The C-type lectin MGL expressed by dendritic cells detects glycan changes on MUC1 in colon carcinoma

Abstract

The epithelial mucin MUC1 is a high molecular weight membrane glycoprotein frequently overexpressed and aberrantly glycosylated in adenocarcinoma. Mucins normally contain high amounts of O-linked carbohydrate structures that may influence immune reactions to this antigen. During malignant transformation, certain glyco-epitopes of MUC1, such as Tn-antigen, TF-antigen and their sialylated forms become exposed. The role of these glycan structures in tumor biology is unknown, but their presence is known to correlate with poor prognosis in several adenocarcinomas. We analyzed the potency of MUC1 containing Tn-antigens (MUC1-Tn) to target C-type lectins that function as carbohydrate recognition and uptake molecules on dendritic cells (DC). We identified the macrophage galactose type C-type lectin (MGL), expressed by both DC and macrophages, as the receptor for recognition and binding of MUC1-Tn. To validate the occurrence of MGL–MUC1 interactions in situ, we studied the binding of MGL to MUC1 in primary colon carcinoma tissue. Isolation of MUC1 out of colon carcinoma tissue showed strong binding activity to MGL. Interestingly, MGL binding to MUC1 was highly correlated to binding by the lectin Helix pomatia agglutinin (HPA), which is associated with poor prognosis in colorectal cancer. The detection of MGL positive cells in situ at the tumor site together with the modified glycosylation status of MUC1 to target MGL on DC suggests that MGL positive antigen presenting cells may play a role in tumor progression.

Introduction

The epithelial mucin, MUC1, is a high molecular weight membrane glycoprotein, normally expressed on the apical surfaces of glandular epithelial cells. It is frequently overexpressed and aberrantly glycosylated in adenocarcinomas, such as breast-, pancreas-, and colon carcinoma [1–3]. MUC1 is believed to provide a protective barrier, preventing adherence by pathogens as well as lubricating the epithelial layer [4]. More recently, MUC1 has been found to be expressed on activated T lymphocytes [5] and mature dendritic cells [6, 7]. The function of MUC1 on leukocytes is currently unknown, although it has been suggested to function in T-cell migration [8].

Mucins are heavily glycosylated molecules. The extracellular part of the protein is composed of around 25–100 tandem repeats, each containing 20 amino acids. Within each tandem repeat there are five potential O-glycosylation sites that carry glycans, and near the membrane proximal region there are a total of five potential N-glycosylation sites [9]. These glycans may comprise ∼80% of the molecular weight [1]. During malignant transformation , MUC1 expression is highly upregulated on cancer cells [4, 10] and the glycans are severely truncated, resulting in reduction of the total amount of carbohydrates present [11, 12]. Both the peptide backbone and new glycan epitopes become exposed to the immune system. The glycan epitopes include Tn-(GalNAcαSer/Thr), Thomsen–Friedenreich (TF, Galβ1-3GalNAcαSer/Thr, core 1) antigens and their sialylated counterparts [13–15]. Expression of these glycan antigens is often increased in colon cancer, whereas they are not expressed in normal colon [16, 17]. Such changes are particularly useful when considering diagnostic markers for cancer. For instance, lectin from the Roman snail Helix pomatia (H. pomatia agglutinin, HPA) can be used as a prognostic indicator for several adenocarcinomas. HPA interacts specifically with Tn-antigens and HPA tissue staining has been shown to correlate with unfavourable prognosis in colorectal carcinomas [18], breast carcinomas [19], and gastric carcinomas [20]. However, little is known on the biological effects of modified glycosylation on the immune system and tumor clearance.

Dendritic cells (DC) play a key role during the initiation of immune responses. Immature DC internalize antigens for processing and presentation, which often requires specific receptors, such as pattern recognition receptors (PRR). Immature DC express various C-type lectin receptors (CLR), that function as PRR, and are involved in antigen recognition, uptake and presentation [21]. CLR specifically recognize carbohydrate structures and engagement of CLR, such as DC-SIGN [22], mannose receptor [23] and DEC-205 [24] is known not to influence DC maturation, and thus antigen uptake through CLR primarily results in an immunosuppressive or tolerogenic phenotype. However, when DC properly mature during CLR mediated antigen uptake the immune suppressive state of CLR can be overcome, leading to induction of antigen specific immune responses [25]. This underlines the important role of C-type lectins in steering immune responses. CLR are likely to be involved in recognition of aberrantly glycosylated tumor antigens, as recently has been shown for the CLR DC-SIGN to recognize specific glycan epitopes on carcinoma embryonic antigen in colon carcinoma [26]. Also for the tumor-associated MUC1-glycosylation changes have been observed to influence the growth rate of tumors in a murine model of mammary carcinoma [27] and it interesting to speculate that they may interact with CLR on DC, thereby influencing immune responses.

In this study, we analyzed the interaction of tumor associated MUC1 containing Tn-antigens (MUC1-Tn) with CLR on DC. We identified the macrophage galactose type C-type lectin (MGL) [28, 29] as a receptor for this MUC1 glycoform. Interestingly, MGL bound specifically to MUC1-Tn in primary colon carcinoma and not with MUC1 from normal epithelial cells.

Materials and methods

Cells

Chinese hamster ovary cells (CHO) were cultured in RPMI containing 10% fetal bovine serum (FBS) and streptomycin/penicillin. Monocyte derived dendritic cells (MoDC) were cultured as described before [30], with modifications. In short, human blood monocytes were isolated from buffy coats by a Ficoll gradient step, followed by MACS beads isolation of CD14+ cells (Miltenyi Biotec, Bergisch Gladbach, Germany). The monocytes were cultured in the presence of IL-4 and GM-CSF (500 and 800 U/mL, respectively; ScheringPlough, Brussels, Belgium) for 4–6 days. The phenotype of immature DC was confirmed by flow cytometry (CD11b^high, CD11c^high, ICAM-1^high, CD80^low, CD83^low, CD86^low). The DC were matured by incubation with lipopolysaccharide (LPS, Salmonella typhosa, 1 μg/mL, Sigma–Aldrich, St. Louis, MO). DC maturation was confirmed by CD80, CD83 and CD86 expression and DC purity was >90%.

Recombinant MUC1-Fc fusion protein

MUC1-Fc containing 32 tandem repeats of MUC1 fused with the Fc part of murine IgG2a was produced in CHO-K1 cells as described before [31]. Purified MUC1-Fc glycoprotein was treated first with V. cholerae neuraminidase (Roche) in 50 mM NaAc pH 5.5, 4 mM CaCl₂ at 37 °C for 16 h, and then with β-galactosidase from bovine testes (Sigma) in 50 mM NaAc pH 4.5 at 37 °C for 16 h, to remove sialic acid and galactose, respectively. The integrity of the MUC1-Fc proteins after this treatment, was checked by SDS-PAGE. The resulting glycoform was determined by ELISA using glycan specific antibodies and lectins (see “Binding assay (ELISA)” section).

The MUC1-Fc protein used for fluorescent labeling was produced in the glycosylation mutant CHO cells Lec3.2.8.1 [32], making only GalNAc O-glycans (and high mannose N-glycans) by the core facility “Mammalian Protein Expression” at Göteborg University and purified by metal chelating chromatography using HiTrap chelating HP (GE Healthcare) loading with Co²⁺. MUC1-Fc was labeled with Alexa-488 according to manufacturers instructions (Molecular Probes, Eugene, OR, USA). The MUC1-Fc contains only one potential N-glycosylation site [31] that is probably differently glycosylated in Lec3.2.8.1 cells and CHO-K1 cells [32]. In spite of this, high mannose N-glycans did not result in altered binding of MUC1-Fc to DC (data not shown).

MUC1 samples

Tissue samples from colon carcinoma were collected after surgical removal of the tumor with informed consent of the patients. Primary colon carcinoma tissue and the corresponding normal tissue were obtained from patients from resection specimens, following national and institutional ethical guidelines regarding the use of human tissues. Tissue samples (∼0.5 to 1.0 cm³) were gently homogenized and incubated in 1–3 mL lysis buffer (TEA buffer containing 1% Triton-X-100, 2 mM CaCl₂, 2 mM MgCl₂ with protease inhibitor coctail tablets (Roche) depending the sample volume for 2 days at 4 °C under rotation. The lysates were centrifuged at 14,000 rpm for 15 min, supernatant collected and stored in aliquots at −80 °C. Protein concentration was measured by BCA assay (Pierce, Rockford, IL, USA).

Antibodies and CLR-Fc fusion proteins

The following antibodies were used: clone 214D4 (provided by John Hilkens, Netherlands Cancer Institute, Amsterdam, The Netherlands) specifically recognizing the amino acid sequence PDTR in the extracellular domain of MUC1 (glycosylation independent); this antibody was biotinylated using a biotinylation kit (Pierce); DC-SIGN specific antibody AZN-D1 [33]; CD68 specific antibody EBM-11; FITC-labeled goat anti-mouse IgG (Jackson Immunoresearch, West Grove, PA, USA); LAMP-1 specific antibody (BD Biosciences, Erembodegem, Belgium); Peroxidase-labeled goat anti-human Fc (Jackson Immunoresearch); Alexa-488 labeled goat anti-mouse IgG1,Alexa-594 labeled goat anti-mouse IgG2a, and Alexa-594 labeled goat anti-mouse IgG1 (Molecular Probes). Anti-glycan antibodies 5E5 (MUC1-Tn) and HH8 (MUC1-TF) were provided by Henrik Clausen, University of Copenhagen, Denmark. The MGL-human Fc fusion protein was made in the following way: the extracellular part of MGL (amino acids 61–289) was amplified on pRc/CMV-MGL by PCR, confirmed by sequence analysis and fused at the C-terminus to human IgG1-Fc in the Sig-pIG1-vector [29]. MGL-Fc was produced by transfection of CHO cells and concentration determined by ELISA. An MGL-murine Fc fusion protein was generated by cloning the extracellular part of MGL (including the CD38 signal sequence from the Sig-pIG1 vector) into an pcDNA3 expression vector containing exons 1–3 of murine IgG2a-Fc [31].

The 18E4 monoclonal antibody to MGL was generated by immunizing Balb/c mice with purified MGL-murine Fc. Hybridoma supernatants were screened for the presence of anti-MGL antibodies on CHO-MGL transfected cells [47].

Beads adhesion assay

Carboxylate-modified TransFluorSpheres (488/645 nm, 1.0 μm; Molecular Probes) were coated with streptavidin as previously described [34]. Streptavidin coated beads were incubated with biotinylated F(ab′)₂ fragments of goat anti-mouse IgG2a (6 μg/mL, Jackson Immunoresearch) in 0.5 mL PBS 0.5% BSA for 2 h at 37 °C, washed and further incubated with MUC1-Fc fusion proteins at 4 °C overnight. Finally, beads were washed, stored in 100 μL PBS 0.5% BSA, and used within 1 week. The beads adhesion assay was performed as previously described [34]. DC were incubated with coated beads in Tris–sodium buffer (20 mM Tris–HCl, pH 7, 150 mM NaCl, 2 mM CaCl₂, 2 mM MgCl₂) with 0.5% BSA and adhesion was determined in the absence or presence of blocking agents (10 mM EGTA, 25 mM monosaccharides (Sigma) or 5 mM mannan) at 37 °C for 45 min. Cells were washed and analyzed by flow cytometry.

Binding assay (ELISA)

The solid phase adhesion assay was performed by coating ELISA plates with purified anti-MUC1 antibodies (clone 214D4) in 0.2 M Na₂CO₃, followed by blocking with 1% BSA in Tris–sodium buffer. Negative controls were included for each sample, where no anti-MUC1 antibodies were added. After washing with Tris–sodium buffer, tissue lysates were added (0.2–1.0 mg/mL protein) and incubated at 4 °C overnight. Plates were washed and further incubated with MGL-human Fc (0.5 μg/mL) or biotinylated HPA (5 μg/mL, Sigma), dioxygene labeled MAA (5 μg/mL, Boehringer Mannheim) or anti-TF-antibodies (clone HH8) for 1 h at room temperature (RT). After washing with Tris–sodium–0.02% Tween the binding was detected by peroxidase labeled anti-human IgG, anti-streptavidin (Jackson Immunoresearch), anti-digoxigenin antibodies or anti-murine IgG. Specificity was determined in the presence of 10 mM EGTA or 100 mM GalNAc monosaccharides. MUC1 coating was determined by biotinylated anti-MUC1 antibodies (clone 214D4, 2 μg/mL) followed by incubation with peroxidase-labeled streptavidin. The reaction was developed by TMB substrate and optical density measured by a spectrophotometer.

Immunofluorescence analysis and confocal microscopy

Cryosections (7 μ) of normal colon epithelium or primary colon carcinoma were fixed in 100% acetone (10 min), washed with PBS, and incubated with first antibody (10 μg/mL) for 1 h at 37 °C. After washing, the final staining was performed with Alexa-594 labeled goat-anti mouse IgG2a or Alexa-488 labeled goat-anti mouse IgG1 and nuclear staining was performed with Hoechst. For staining with MGL-Fc, cryosections were fixed in 2% paraformaldehyde and all incubation steps were performed in Tris–sodium buffer. The final staining was performed with Alexa-488 labeled goat-anti mouse IgG.

To study internalization, Alexa-488 labeled MUC1-Fc (30 μg/mL) was incubated with DC for 2 h at 37 °C. The Transferrin receptor was detected by incubating the cells with Alexa Fluor 594-conjugated Transferrin (10 μg/mL; Molecular Probes) for 15 min at 37 °C prior to fixation. Labeled cells were fixed in 3% paraformaldehyde in PBS and permeabilized in PBS–0.1% saponin prior to staining. Cells were stained with antibodies to LAMP-1, a lysosomal marker, and subsequently with Alexa-594 conjugated anti-mouse IgG1. Next, cells were allowed to adhere to poly-l-lysine coated glass slides and mounted in anti-bleach reagent. Fixed slides were imaged with a Leica TCS SP2 AOBS confocal laser-scanning microscope (Leica Microsystems, Heidelberg, Germany).

Statistical analysis

Statistical differences between groups were analyzed by Wilcoxon matched pairs test using Graphpad Instat from Graphpad Software, Inc. Correlation was analyzed by a non-parametric Spearman ranks test. Significance was accepted at the p < 0.05 level.

Results

Characterization of MUC1-Fc fusion proteins

The extracellular domain of MUC1, including 32 tandem repeats with five O-glycosylation sites each, was fused to a murine IgG Fc tail and purified in Chinese hamster ovary cells, as previously described [31]. To obtain a glycoform of MUC1 with mainly Tn-antigen, the protein was desialylated and treated with β-galactosidase, leaving only the core GalNAcs on the protein. The resulting MUC1-Tn reacted with antibodies against Tn, but not against TF in ELISA (Fig. 1). These results indicate that the glycosylation mainly consisted of GalNAcs.

Fig. 1.

MUC1-Fc fusion proteins contain Tn-antigens. MUC1-Fc was serially diluted and coated on ELISA plate. Binding of the following antibodies/lectins was analyzed: Tn-specific antibodies (5E5), HPA (specific for Tn-antigens) and TF-specific antibodies (HH8)

MUC1-Fc containing Tn-antigens interacts with monocyte derived dendritic cells

To study the potency of tumor associated MUC1 to target DC, MUC1-Fc containing Tn-antigens (GalNAcαSer/Thr) was coupled to fluorescent beads and validated for interaction with monocyte derived DC (MoDC). Both immature DC and DC matured with lipopolysaccharide were analyzed for binding. MUC1-Tn bound efficiently to immature DC and binding was calcium dependent, suggesting a role for CLR (Fig. 2a). Mature DC showed strongly reduced interaction with MUC1-Tn, concomitant with the fact that mature DC down-regulate expression of CLR on their cell surface (Fig. 2a). Fluorescent beads coated with polyacrylamide-coupled Lewis-X, a carbohydrate epitope for the CLR DC-SIGN [35], served as a positive control. As expected, Lewis-X bound strongly to immature DC that express high levels of the CLR DC-SIGN and binding to mature DC, that express low levels of DC-SIGN, was significantly reduced (Fig. 2b). Beads coated only with goat anti-mouse antibodies did not bind to MoDC (data now shown).

Fig. 2.

MUC1-Tn interacts with dendritic cells. MUC1-Fc proteins were coupled to fluorescent beads and incubated with day 5 immature or mature monocyte derived DC in the presence or absence of 10 mM EGTA (a). Beads coated with polyacrylamide coupled Lewis X antigens (ligand for DC-SIGN on DC) served as a positive control for beads binding (b). The figure shows the percentage of cells binding beads as determined by flow-cytometry. Mature DC were obtained by incubating DC in the presence of 100 ng/mL lipopolysaccharide overnight. Samples were analyzed in triplicates (error bars represent standard deviation of the mean) and one representative experiment is shown out of 3

To determine the specificity of MUC1 binding, DC were pre-incubated with different carbohydrate monosaccharides (25 mM) to block interaction with CLR (Fig. 3a). Only N-acetylgalactosamine (GalNAc) monosaccharides partially blocked binding of MUC1-Tn to DC, suggesting the involvement of a GalNAc-specific receptor (see below). No block was observed by pre-incubation with galactose, mannose or mannan, suggesting that CLR such as DC-SIGN or mannose receptor were not involved. Furthermore, MUC1 binding was abrogated after DC maturation, indicating down-modulation of the interacting receptor following DC maturation (Fig. 2a).

Fig. 3.

The macrophage galactose-type C-type lectin (MGL) binds to MUC1. a To determine the specificity of MUC1 interaction with DC, cells were pre-incubated with different saccharides before adding MUC1-Fc coated beads. Only GalNAc structures reduced binding of MUC1 to DC and mannan blocked binding of Lewis X. b MGL expression on day 5 immature and mature DC was determined by mAb 18E4. c DC were pre-incubated with different anti-CLR blocking antibodies before incubation with MUC1-Fc coated beads. The figure shows the percentage of cells binding beads as determined by flow-cytometry. Samples were analyzed in triplicates (error bars represent standard deviation of the mean) and one representative experiment is shown out of 3

A candidate C-type lectin, expressed on immature DC and macrophages, with specificity for GalNAc carbohydrate structures [29, 36], is the macrophage galactose-type C-type lectin (MGL) [28]. MGL expression is down-modulated upon DC maturation (Fig. 3b), coinciding with loss of MUC1 binding. Pre-incubation of MoDC with anti-MGL antibodies reduced binding of MUC1-Tn, indicating that MGL is a receptor for this MUC1 glycoform, although the block was not complete. EGTA, which inhibits the lectin domain of CLR to recognize glycans also inhibited the interaction to the same level as anti-MGL antibodies. We therefore concluded that all CLR binding activity was mediated by MGL (Fig. 3c). Neither anti-DC-SIGN nor anti-mannose receptor antibodies blocked this interaction. The specificity of MUC1-Tn for MGL was confirmed using MGL transfected Chinese Hamster Ovary (CHO) cells (Fig. 4a, b). MUC1-Tn bound strongly to MGL transfected cells and interaction was blocked by EGTA, GalNAc, or anti-MGL blocking antibodies, confirming the specificity of the interaction (Fig. 4b). Together, these data strongly indicate that MGL on DC specifically interacts with tumor associated MUC1, in particular the Tn-antigen.

Fig. 4.

MUC1 binds to MGL transfected CHO cells. a MGL expression on CHO cells transfected with MGL (filled histogram). Non-transfected CHO cells served as a negative control. MUC1-Fc-coupled beads were incubated with non-transfected CHO cells or CHO cells transfected with MGL (b). Cells were pre-incubated with EGTA or 25 mM GalNAc monosaccharides to block specific interaction with MGL. One representative experiment is shown out of 3

To study internalization of MUC1, immature DC were incubated with fluorescently labeled MUC1-Fc for 2h at 37 °C. The cells were also stained for the transferrin receptor (an early endosomal marker) and LAMP-1 (a lysosomal marker). DC efficiently internalized MUC1 and co-localization was observed with both LAMP-1 and transferrin receptor, suggesting that MUC1 is internalized for lysosomal degradation (Fig. 5). In contrast, the control Fc-protein was not clearly detected in vesicles, but showed diffused staining of much lower intensity.

Fig. 5.

MUC1 co-localizes with LAMP-1 and transferrin after internalization. DC were incubated with 30 μg/mL Alexa Fluor-488 labeled MUC1-Fc for 2 h at 37 °C. Before fixation, the DC were incubated with Alexa Fluor-594 labeled transferrin for 15 min, which is specifically transferred to the early endosomes. After fixation and permeabilization the DC were stained with anti-LAMP-1 antibodies (a lysosomal marker), followed by Alexa Fluor-594 labeled goat anti-mouse IgG1

MGL binds MUC1 derived from primary colon carcinoma

To investigate whether MGL specifically recognizes MUC1 in primary tumors, tissue samples (n = 20) were obtained after surgical removal of colorectal tumors from carcinoma patients. The surgical procedure was performed at the VU University Medical Center (VUmc), Amsterdam, The Netherlands and both normal and malignant tissue was obtained from each patient. Total protein concentration was measured in the tissue lysates and found to be significantly higher in the tumor lysates (p < 0.001) than in normal tissue lysates, probably due to differences in protein content. MUC1 concentration was also significantly higher (p < 0.0001) in tumor tissue lysates (Fig. 6a, b), however, there was no correlation between total protein concentration and concentration of MUC1.

Fig. 6.

MGL interaction with colon derived MUC1 mucins. MUC1 was captured from tissue lysates of a primary colon tumor (n = 20), the corresponding normal colon (n = 20) by mAb 214D4 in ELISA. The mucins were detected by the same antibody and binding by MGL-Fc, HPA, mAb HH8 and MAA investigated. a The relationship between MGL-Fc binding and MUC1 concentration in normal colon lysates. b The relationship between MGL-Fc binding and MUC1 concentration in colon carcinoma lysates. Relationship between MGL-Fc and c HPA binding, d HH8 binding and e MAA binding. Each patient is represented by one symbol that can be found in panels c–e. Samples were analyzed in duplicates and one representative experiment of 2 is shown

When we analyzed MGL-Fc binding to tumor MUC1, binding was significantly higher than to MUC1 out of normal epithelial tissue from the same patient (p < 0.0001). Low binding to MUC1 of normal epithelium could be due to low MUC1 concentration in normal colon lysates. Interestingly, however, MGL-Fc binding and MUC1 concentration in tumor samples did not correlate (r = 0.25; p = 0.28, Fig. 6a, b) suggesting that there were qualitative differences present that determined MGL-Fc interaction with tumor associated MUC1. To investigate this further, we analyzed the carbohydrate content of isolated MUC1 from tumor samples and normal epithelial tissues using plant lectins and carbohydrate specific antibodies. As expected, MUC1 samples varied extensively in carbohydrate reactivity between the different colorectal cancer patients. Some contained Tn-antigens, others TF-antigens or sialylated structures. Interestingly, a strong correlation was observed between MGL-Fc binding and HPA binding (r = 0.88; p < 0.0001), indicating that MGL primarily reacts with Tn-antigens on MUC1 in primary colon carcinoma (Fig. 6c). In contrast, no correlation was observed between MGL-Fc binding and the presence of TF-antigens or α2–3 sialic acid (Fig. 6d, e). In fact, there was a tendency for inverse correlation, further confirming the data that additional carbohydrate structures would mask the GalNAc epitope interacting with MGL.

These data show that both the in vitro-generated Tn glycoform of MUC1 and the in situ tumor associated MUC1 present in colorectal cancer patients bind strongly to MGL on immature DC.

MGL-expressing cells are detected in the colon

To determine whether MGL expressing cells are present in colon tissue and whether these could interact with MUC1 in tumors, tissue sections were analyzed for cells expressing MGL and MUC1. MGL expressing cells were detected within the normal colonic mucosa of four individuals analyzed as well as in primary tumor sections of all these individuals. MGL-expressing cells did not appear to express the macrophage marker CD68 in the tumor sections (Fig. 7a), although some co-localization was observed with the CLR DC-SIGN, suggesting that MGL may be expressed on certain DC populations (Fig. 7b).

Fig. 7.

Characterization of MGL expressing cells. Cryosections were stained with anti-MGL (18E4), anti-CD68 (EBM11), anti-DC-SIGN (AZN-D1) or anti-MUC1 antibodies (mAb 214D4) for 2 h at 37 °C. After washing the sections were incubated with Alexa-594 labeled goat anti-mouse IgG2a (18E4) and Alexa-488 labeled goat anti-mouse IgG1. a Colon carcinoma section stained for MGL (red) and CD68 (green). b Colon carcinoma section stained for MGL (red) and DC-SIGN (green). MUC1 staining was performed on both normal tissue (c) and colon carcinoma (d)

In normal tissue, MUC1 positive cells were detected at the epithelial layer and some MUC1 positive cells were scattered throughout the mucosa (Fig. 7c, d). However, as expected MUC1 staining was strong in the tumor tissue. The finding that the MGL expressing cells are located in the vicinity of MUC1 expressing cells demonstrate that MGL positive antigen presenting cells are present in close contact with epithelial tumor cells. To determine whether tumors expressed MGL-binding glycan epitopes, we analyzed the binding activity of MGL-Fc to these tissues. MGL-Fc showed strong staining of colon carcinoma cells and not with normal colon epithelium (Fig. 8). Furthermore, the fact that the binding of MGL-Fc was blocked by GalNAc monosaccharides, demonstrates the requirement of the carbohydrate recognition domain for staining.

Fig. 8.

MGL-Fc stains colon carcinoma. Cryosections were stained with MGL-murine Fc for 2 h at 37 °C, in the absence (a and b) or presence (c and d) of 100 mM GalNAc monosaccharides. After washing the sections were incubated with Alexa-488 labeled goat anti-mouse IgG. Both normal colon epithelium (a and c) and primary colon carcinoma (b and d) were stained from the same patient. Results from one representative patient out of 4 is shown

Discussion

The epithelial mucin MUC1 is a tumor antigen where post-translational modification is altered during malignant transformation [15]. The protein backbone is retained, but O-linked glycan structures are modified. This creates epitopes that may be recognized by the immune system, including the protein backbone and carbohydrate structures generally not present in non-malignant tissues. Adenocarcinoma patients may develop humoral and cellular immune responses to the MUC1 protein. These responses are, however, weak and not sufficient to eradicate the tumor [37, 38]. Nonetheless, anti-MUC1 antibody levels have been shown to correlate with survival [39] indicating the importance of this tumor antigen for protection against cancer.

Here, we have studied how tumor related MUC1 containing Tn-antigens may be optimally targeted to DC. DC play a key role in the initiation of immune responses and DC-based tumor vaccines are a promising tool for immunotherapeutic strategies against cancer. Therefore, it is important to understand the molecular mechanisms involved in how tumor antigens are recognized and internalized by DC to trigger efficient anti-tumor immune responses.

Using recombinant MUC1-Fc proteins containing 32-tandem repeats of the extracellular domain, and a sensitive flow cytometry-based binding assay, we demonstrated that MUC1-Tn bound efficiently to DC. The interaction was calcium dependent and was significantly contributed by the CLR MGL, that is expressed on immature DC, a known receptor for GalNAc containing carbohydrate structures [29, 40]. DC maturation resulted in abrogation of MUC1-Tn binding coinciding with loss of MGL expression. We did not observe any involvement of the mannose receptor, that has previously been described to be a receptor for glycosylated MUC1 [41]. The discrepancy between the two studies is likely due to differences in MUC1 glycoforms. The MUC1 glycoforms studied by Hiltbold et al., may have contained additional glycan epitopes that target the mannose receptor. The glycoform of MUC1 in our study, primarily contained GalNAc moieties that do not interact with the mannose receptor. We demonstrated MGL-Fc interaction with MUC1 present in tissue samples of primary colon carcinoma patients (n = 20) that strikingly correlated strongly with Tn-antigen expression (binding by the lectin H. pomatia agglutinin) whereas no correlation was observed with expression of TF-antigens or sialylated MUC1. Tn-antigens are known to be highly expressed in colon carcinoma, but not in normal colon tissue [17]. Detection of these carbohydrate epitopes by HPA (a GalNAc-specific lectin from the snail H. pomatia) has been shown to be useful for identifying adenocarcinomas with metastatic potential [42, 43]. Primary tumors of epithelial origin, which are stained positively with HPA, are more likely to metastasize and this is highly correlated with poor prognosis. It further suggests that these epitopes may play an important role during metastasis [18, 19] although it is not clear how this may be achieved or whether there is a role for such structures in tumor cell survival. Interaction of these carbohydrate structures with antigen presenting cells may promote efficient antigen uptake and presentation to T-cells. We demonstrated that MUC1 was internalized by DC and targeted to the lysosomes, which suggests that MUC1 containing Tn-antigens may be processed and presented to T-cells. However, signalling that occurs during tumor antigen uptake may significantly contribute to the quality of the T-cell response. For example, antigen uptake in the absence of Toll-like receptor signalling may result in immunological tolerance [44], where tumor cells may evade immune responses.

We demonstrated the presence of MGL positive cells both within normal colon mucosa and within primary carcinomas, demonstrating that these cells may be in close contact with the MUC-Tn expressing tumor cells. Interestingly, these cells seemed to exhibit DC-like phenotype, and they did not express the macrophage marker CD68. MGL-Fc bound strongly to colon carcinoma cells, supporting the hypothesis that MGL positive cells may interact with tumor cells.

Tumor cells may shed MUC1 into the micromilieu and it may be detected in serum of cancer patients [4]. These forms of MUC1 lack a cytoplasmic tail and can (in advanced colon cancer patients) reach such high levels that they can interfere with cell interactions, including selectin interactions [45]. These shed forms of MUC1 thus have the possibility to interfere with the immune system and help the tumor cells escape immune surveillance [45, 46]. Interaction with MGL on antigen presenting cells may similarly influence immune responses. It is interesting to mention that MGL is expressed on human immature DC and in particular highly expressed on tolerogenic DC and/or macrophages that are cultured in the presence of dexamethazone [47].

In conclusion, our data demonstrate that immature DC may interact with tumor-associated glycoforms of MUC1 via the CLR MGL. Our finding that MGL binding to MUC1 highly correlates with HPA binding, and the fact that MGL is highly expressed on immature or tolerigenic APC, suggests that this interaction may lead to immunosuppressive effects. Whether these suppressive effects are due to enhanced antigen uptake, accompanied by an reduced maturation of APC that leads to the induction of regulatory T cell responses, is currently under investigation.