-Aldrich) for 48 hours and 2 × 105 MC38 WT or treated (MC38 KF) cells were subcutaneously inoculated into the right flank of 6- to 8-week-old C57BL/6 WT mice. After tumor becomes visible, KF (100 μL, 20 μmol/L) was injected directly in the tumor every 2 days to maintain KF efficiency. Similarly, MC38 WT and T5KO cells (2 × 105 cells) were subcutaneously inoculated into the right flank of WT and Rag2−/−Il2rg−/− C57BL/6 mice. Beginning day 7 postinoculation, tumor size was measured with a caliper every two days, and tumor volume (mm3) was monitored using the formula (length × width2)/2. Mice were euthanized when one of the tumors from one mouse reached a volume greater than 2,000 mm3 (human endpoint established). The collected tumors were divided in 2 pieces for explant ex vivo assays and to isolate and analyze immune cell infiltrates.
The tumors were manually minced, digested in a dissociation medium (RPMI1640 GlutaMAX medium (Gibco) with collagenase IV (1 mg/mL; Sigma-Aldrich) supplemented with 10% FBS, penicillin/streptomycin (100 U/mL), 1 mmol/L CaCl2 and 1 mmol/L MgCl2) for 45 minutes with agitation at 37°C, and subjected to two semiautomated dissociation steps, before and after enzymatic digestion, using the gentleMACS Dissociator (Miltenyi Biotec), program 02.01 and 03.01, respectively, of tumor dissociation, using gentleMACS M Tubes (Miltenyi Biotec). Then, cells were filtered through 70-μm cell strainers. The mononuclear cells were enriched by Lymphoprep gradient density centrifugation at 800 × g with no break for 30 minutes. The isolated cells were after incubated with ammonium chloride potassium (ACK) erythrocyte lysing buffer [from in-house made 10× buffer: 1.5 mol/L ammonium chloride (NH4Cl), 100 mmol/L potassium bicarbonate (KHCO3), 10 mmol/L EDTA; all from Sigma-Aldrich] for 3 minutes. The cells were then stained for flow cytometry as described below.
The other smaller piece of the tumor was placed in 300 μL of RPMI medium for 24 hours in a 48-well plate. The weight of tumor pieces was determined, and supernatants were stored at −80°C for cytokine analysis by ELISA.
Colorectal cancer spontaneous mouse models—ApcMin/+ and VCMsh2LoxP/LoxP
The ApcMin/+ mice crossed with Mgat5−/− mice (ApcMin/+Mgat5−/−) or WT to Mgat5 were euthanized at 140 days (based on ref. 25). The number and size of lesions per zone of small intestine (proximal, medial, and distal) and colon were evaluated using a caliper. To evaluate the tumor immune infiltrates, a portion of 3 cm of distal small intestine, region with more incidence of lesions, with several lesions (WT), and without or with small lesions (Mgat5−/−) were collected, dissociated as described for the tumorigenicity assays, and CD45+ cells were purified using Mouse CD45 MACS Microbeads (Miltenyi Biotec) by magnetic separation with MS columns according to manufacturer's instructions. The survival rate of mice was carried out until the same human endpoint established in-house appeared and when animals were euthanized. Kaplan–Meier survival plots were used to evaluate the survival differences between ApcMin/+Mgat5−/− and ApcMin/+Mgat5+/+ mice.
VCMsh2LoxP/LoxP mice are described to have a median survival around 12 months and develop up to four lesions in small intestine and colon (26), as opposed to the ApcMin/+ model that develops several lesions. After 330 days, VCMsh2LoxP/LoxPMgat5−/− or WT background mice were euthanized, and tumor burden was evaluated as described previously. The region with lesion (WT) and a portion of proximal small intestine (region with more incidence of lesions) without lesion (Mgat5−/−) were collected, dissociated, and CD45+ cells were isolated, as described above, for flow cytometry analysis.
Another portion was placed in 300 μL of RPMI medium for 24 hours in 48-well plates. The weight of tumor pieces was determined, and supernatants were stored at -80°C for later cytokine analysis by ELISA.
Flow cytometry
For lectin staining, performed in tumor cell lines and isolated tumor cells (1 × 106), the cells were incubated with conjugated lectins (Vector Laboratories): L-PHA-fluorescein, GNA-fluorescein, and ConA-biotin for 15 minutes after incubation with FVD–APC-eFluor780 (eBioscience). The biotinylated lectin was then incubated with streptavidin-PE (BioLegend) for 30 minutes.
PBMCs and triple cocultures were stained for transcription factors and intracellular cytokines. For transcription factors, dead cells were excluded by staining with FVD–APC-eFluor780 (eBioscience), followed by extracellular staining with anti-CD45-BV510 (HI30, BioLegend) and anti-CD4-eFluor450 (RPA-T4, eBioscience). For intracellular staining of transcription factors, fixation and permeabilization were performed using Foxp3/Transcription Factor Staining Buffer Set (eBioscience) according to manufacturer's instructions. After blocking with 2% mouse serum (obtained in-house from C57BL/6 mouse), anti-Tbet-PerCP-Cy5.5 (eBio4B10, eBioscience) was added.
Intracellular cytokine analysis was performed using Golgi blockade with Brefeldin A. Cells were collected, washed, and stained with FVD-eFluor780, followed by anti-CD45-BV510 (HI30, BioLegend) and anti-CD4-PercP-Cy5.5 (RPA-T4, BioLegend). Cells were washed and fixed with 2% formaldehyde (PanReac ApplieChem). Permeabilization was done with 0.5% saponin (Sigma-Aldrich), and staining with anti-IFNγ-APC (4S.B3) and anti-TNFα-AF488 (MAb11; all from eBioscience) was performed.
For cell surface markers analysis on moDCs, cells from cocultures were stained with anti-CD45-BV510 (HI30, BioLegend), anti-CD11c-FITC (BU15, Immunootools) or anti-CD11c-APC (BU15, eBioscience), anti-CD14-PE (MEM-18, Immunotools) or anti-CD14-PE (61D3, eBioscience), anti-HLA-DR-PeCy7 (G46–6, BD Pharmingen) or anti-HLA-DR-PeCy7 (L243, eBioscience), and anti-CD86-PE-Cy5 (IT2.2, eBioscience) after FVD staining and FcγR blocking (Fc Receptor binding inhibitor polyclonal antibody, eBioscience).
For tumorigenicity assays, single-cell suspensions enriched in mononuclear cells were stained with anti-CD45-FITC (30-F11), anti-CD3-eFluor 506 (17A2), anti-CD4-eFluor 450 (RM4–5), anti-CD8a-PE-Cy7 (53–6.7), anti-CD279-PE (J43), and anti-TCR γδ-APC (eBioGL3) to assess adaptive immune responses, or with anti-MHCII-Pe-Cy5 (M5/114.15.2), anti-CD24-PE (M1/69), anti-CD64-APC (X54–5/7.1), anti-CD11c-eFluor 450 (N418), anti-CD11b-eFluor 506 (M1/70), and anti-CD206-PE-Cy7 (MR6F3; all from eBioscience) to assess innate immune responses after staining with FVD and blocking with 2% rat serum (StemCell Technologies).
For tumors from spontaneous models, isolated CD45+ cells were stained with FVD-eFlour780 before the staining with anti-CD3-FITC (17A2, Invitrogen), anti-CD279-PE (J43, eBioscience), anti-CD4-PerCP-eFluor 710 (GK1.5, eBioscience), anti-CD8a-V500 (53–6.7, BD Horizon), and anti-CD25-PE-Cy7 (PC61.5, eBioscience). After fixating and permeabilizing the cells using the Foxp3/Transcription Factor Staining Buffer Set (eBioscience) according to manufacturer's instructions, cells were blocked with 2% rat serum and stained with anti-Foxp3-APC (FJK-16s, Invitrogen) and granzyme B (GZMB)-eFluor450 (NGZB, eBioscience). To determine the innate immune profile, the same protocol used in tumorigenicity assays was performed. Samples were acquired on a FACSCanto II machine (BD Biosciences) and the data were analyzed using FlowJo software v10. The gating strategy to assess to all cell populations is presented in the Supplementary Data.
ELISA and Cytometric Bead Array
Supernatants from tumor explants were collected and the concentrations of TNFα, IFNγ, IL1β, IL17A (mouse ELISA Ready-SET-Go! Kits from eBioscience) and IL10 (mouse DuoSet ELISA Kit from R&D systems) were measured by ELISA, according to the manufacturer's protocol and using as substrate 3,3′,5,5′-tetramethylbenzidine (TMB, eBioscience) and stop solution 2 N H2SO4. The absorbance was detected using a microplate reader (Biotek Instruments) at 450 nm and 570 nm, the concentration was determined using a standard curve, and the results were posteriorly normalized to the tumor explant dry weight (g).
For cocultures, cytokine concentrations were analyzed by flow cytometry using two different Cytometric Bead Arrays (CBA): the BD Cytometric Bead Array Human Th1/Th2/Th17 Kit and the BD Cytometric Bead Array Human Inflammatory Cytokine Kit (BD Biosciences), following manufacturer's instructions. The samples were measured on BD Accuri C6 Machine (BD Biosciences) using a specific template provide by BD Biosciences. The human TGFβ1 DuoSet ELISA (R&D Systems) was also used to assess the concentration of TGFβ in moDCs cocultures.
C-type lectin receptor–binding glycan assay
The production of the C-type lectin receptor (CLR)–Fc fusion proteins used in this experiment was performed as described previously (27, 28). SDS-PAGE with subsequent Coomassie stain, as well as Western blot analysis, were performed to check the identity and purity of the respective CLR-Fc fusion proteins. CLR-Fc concentrations were determined using the Micro BCA Protein Assay Kit (Thermo Fisher Scientific). The glycan domains were assessed by flow cytometry, where MKN45 Mock and MKN45 T5 cells were first stained with FVD–APC-eFluor780; eBioscience), followed by incubation with DC-SIGN-Fc (2 μg/mL) and MGL1-Fc (2 μg/mL) fusion proteins in lectin-binding buffer (50 mmol/L HEPES, 5 mmol/L MgCl2, and 5 mmol/L CaCl2) for 1 hour at 4°C. MGL-1, which recognizes galactose residues, was used because that the glycan motifs recognized in humans are similar to those recognized in mice, as described previously (29). After washing with lectin-binding buffer, the cells were stained with a PE-conjugated goat anti-human Fc (Dianova) for 25 minutes at 4°C. Samples were analyzed on a BD FACSCanto II machine (BD Biosciences), and the data were analyzed using FlowJo v10 software (FlowJo LLC).
Immunofluorescence
FFPE sections from small intestine of ApcMin/+Mgat5+/+ and ApcMin/+Mgat5−/− mice (n = 4/group) were deparaffinized and hydrated as before, and antigen retrieval was performed using vapor heat for 40 minutes (1 mol/L citrate buffer, pH = 6). The slices are blocked with 10% BSA and 20% rabbit normal serum (Vector Laboratories) for 1 hour and incubated with primary anti-IFNγ-APC (5 μg/mL, XMG1.2) or anti-IL17A-PE (5 μg/mL, eBio17B7) diluted in 5% BSA overnight. Nuclei were stained with DAPI, and cell fluorescence was visualized using the fluorescence microscope Zeiss Axio Imager Z1 (Carl Zeiss, Germany). Analysis was performed in Fiji software (ImageJ, NIH, Bethesda, MD), and qualitative classification (±, +, ++, +++) was used to score three different areas from the same tissue. The mean of staining for each was considered to classify the samples.
Microbiome and 16S bioinformatic analysis
Before euthanasia, the feces from ApcMin/+ Mgat5+/+ and ApcMin/+ Mgat5−/− were directly collected from mice into Eppendorf tubes and frozen immediately in liquid nitrogen. Bacterial amplicon sequence analysis targeted a 16S DNA gene fragment comprising the V3/V4 hypervariable regions (16S sense 5′-TACGGRAGGCAGCAG-3′ and antisense 5′-CTACCNGGGTATCTAAT-3′). 16S amplicon library preparation was performed according to an optimized and standardized protocol (Metabiote, GenoScreen). In brief, PCR of the 16S DNA was performed using 5 ng of genomic DNA according to the manufacturer's protocol (Metabiote), utilizing bar-coded primers (Metabiote MiSeq Primers) at 0.2 μmol/L final concentration. An annealing temperature of 50°C for 30 cycles was used. The Agencourt AMPure XP-PCR purification system (Beckman Coulter) was used to purify PCR products and following quantification according to the manufacturer's instructions, they were multiplexed at equal concentrations. 250 bp paired-end sequencing was performed on an Illumina MiSeq platform (Illumina) at GenoScreen (accession number PRJNA660736). Raw paired-end reads were subjected to the following processes: (i) trimming of low-quality bases from the 3′ end (quality <30, based on the Phred algorithm); (ii) paired-end read assembly using fast length adjustment of short reads to improve genome assemblies (30), requiring a minimum overlap of 30 bases and a 97% overlap identity; and (iii) removal of forward and reverse primers using CutAdapt (31), with no mismatches allowed in the primer sequences. Sequences without perfectly matching forward and reverse primers were eliminated.
The merged sequences were analyzed using Deblur in the Qiime 2 environment (version 2019.7), which allows for theoretical single-nucleotide resolution [although we continued to use the conventional operational taxonomic unit (OTU) nomenclature]. Taxonomic classification was performed using a classifier trained specifically on the primers (available through the Qiime2 pipeline (version 2019.7) using the Silva reference database (version 132; www.arb-silva.de/silva-license-information). Bacterial OTUs that could not be assigned at the Phylum-level were excluded. Alpha diversity was estimated by the Shannon diversity index or the number of observed OTUs. Differences in diversity between ApcMin/+Mgat5+/+ and ApcMin/+Mgat5−/− were tested using the Wilcoxon rank-sum test. Diversity values are expressed as median and range. Comparisons in taxonomic composition were made using the linear discriminant analysis with effect size (LEfSe) algorithm using recommended parameters. Statistical analysis and plotting were performed in R (version 3.6.1) using the phyloseq (version 1.28.0), ggplot2 (version 3.2.1), and ggpubr (version 0.2.3) packages. The single OTU assigned to the genus Enterococcus had the sequence (TAGGGAATCTTCGGCAATGGACGAAAGTCTGACCGAGCAACGCCGCGTGAGTGAAGAAGGTTTTCGGATCGTAAAACTCTGTTGTTAGAGAAGAACAAGGACGTTAGTAACTGAACGTCCCCTGACGGTATCTAACCAGAAAGCCACGGCTAACTACGTGCCAGCAGCCGCGGTAATACGTAGGTGGCAAGCGTTGTCCGGATTTATTGGGCGTAAAGCGAGCGCAGGCGGTTTCTTAAGTCTGATGTGAAAGCCCCCGGCTCAACCGGGGAGGGTCATTGGAAACTGGGAGACTTGAGTGCAGAAGAGGAGAGTGGAATTCCATGTGTAGCGGTGAAATGCGTAGATATATGGAGGAACACCAGTGGCGAAGGCGGCTCTCTGGTCTGTAACTGACGCT). This was submitted to nucleotide blast (megablast) against the 16S ribosomal RNA sequence database using standard settings. Raw sequence data are accessible in the Sequence Read Archive (accession number PRJNA660736).
Quantification and statistical analysis
Data were processed using GraphPad Prism 7.0 (GraphPad Software) or SPSS v26 Software (IBM). Flow cytometry analysis were collected with BD FACSDiva v8.0.3 Software (BD Biosciences) and analyzed with FlowJo v10 Software (FlowJo LLC). Means and SEM or SD are presented as averages and error bars as indicated in the figure legends. For in vitro studies, points represent biological replicates, from mean of technical duplicates/triplicates, and the bars represent the mean of different experiments ± SD. When described, the mean fluorescence intensity (MFI) or % of cells were normalized (fold change) to the mean of MKN45 Mock cocultures of each experiment. Unpaired t test was performed to compare two different groups. Ordinary one-way or two-way ANOVA was performed to compare more than two experimental groups. Tumor measurements of in vivo experiments are represented as tumor volume (mm3) and represented as mean tumor volumes of a group of mice ± SEM and a two-way ANOVA was performed. Survival differences between groups were analyzed by Kaplan–Meier estimator, performed by SPSS v26 Software (IBM). Multiple t tests were used to compared different immune populations, determined by flow cytometry, between the groups of mice. Statistical significance was considered at P ≤ 0.05, with (*), (#), (§), or ($) representing the following scheme: *, P < 0.05; **, P < 0.01; ***, P < 0.001; and ****, P < 0.0001.
Results
Complex branched N-glycans are upregulated during colorectal cancer: Correlation with Tregs
Using a well-characterized cohort of human clinical samples from colorectal carcinogenesis, we evaluated the correlation between the in situ expression of β1,6-GlcNAc–branched N-glycans and the local tumor immune response. We showed a clear increased expression of branched N-glycans, predominantly at the tumor cell surface, that were positive for L-PHA lectin, along colorectal carcinogenesis (Fig. 1A). This gradual increase of β1,6-GlcNAc–branched N-glycan expression occurred early, reaching the highest expression in high-grade dysplasia and colon adenocarcinoma (Fig. 1A). In colon adenocarcinoma stage, low expression of mannosylated N-glycans (positive to GNA lectin) were detected (Fig. 1A). This in situ glycoprofile was accompanied with an increase in Foxp3-expressing regulatory T cells (Tregs) and decreased expression of Tbet-expressing cells in the tumor microenvironment (Fig. 1A; Supplementary Fig. S1A). This colon cancer–associated glycoimmune phenotype suggests a potential relationship between branching N-glycans and the local tumor immune response.
This correlation was further validated at the transcriptional level (Fig. 1B–D), where a significant increase of MGAT5 mRNA expression (gene encoding GnT-V, mediate β1,6-GlcNAc–branched N-glycan) was observed from normal to dysplasia stages, with a slight decrease in colon adenocarcinomas (Fig. 1B). Concomitantly with the upregulation of MGAT5 expression, the FOXP3 expression also increased gradually from normal to dysplasia to colon adenocarcinoma (Fig. 1C). This further supported a correlation between MGAT5 and FOXP3 expression, both at protein/glycan and mRNA levels. The Pearson correlation coefficient, performed at normal, dysplasia, and colon adenocarcinoma stages, revealed a significant positive association between MGAT5 and FOXP3 expression (Fig. 1D). To further validate these results, we used data from TCGA and GTEx databases, and we observed a significant increase in MGAT5 and FOXP3 mRNA expression in this larger cohort of patients with colorectal cancer compared with the normal context (Fig. 1E). A significant positive correlation between MGAT5 mRNA expression and FOXP3 mRNA expression or a Treg signature (composed of ENTPD1, FOXP3, and CTLA4 genes) were also observed in this large colorectal cancer cohort (Fig. 1F).
The analysis of other glycogenes that participate in N-glycan branching pathway showed a trend increase of MAN2A1 (encoding α-mannosidase II) expression and a trend decrease in MGAT3 (encoding GnT-III glycosyltransferase) expression in colon cancer compared with normal colon (Supplementary Fig. S1B and S1C), which is in line with the upregulation of MGAT5-mediated branched N-glycans. Although without statistical significance, a decreased expression of POMT1/2 (encoding O-mannosyltransferase) was also observed in colon cancer compared with normal controls (Supplementary Fig. S1D), in accordance with a decreased expression of mannose glycans observed in colorectal cancer (Fig. 1A). This transcriptional profile further supported the observed overexpression of branched N-glycans and the downregulation of mannose-enriched glycans along colorectal carcinogenesis.
To validate these observations, we analyzed a gastric carcinogenesis cascade. We observed that during gastric carcinogenesis, branched N-glycan expression also gradually increased, early in premalignancy (gastric atrophy/metaplasia stage; Supplementary Fig. S1E). At the gastric adenocarcinoma stage, overexpression of branched N-glycans was also accompanied by an increased frequency of Foxp3+ cells and low frequency of Tbet-expressing cells compared with normal gastric mucosa (Supplementary Fig. S1E). Using data from TCGA, we also observed a significant increase of MGAT5 and FOXP3 mRNA expression in gastric cancer compared with normal stomach (data from GTEx), further supporting our data (Supplementary Fig. S1F). A significant positive correlation between MGAT5 mRNA expression and the Treg signature were also observed in the same gastric cancer cohort (Supplementary Fig. S1G). Taken together, these results suggest a positive correlation between the expression of β1,6-GlcNAc–branched N-glycans at the tumor cell surface (concomitantly with low mannose–enriched glycans) and an increased differentiation and/or recruitment of Foxp3+ Tregs associated with immunosuppression in the tumor microenvironment.
Branched N-glycans expression hampers immune recognition of tumor cells
To gain further insights on the impact of the overexpression of β1,6-GlcNAc–branched N-glycans by tumor cells in the modulation of the surrounding immune response, we performed short-term cocultures using the gastrointestinal cancer cell line MKN45, stably transfected with GnT-V enzyme (MKN45 T5) and overexpressing β1,6-GlcNAc–branched N-glycans (with reduced levels of mannosylated glycans; Supplementary Fig. S2A–B). Short-term cocultures of MKN45 T5 or Mock cells and PBMCs from healthy donors were performed for 18, 24, 48, and 72 hours. Immunophenotyping (gating strategy in Supplementary Fig. S2C) and cytokine release were analyzed. The most pronounced differences were observed at 18 hours, and therefore this time point was selected for further experiments (Supplementary Fig. S2D). We observed a significant decrease in the MFI of Tbet and IFNγ in CD4+ T cells when cancer cells overexpressed branched glycans, compared with Mock cells (Fig. 2A and B). We also detected decreased IFNγ in the supernatants of PBMCs and MKN45 T5 cell cocultures (Fig. 2C). We further observed that overexpression of branched N-glycans by MKN45 T5 cells is associated with decreased expression of HLA-ABC (MHC-I) in tumor cells, when compared with Mock cells, further supporting a role of branched N-glycans on immune escape (Fig. 2D). Higher expression of MHC-I was observed intracellularly in MKN45 T5 cells when compared with cell surface expression, suggesting an internalization of MHC-I when branched N-glycans were overexpressed in tumor cells (Fig. 2D).
Considering antigen-presenting cells (APC), such as DCs, express glycan-recognizing receptors, namely DC-SIGN and MR, we cocultured tumor cells with moDCs from healthy donors to further explore the cross-talk between tumor cells overexpressing branched N-glycans and immune cells. No differences in the expression of surface/activation markers CD86 and HLA-DR was observed in different cocultures (Supplementary Fig. S2E). However, analysis of the supernatants from MKN45 T5 cocultures showed a significantly decreased concentration of proinflammatory cytokines IL6 and IL8 compared with Mock cell cocultures (Fig. 2E). A trending increase of the inhibitory cytokine TGFβ was also observed when moDCs were cocultured with MKN45 T5 cells (Fig. 2E).
To test whether this immunosuppressive phenotype was mediated by specific glycan-binding proteins (GBP) expressed on moDCs, we analyzed the expression of the glycan-binding domains on MKN45 cells that are recognized by DC-SIGN and macrophage galactose type lectin-1 (MGL-1). We found that MKN45 Mock cells (with increased surface expression of mannose-enriched glycans and decreased expression of branched N-glycans; Supplementary Fig. S2B) displayed higher DC-SIGN–recognized glycan domains that significantly decreased with branched glycans overexpression (Fig. 2F). These results suggested that the specific recognition of mannose glycans by DC-SIGN exposed on Mock cells was significantly hampered when branched glycans were overexpressed in MKN45 T5 cells (Fig. 2F; Supplementary Fig. S2B).
To further validate this interplay, we inhibited the glycan-binding recognition of DC-SIGN and MR on MKN45 Mock and T5 cells (Fig. 2G). When Mock cells (display higher mannose-enriched glycans than T5 cells) were cocultured with moDCs, CD86 expression significantly decreased after treatment with inhibitory anti-MR or anti-DC-SIGN (Fig. 2G). This observation supports the role of these GBPs in the immunomodulation, whereby overexpression of branched N-glycans in T5 cells appeared to mask and hamper the recognition of mannose-enriched glycans by DC-SIGN (Fig. 2F and G), thus impairing immune activation. This contact-dependent mechanism between glycans expressed by tumor cells and GBP present on the DC surface was confirmed using transwell assays, which showed that in the absence of cell–cell contact, a decreased production of IL8 was observed (Supplementary Fig. S2F).
To further assess the consequence of DC glycans recognition on T-cell function, triple cocultures with tumor cells, moDCs, and autologous naïve CD4+ T cells were performed. CD4+ T cells exhibited decreased expression of Tbet, IFNγ, and TNFα in the cocultures with cells overexpressing complex branched N-glycans compared with Mock cells (Fig. 2H and I). Altogether, these evidences support the functional effects of branched N-glycans on tumor cells in the modulation of immunosuppressive networks through masking and hampering immune recognition by specific GBPs, together with an apparent internalization of MHC-I molecules, revealing a “sweet” escape strategy and immune checkpoint in cancer cells.
Removing branched N-glycans exposes immunogenic glycans on colorectal cancer cells
Considering the immunosuppressive profile of branched N-glycans expressed at tumor cell surface, we inhibited the synthesis of branching N-glycosylation using KF, a potent inhibitor of the α-mannosidase I enzyme that blocks the biosynthesis of branching N-glycosylation, generating high mannose glycoproteins. Treatment of MC38 murine colon adenocarcinoma cells with KF (MC38 KF) showed a significant overexpression of high-mannose N-glycans and decreased expression of β1,6-GlcNAc–branched N-glycans, revealed by the increased reactivity to GNA/ConA lectins and decreased L-PHA binding, respectively (Fig. 3A). The subcutaneous inoculation of MC38 KF cells in C57BL/6 WT immunocompetent mice resulted in a significant delay in tumor growth compared with vehicle-treated MC38 cells (Fig. 3B and C). These results demonstrated that the removal of branched N-glycans promoted the exposure of mannose residues associated with suppression of tumor growth.
To further investigate the role of branched versus nonbranched/mannose-enriched N-glycans in colorectal cancer immunoediting, the Mgat5 glycogene was silenced using CRISPR-Cas9 in MC38 cells. Among all clones obtained, three different MC38 clones were selected, namely clones 1, 2, and 3, and validated for the presence of insertion–deletions (INDEL) in Mgat5 gene. The three clones showed the presence of different INDELs on Mgat5 gene associated with a decrease in Mgat5 mRNA transcription (Supplementary Fig. S3A and S3B). The three clones also showed lower β1,6-GlcNAc–branched N-glycans at their cell surfaces, together with higher expression of high-mannose glycans compared with MC38 WT (Fig. 4A; Supplementary Fig. S3C and S3D). MC38 T5KO clone 2 was chosen to proceed on the basis of the lowest expression of L-PHA–recognizing β1,6-GlcNAc–branched N-glycans and the highest GNA-recognizing mannose glycans compared with the other clones (Supplementary Fig. S3C). Clone 2 showed no impact on Mgat5b mRNA transcription level, which could be associated with a compensatory mechanism for complex branched N-glycans biosynthesis, and no impact on cell proliferation was observed using CFSE MFI versus MC38 WT cells (Supplementary Fig. S3E and S3F).
Tumorigenicity assays were conducted using the selected MC38 T5KO clone 2. We found that this clone into immunocompetent WT mice resulted in a significant suppression of tumor growth compared with controls (MC38 WT; Fig. 4B and C), which was more evident compared with KF-treated cells (Fig. 3B and C) due to the stable and effective removal of the branched N-glycans upon Mgat5 gene KO. Only 1 of 11 mice in this group developed a very small tumor (Fig. 4C), demonstrating the effects of exposing the mannose structures (Fig. 4A) in the suppression of colorectal cancer development and progression.
To address the effects of this on cancer immunoediting and to determine how this MC38 T5KO glycoprofile modulated the surrounding immune responses, tumors were collected for immunophenotyping of immune infiltrates and evaluation of cytokines. We observed a slight increase in the percentage of infiltrating monocytes, macrophages, and DCs after removal of complex branched N-glycans (MC38 T5KO) compared with MC38 WT tumors (Fig. 4D). The innate immune response analysis is adapted from ref. 32 and described in Supplementary Fig. S4A. For adaptive immune responses (Supplementary Fig. S4B), MC38 T5KO tumors exhibited an increase in the number of CD3+ T cells, most likely due to the increase in γδ T cells, compared with MC38 WT tumors (Fig. 4E). Cytokine analysis of the tumor explants was also in line with an immunostimulatory phenotype, showing higher concentrations of TNFα released by MC38 T5KO tumors compared with controls (Fig. 4F). No major differences were detected in the concentrations of IFNγ and IL1β between the two conditions (Fig. 4F). Together, these observations, although not significant due to the limitation of tumor number availability for analysis, indicated that removal of β1,6-GlcNAc–branched N-glycans exposed mannose glycans that contributed to tipping the balance toward immunostimulation and antitumor immune attack. To overcome the aforementioned technical limitation and to further validate our results, MC38 WT and T5KO cells were cocultured with BMDCs, and cytokines were evaluated by ELISA. We observed that MC38 T5KO cells cocultured with BMDCs induced an increased production of the proinflammatory cytokine IL23 compared with controls (Fig. 4G). No major differences were detected in the concentrations of other cytokines (including TNFα, IL1β, IL12p70, and IL6) at different time points. Together, these results corroborated the proinflammatory and antitumor immune effects of branched N-glycans removal through exposure of immunogenic mannose glycan epitopes.
Branched N-glycans are direct modulators of the immune microenvironment
To further demonstrate the existence of a dynamic cross-talk between tumor-expressing branched N-glycans and immune cells in the microenvironment, MC38 WT and T5KO cells were subcutaneously inoculated in syngeneic immunocompetent C57BL/6 WT mice or immunodeficient Rag2−/−IL2rg−/− mice (deficient in functional T, B, and natural killer (NK) cells and reduced myeloid cells). Our results showed that the removal of complex branched N-glycans only resulted in nontumor growth in immunocompetent mice, which demonstrated the dependency on immune recognition of mannose glycans to control tumor growth (Fig. 4H and I). MC38 T5KO cells exposing mannose glycans (through removal of branched N-glycans) in immunodeficient mice exhibited an increase in tumor growth, whereas immunocompetent mice showed regression of tumors (Fig. 4H and I). The same was observed in MC38 WT cells expressing branched N-glycans (Supplementary Fig. S5A–S5C). These results demonstrated the interplay and functional cross-talk between glycans expressed on cancer cells and the immune cells in the tumor microenvironment.
Deletion of branched glycosylation results in spontaneous suppression of colorectal cancer
To translate the observations into a preclinical model of spontaneous colorectal cancer, we crossed the ApcMin/+ mice (widely used to study intestinal tumorigenesis; ref. 33) with Mgat5−/− mice and after 140 days (based on ref. 25) the tumor burden was evaluated, as well as the tumor-infiltrating immune cells. The ApcMin/+Mgat5−/− mice developed a lower number of tumor lesions with smaller size in the different parts of small intestine compared with ApcMin/+Mgat5+/+ mice (Fig. 5A–C). Analysis of overall survival showed that ApcMin/+Mgat5−/− mice survived longer compared with ApcMin/+Mgat5+/+ mice (Fig. 5D). These findings highlight that β1,6-GlcNAc–branched N-glycan removal, and the consequent exposure of mannose glycans (Fig. 5E), can inhibit colorectal cancer development and progression.
Given the in vivo antitumor effects of the Mgat5 deletion, we next examined its impact in the modulation of antitumor responses. A segment of the distal part of the small intestine was collected from ApcMin/+ Mgat5+/+ and ApcMin/+ Mgat5−/− mice and the immunoprofile was analyzed. Our results showed that inhibition of tumor growth in ApcMin/+Mgat5−/− mice was accompanied with an increased infiltration of DCs and lymphocytes, together with a significant decreased expression of MR+ M2 protumoral macrophages (Fig. 5F and G). We further showed that colorectal cancer development in Mgat5-null mice was accompanied by a significant decrease of Foxp3+ Tregs in the tumor microenvironment and a significant decrease of the immune checkpoint molecule PD-1, predominantly in CD4+ T cells (Fig. 5H). Intestinal explant supernatants collected from ApcMin/+Mgat5−/− mice released higher concentrations of proinflammatory cytokines, predominantly IL17A, compared with ApcMin/+Mgat5+/+ lesions. The concentration of the anti-inflammatory cytokine IL10 was slightly lower in ApcMin/+Mgat5−/− samples compared with ApcMin/+Mgat5+/+ samples (Fig. 5I). In accordance, a higher Il17a mRNA expression and a slight decrease of Il10 expression was detected in ApcMin/+Mgat5−/− mice tissues compared with the controls (Supplementary Fig. S6A). No major differences were observed in Ifng mRNA expression (Supplementary Fig. S6A). The data obtained by immunofluorescence also showed a higher expression of IFNγ and IL17A in ApcMin/+Mgat5−/− mice intestinal samples compared with ApcMin/+Mgat5+/+ mice (Fig. 5J). We also observed these immunostimulatory effects upon increased exposure of mannose-enriched glycans in Mgat5 KO mice (Fig. 5E).
As proof-of-concept of the in vivo effects of Mgat5 deletion in the suppression of colorectal cancer progression, via immunestimulation, we used a different colorectal cancer mouse model based on conditional knockout of Msh2 in villin-expressing intestinal tissue, VCMsh2LoxP/LoxP mice. These mice develop intestinal adenomas and carcinomas similar to human Lynch syndrome and are microsatellite instability-high (MSI-high). VCMsh2LoxP/LoxPMgat5−/− mice developed a lower number of intestinal lesions with smaller size than control mice (VCMsh2LoxP/LoxPMgat5+/+; Fig. 6A–C). As was observed in the ApcMin/+ model, the immunoprofiling showed that the removal of complex branched glycans also skewed the balance towards immunostimulation, with decreased frequency of inhibitory Foxp3+ Tregs, associated with control of tumor growth (Fig. 6D–F). Together, our data demonstrated that the abnormal branched N-glycan expression in colorectal cancer constituted an immune checkpoint mechanism in this cancer type. We demonstrated that glycoengineering through Mgat5 deletion results in a significant exposure of relevant immunogenic glycan epitopes, such as mannose glycans, leading to suppression of tumor growth associated with immune stimulation and effective antitumor responses.
The gut microbiome composition in the absence of branched N-glycans: Impact on antitumor immune responses
The gut microbiota has been described as a major player in shaping immune response and inflammation (34). Changes in microbiota composition have been associated with major intestinal inflammatory disorders, such as inflammatory bowel diseases (IBD) and cancer (reviewed in refs. 35–37). In the cancer context, it is known that intestinal microbiota contribute to the immune system activation against tumors and response to chemotherapy (38, 39). However, the contribution of glycans to the composition and function of the gut microbiota, and their potential implications in the immunomodulation associated with cancer progression, remains completely unknown. To establish a correlation between changes in N-glycan expression, upon Mgat5 deletion, and the modulation of gut microbiota composition associated with antitumor immune responses in colorectal cancer, stools from ApcMin/+Mgat5+/+ and ApcMin/+Mgat5−/− mice were collected, and the microbiome composition was analyzed by 16S rDNA gene sequencing. We observed a reduction in alpha diversity in ApcMin/+Mgat5−/− mice compared with ApcMin/+Mgat5+/+ mice (Fig. 7A). Alterations in composition were also detected between ApcMin/+Mgat5+/+ and ApcMin/+Mgat5−/− mice (Fig. 7B and C). Our results suggest that the absence of complex branched N-glycans modulated the gut microbiota composition, with an increase in Enterococcus in ApcMin/+Mgat5−/− mice compared with ApcMin/+Mgat5+/+ mice (Fig. 7C). The single strain of Enterococcus detected, although not assigned to a species by the classifier, had a 100% 16S sequence identity match to Enterococcus faecalis NBRC 100480 and other Enterococcus faecalis strains by blast. In accordance, it was reported that different strains from Enterococcus bacteria exhibited an immunostimulatory effect, being able to promote DC activation and IFNγ production and associated with reduction of polyps progression in ApcMin/+ mice, which supports the hypothesis that these microorganisms have a role in the modulation of the tumor microenvironment (40–42). Altogether, our data suggest that the absence of branched N-glycan expression potentiates an effective antitumor immune response associated with the modulation of intestinal microbiota composition and function.
Discussion
Changes in protein glycosylation is a hallmark of cancer development and progression, and particularly, the overexpression of branched N-glycans by cancer cells is one of the most widely occurring cancer-associated glycan modifications (13–15). In gastrointestinal cancer, overexpression of GnT-V–mediated branched N-glycans by cancer epithelial cells exert a proinvasive and prometastatic behavior in cancer cells through mechanistic interference with tumor cell adhesion (13–15). The detection of this aberrant glycoform in patients with gastric cancer also correlates with poor survival (16). The role of GnT-V–mediated branched N-glycans in chronic inflammatory gastrointestinal disorders, such as IBD, has been demonstrated and shows that branched N-glycans regulate T-cell–mediated immune responses (22, 35, 43, 44). However, whether this abnormal expression of cancer-associated branched N-glycans is associated with immune modulation and immune evasion correlated with cancer progression remains unknown.
The immune system is primed with a variety of different glycan-binding receptors (collectively designated as lectins) that have the ability to sense changes in cellular glycosylation and instruct either immunosuppressive or immunostimulatory responses. GBPs involved in glycan immune recognition and signaling include the C-type lectins, galectins, and siglecs and can be expressed or secreted by DCs, macrophages, B cells, T cells, and NK cells (21, 45). Changes in both O- and N-linked glycans occur in cancer (11) and both types of glycosylation can be sensed and recognized by specific immune cell glycan-binding receptors (46, 47). We previously reported a correlation between N- and O-glycosylation and cancer cell behavior (15); however, the interplay in immune evasion remains to be defined and is worth exploring further.
Here, we investigated how the aberrant expression of complex branched N-glycans by gastrointestinal cancer cells affects immune responses within the tumor microenvironment by addressing the impact of removing branched N-glycosylation on the exposure of immunogenic (glycan)epitopes (like mannose-enriched glycans) and how this affects antitumor immune recognition and surveillance, along with exploring the relationship with the gut microbiome. We demonstrated that MGAT5-mediated branched N-glycans are an immune checkpoint in colorectal cancer and are overexpressed during colorectal cancer carcinogenesis. This process displaying a positive correlation with Foxp3 expression. In vitro studies further showed that cancer cell branched N-glycans prevented immune recognition and contributed to tumor immune escape. The overexpression of complex branched N-glycans by gastrointestinal tumor cells imposed an immunosuppressive response within the microenvironment by inhibiting IFNγ production. The role of branched N-glycans in immune escape was further demonstrated by removing this glycan by KF or Mgat5 CRISPR-Cas9 editing. The elimination of Mgat5-mediated branched N-glycans in colon epithelial cancer cells revealed and exposed mannose-enriched glycans associated with suppression of tumor growth in vivo, along with an increased inflammatory response. Accordingly, in colorectal cancer clinical samples, mannose-enriched glycans were downregulated during cancer progression. These evidences support that branched glycans were covering/masking immunogenic glycans, such as mannose, that upon exposure, could unleash effective antitumor immune responses.
In fact, high-mannose glycans are typically found in lower organisms and can be recognized as “non-self” by specific C-type lectins, such as MR and DC-SIGN expressed by macrophages and DCs, eliciting a proinflammatory immune response (48). We showed that treatment of gastrointestinal cancer cells with inhibitory anti-MR/DC-SIGN resulted in downregulation of the activation marker CD86, which supports the role of mannose-recognizing C-type lectins in the recognition of glycans and tumor immunomodulation. Accordingly, treatment with mannose is shown to cause tumor growth retardation via enhancement of tumor cell death (49). The immunogenic role of mannose glycans in autoimmunity shows that deficiency of α-mannosidase-II results in the expression of mannosylated glycans and induction of an autoimmune-like disease in mice (50). Moreover, the lack of branched glycans (by blocking galectin-1 binding) reduces angiogenesis and can increase immune cell infiltration (51). On the other hand and in line with the immunosuppressive functions of the β1,6-GlcNAc–branched glycans, the presence of sialic acid, commonly found in the branched N-glycans terminal part, associates with inhibition of effector T cells and induction of de novo Tregs through recognition by Siglecs (52).
The dynamic cross-talk between branched/mannose glycans expressed at the tumor cell surface and on immune cells in the tumor microenvironment was demonstrated when glycoengineered colon cancer cells were inoculated either in immunocompetent or immunodeficient mice. We demonstrated that mannose glycan exposure unleashed an antitumor immune attack in immunocompetent mice, suggesting a direct interaction between glycans and immune cells needs to occur to modify antitumor immunity and tumor control. These results provide a strategy whereby glyco-engineering can unmask immunogenic glycan epitopes, such as mannose, resulting in enhanced antitumor immune responses.
The proof-of-concept of these findings was revealed in our preclinical models of colorectal cancer. When Mgat5-null mice were crossed with Apc or Msh2-mutated mice, we observed suppression of tumor growth that associated with an increased survival rate. The absence of branched N-glycans, and the consequent exposure of immunogenic mannose-enriched glycans, was found to trigger a proinflammatory immune response characterized by a significant decrease of Tregs and decreased expression of classical checkpoint molecules, such as PD-1. These immunomodulatory effects imposed by the absence of branched N-glycans and the revealing of mannose glycans led to an effective antitumor immune attack associated with colorectal cancer prevention. The future identification of relevant protein carriers of these branched/mannose–enriched N-glycans will be important for developing future targeted immunotherapies.
Given that the colon mucosa is covered with a dense and complex coat of glycans, called the glycocalyx, and that the gut microbiota use glycans as a source of energy, we evaluated the impact of gut mucosa glycosylation in cancer immunoediting, immunosurveillance, and its complementary/synergistic effects on the microbiota composition. We also asked whether and how an altered host glycoprofile (via Mgat5-glyco-engineering and the consequent remodeling of mucosa branched N-glycosylation) affected the microbiome composition that associated with immunomodulation in colorectal cancer. We found that in ApcMin/+ mice, the Mgat5 deletion significantly altered the microbiome diversity and composition whereby the absence of branched N-glycans led to an increased abundance of Enterococcus bacteria. Accumulating evidences support an important role of gut microbiome in antitumor immunity (38, 53, 54). Changes in intestinal microbiome composition associate with resistance to anti-PD-1–based immunotherapy (55). Here, the identification of Enterococcus bacteria that associated with suppression of colorectal cancer growth and antitumor immune responses (triggered by the removal of branched N-glycans) supports the effects of altered host glycosylation in tumor immunomodulation and tumor growth. These evidences are in line with a study showing that Enterococcus faecalis CECT7121 induces immune activation, characterized by the activation of DCs and IFNγ production (40, 42). Another study describes that heat-killed Enterococcus faecalis EC-12 is able to reduce polyp progression in ApcMin/+ mice by suppressing β-catenin signaling (41). Our evidence reveals the effects of the remodeling of mucosa glycosylation in the selection of specific microorganisms that together converge to regulate the antitumor immune response and tumor growth.
Altogether, this study described a glycan-associated immune checkpoint in colorectal cancer, in which the expression of abnormal branched N-glycans induced immunosuppressive networks in the tumor microenvironment through hidden immunosurveillance. This tumor-specific glycosignature could be a tool to identify high-risk individuals of colorectal cancer progression that can be potentially eligible for early immunopreventive interventions. From the clinical point of view, we also demonstrated that glycoengineering to remove tolerogenic branched N-glycan is a strategy to expose relevant immunogenic glycan epitopes that not only impose changes in microbiome composition, but also triggers an effective antitumor immune response (Supplementary Fig. S7, summary of data).
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
M.C. Silva: Conceptualization, software, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. Â. Fernandes: Conceptualization, software, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. M. Oliveira: Investigation, methodology. C. Resende: Resources, investigation, methodology. A. Correia: Resources, investigation, methodology. J.C. de-Freitas-Junior: Conceptualization, formal analysis, investigation, methodology. A. Lavelle: Software, formal analysis, methodology. J. Andrade-da-Costa: Investigation, methodology. M. Leander: Methodology. H. Xavier-Ferreira: Methodology. J. Bessa: Resources. C. Pereira: Resources. R.M. Henrique: Resources. F. Carneiro: Resources. M. Dinis-Ribeiro: Resources. R. Marcos-Pinto: Resources. M. Lima: Resources. B. Lepenies: Resources. H. Sokol: Resources. J.C. Machado: Resources. M. Vilanova: Resources. S.S. Pinho: Conceptualization, resources, supervision, funding acquisition, investigation, writing–original draft, project administration, writing–review and editing.
Acknowledgments
This article is a result of the project NORTE-01-0145-FEDER-000029, supported by the Norte Portugal Regional Programme (NORTE 2020) under the PORTUGAL 2020 Partnership Agreement through the European Regional Development Fund. This work was also funded by Fundo Europeu de Desenvolvimento Regional (FEDER) funds through the COMPETE 2020—Operacional Programme for Competitiveness and Internationalization (POCI), Portugal 2020, and by Portuguese funds through the Portuguese Foundation for Science and Technology in the framework of the projects (POCI-01/0145-FEDER-016601/PTDC/DTP-PIC/0560/2014 and POCI-01-0145-FEDER-028772). M.C. Silva acknowledges Fundação para a Ciência e a Tecnologia (FCT; SFRH/BD/136388/2018) for a fellowship. J. Bessa acknowledges FCT, for an FCT Investigator position (Grant IF/00654/2013) and support from European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme (grant agreement no. ERC-2015-StG-680156-ZPR). J.C. Machado acknowledges FCT for funding (PTDC/MED-PAT/32462/2017; PTDC/BIM-MEC/2834/2014). The authors thank Prof. Naoyuki Taniguchi for kindly providing the MKN45 Mock and T5 cell lines. They thank Prof. Michael Pierce for kindly providing the Mgat5-knockout mice. The authors thank Prof. Winfried Edelmann, from Albert Einstein College of Medicine, New York, NY, for kindly providing the VCMsh2loxP/loxP mice. They also thank Prof. Adriane Todeschini for kindly providing the MC38 cell line. The authors thank Renata Carriço, from i3S, for the technical support in the CRISPR/Cas9 approach. They acknowledge the support of the i3S Scientific Platforms: Animal Facility and Translational Cytometry (TraCy). The Institute of Molecular Pathology and Immunology of the University of Porto integrates the i3S research unit, which is partially supported by the Portuguese Foundation for Science and Technology.
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.