Dendritic Cells Recognize Tumor-Specific Glycosylation of Carcinoembryonic Antigen on Colorectal Cancer Cells through Dendritic Cell–Specific Intercellular Adhesion Molecule-3–Grabbing Nonintegrin

Dendritic cells have the unique capacity to induce primary immune responses. Subepithelial immature dendritic cells have a high phagocytic potential and are able to sample antigens from the periphery. During inflammation, these immature dendritic cells undergo maturation, which enables them to migrate to the lymph nodes, and to present peptides derived from the acquired antigens to lymph node–residing T cells (1). Immature dendritic cells have been detected within fields of breast and colon cancer cells in contrast to mature dendritic cells that were confined to peritumoral areas (2, 3). Despite the capacity of immature dendritic cells to migrate into tumor beds and to capture tumor-derived antigens for presentation to specific T cells, adaptive immune responses are often hampered in cancer patients. Tumor cells are able to dampen T-cell responses through release of soluble MHC class I–related molecules (4); down-regulation of membrane MHC molecules; and up-regulation of vascular endothelial growth factor (VEGF; ref. 5), prostanoids (6), and immunosuppressive cytokines such as interleukin 10 (IL-10), transforming growth factor-β (TGF-β), and IL-6 (7), which inhibit dendritic cell differentiation and maturation. Immature dendritic cells are unable to provide sufficient costimulation for the induction of vigorous T-cell responses; rather, they induce regulatory T cells and T-cell tolerance (8).

Recently, dendritic cell–specific intercellular adhesion molecule (ICAM)-3–grabbing nonintegrin (DC-SIGN) has been detected on immature dendritic cells that were associated with melanoma and myxofibrosarcoma (9, 10). DC-SIGN is a C-type lectin that is specifically expressed on dendritic cells and has previously been identified on monocyte-derived dendritic cells in vitro and on dendritic cell subsets of skin, mucosal tissues, tonsils, lymph nodes, and spleen in vivo (11, 12). DC-SIGN displays affinity for high-mannose moieties (13) and functions as an internalization receptor for HIV-1 (12), HCV (14), Mycobacterium tuberculosis (15), and other pathogens (16–18) that express mannose-containing carbohydrates. DC-SIGN also mediates cellular interactions of dendritic cells with endothelial cells and T cells through high-mannose moieties on ICAM-2 (19) and ICAM-3 (11), respectively. In addition to high-mannose moieties, DC-SIGN recognizes nonsialylated Lewis^x and Lewis^y glycans and binds to Lewis^x-expressing pathogens such as Schistosoma mansoni and Helicobacter pylori (20, 21).

Carcinoembryonic antigen (CEA) is a tumor-associated antigen that is overexpressed on almost all human colorectal, gastric, and pancreatic adenocarcinomas; 70% of non–small cell lung carcinomas; and 50% of breast carcinomas (22). CEA belongs to a small immunoglobulin gene family that includes CEACAM1, which is also present on carcinoma epithelium (22). The glycosylation pattern of CEA changes upon malignant transformation (23). In contrast to CEA on normal colon epithelial cells, CEA on colorectal cancer cells that have metastasized to the liver has a high degree of fucosylation, contains higher levels of Lewis^x carbohydrates, and has de novo expression of Lewis^y moieties (24–26). Based on the specificity of DC-SIGN for Lewis^x and Lewis^y moieties and the presence of these carbohydrates on tumor-associated CEA but not on CEA from normal colon epithelium, we hypothesized that immature dendritic cells are able to distinguish malignant from normal colon epithelium through DC-SIGN. Here, we show that DC-SIGN⁺ dendritic cells of immature phenotype are present within primary colon colorectal cancer and that immature dendritic cells in contrast to mature dendritic cells interact in a DC-SIGN-specific manner with both breast and colorectal cancer cells. DC-SIGN recognizes CEA on colorectal cancer cells through Lewis^x and Lewis^y moieties but does not bind to CEA derived from normal colon epithelial cells. This indicates that the carbohydrate specificity of DC-SIGN enables immature dendritic cells to recognize colorectal cancer cells through tumor-specific glycosylation. The interaction between DC-SIGN and CEA may play a role in tolerance induction to tumor cells.

Antibodies and reagents. The following antibodies were used: 12A2 (IgG1 isotype control; ref. 27); 7F7 (IgG2a isotype control; ref. 28); Col-1 (anti-CEA, from Becton Dickinson, San Jose, CA); CLB-gran/10 (anti-CEA and anti-CEACAM1, from Sanquin, Amsterdam, The Netherlands); AZN-D1, AZN-D2, and CSRD (anti-DC-SIGN; refs. 11, 29); 6H3 (anti-Lewis^x, kind gift of Ben Appelmelk, Department of Medical Microbiology and Infection Control, VU University Medical Center, Amsterdam, The Netherlands); and F3 (anti-Lewis^y, from Calbiochem, La Jolla, CA).

DC-SIGN-Fc contains the extracellular portion of DC-SIGN (amino acid residues 64-404) and the Fc domain of human IgG1 fused at the COOH terminus of DC-SIGN (30). DC-SIGN-Fc was produced in Chinese hamster ovary K1 cells after transfection with the DC-SIGN-Sig-pIgG1-Fc vector (20 μg). ICAM-3-Fc (kind gift of D.L. Simmons, Celltech Ltd., Cambridge, United Kingdom) that contains an Fc-domain of the same isotype was used as a negative control (control-Fc).

Cells. The breast cancer cell line SKBR3 and the colon cancer cell lines SW948, WIDR, and HT29 were cultured in RPMI 10% FCS. Primary normal and malignant colon epithelial tissue was obtained from leftover material from surgical resection specimens, following national ethical guidelines regarding the use of human tissues.

Stable transfectants of K562 cells expressing DC-SIGN were obtained by electroporation of pRC-CMV-DC-SIGN (400 V, 960 μF) as described (31).

Immature dendritic cells were cultured from monocytes as described (32, 33). Briefly, monocytes were isolated from buffy coats of healthy donors through Ficoll gradient centrifugation and positive selection of CD14⁺ cells using MACS sorting according to the manufacturer's instructions (Miltenyi Biotec, Bergisch Gladbach, Germany). Isolated monocytes were cultured in RPMI 10% FCS in the presence of IL-4 and granulocyte macrophage colony-stimulating factor (600 and 800 units/mL, respectively; Schering-Plough, Brussels, Belgium) for 7 days. Dendritic cell cultures were over 90% pure, and impurities mainly consisted of T cells. Expression of the monocyte and macrophage marker CD14 was not observed (Fig. 1A). Maturation was induced by culture of immature dendritic cells with 2 μg/mL of lipopolysaccharide (from Escherichia coli; Sigma-Aldrich, St. Louis, MO; 2 μg/mL) in the final 2 days. Dendritic cell maturation was confirmed by expression of the dendritic cell maturation marker CD83 on mature but not immature dendritic cells (Fig. 1A).

Cell-cell adhesion assay. To examine cellular interactions with dendritic cells and DC-SIGN-expressing K562 transfectants, SKBR3 and SW948 cells were grown to confluent cultures in flat-bottomed 96-well plates. Dendritic cells and K562-DC-SIGN were labeled with the green fluorescent dye Calcein-AM (1 μmol/L; Molecular Probes, Eugene, OR) for 15 minutes at 37°C and incubated with SKBR3 and SW948 cells in TSM [20 mmol/L Tris-HCl (pH 8.0), 150 mmol/L NaCl, 1 mmol/L CaCl₂, 1 mmol/L MgCl₂] 0.5% bovine serum albumin (BSA) for 2 hours at 37°C. To determine DC-SIGN-specific adhesion, dendritic cells and K562-DC-SIGN were preincubated with anti-DC-SIGN antibodies (20 μg/mL) for 30 minutes at 37°C. Unbound cells were washed away, and remaining cells were lysed in lysis buffer [50 mmol/L Tris-HCl (pH 8.0) and 0.1% SDS]. The cell lysates were analyzed by fluorimetry at 488 nm (Fluostar Galaxy, BMG Labtechnologies, Inc., Durham, NC). Cell-cell adhesion was calculated as the percentage of maximal binding, which was determined by fluorimetry analysis of the lysed total amount of added Calcein-AM-labeled cells.

Immunoprecipitation and Western blot. SW948 cells, normal colon epithelial tissue, and primary colorectal cancer tissue were lysed in lysisbuffer [1% Triton X-100, 10 mmol/L TEA (pH 8.2), 150 mmol/L NaCl, 1 mmol/L MgCl₂, 1 mmol/L CaCl₂] containing a cocktail of EDTA-free protease inhibitors (Roche Diagnostics, Penzberg, Germany). CEA and CEACAM1 were immunoprecipitated from lysates using protein A sepharose beads (CL-4B, Pharmacia, Uppsala, Sweden) coated with either DC-SIGN-Fc or antibodies that recognize CEA and CEACAM1. Immunoprecipitates were run on SDS-PAGE (reducing conditions, 7% polyacrylamide gel), transferred onto nitrocellulose blot, and immunoblotted with anti-CEA and -CEACAM1 antibodies and secondary peroxidase-conjugated goat-anti-mouse antibodies (Jackson ImmunoResearch Lab., Inc., West Grove, PA) or DC-SIGN-Fc and secondary peroxidase-conjugated goat anti-human antibodies (Jackson Immunoresearch Lab.) in TSM 1% BSA.

DC-SIGN-Fc binding. Cells were incubated with DC-SIGN-Fc (10 μg/mL) for 30 minutes at 37°C in TSM 0.5% BSA and subsequently with FITC-conjugated goat anti-human secondary antibodies (Jackson ImmunoResearch Lab.) for 30 minutes at 37°C in TSM 0.5% BSA. Specific binding of DC-SIGN-Fc was determined by preincubation of DC-SIGN-Fc with anti-DC-SIGN antibodies (50 μg/mL) for 10 minutes at room temperature. Binding of recombinant DC-SIGN to the cells was monitored by flow cytometry.

ELISA-based binding assay. Goat anti-mouse antibodies (4 μg/mL, Jackson Immunoresearch Lab.) were coated onto NUNC maxisorp 96-well plates (Nalge Nunc International, Rochester, NY) for 1 hour at 37°C. Next, the ELISA plates were blocked with TSM 1% BSA for 30 minutes at 37°C, washed, and incubated with anti-CEA or anti-CEA and anti-CEACAM1 antibodies (1 μg/mL) in TSM 1% BSA for 1 hour at 37°C. After washing, CEA and CEACAM1 were captured from cell lysates of HT29, WIDR, SW948, and SKBR3 cells; normal colon epithelial tissue; and primary colorectal cancer tissue during 18 hours at 4°C. Next, DC-SIGN-Fc (1 μg/mL) was added and incubated during 2 hours at room temperature followed by detection using secondary peroxidase-conjugated goat anti-human antibodies (30 minutes at room temperature) in TSM 1% BSA. Similarly, Lewis^x and Lewis^y expression was determined on CEA and CEACAM1 using antibodies specific for these carbohydrates and secondary peroxidase-conjugated rabbit anti-mouse immunoglobulin M antibodies (Nordic Immunologic Laboratories, Tilburg, The Netherlands). After development, binding or expression was determined by measuring the absorbance at 450 nm using an ELISA plate reader (Benchmark Microplate Reader, Bio-Rad, Hercules, CA).

Fluorescent bead adhesion assay. Fluorescent beads coated with CEA or both CEA and CEACAM1 derived from SW948 cells were manufactured as described (34). Briefly, streptavidin was covalently coupled onto TransFluoSpheres (488/645 nm, 1.0 μm, Molecular Probes) according to the manufacturer's instructions. Next, streptavidin-coated beads were allowed to bind biotinylated goat anti-mouse Fc Fab₂ fragments (5 μg/mL, Jackson Immunoresearch Lab.) in PBS 0.5% BSA for 2 hours at 37°C. Next, beads were washed, incubated for 18 hours at 4°C with mouse antibodies directed against CEA alone or both CEA and CEACAM1, washed again, and incubated for 48 hours at 4°C with SW948 cell lysate (4 × 10⁶ cells/mL) to coat CEA and CEACAM1 onto the beads. Adhesion experiments were done as follows. Cells were incubated with ligand-coated beads for 45 minutes at 37°C in TSM 0.5% BSA. DC-SIGN-specific adhesion was determined by preincubation of cells with anti-DC-SIGN antibodies for 30 minutes at 37°C. The percentage of cells that bound fluorescent beads was monitored using fluorescence-activated cell sorting analysis (FACSCalibur, Becton Dickinson).

Immature dendritic cells bind breast and colorectal cancer cells through dendritic cell–specific intercellular adhesion molecule-3–grabbing nonintegrin. In contrast to mature dendritic cells that localize peritumorally, immature dendritic cells are present within breast and colorectal cancer, suggesting that interactions may occur between immature dendritic cells and tumor cells (2, 3). To investigate this, we incubated the breast cancer cell line SKBR3 and the colon cancer cell line SW948 with monocyte-derived immature and mature dendritic cells and measured cellular adhesion. Immature dendritic cells strongly bound to the cancer cell lines SKBR3 and SW948 (Fig. 1B). Little is known about receptors that mediate interactions of immature dendritic cells with cancer cells. The Ca²⁺ chelator EGTA completely blocked adhesion of immature dendritic cells to the tumor cell lines, indicating that Ca²⁺-dependent receptors mediate binding (Fig. 1B). Previously, we have shown that the Ca²⁺-dependent lectin DC-SIGN is an important adhesion receptor on immature dendritic cells, which establishes cellular interactions with both T cells and endothelial cells (11, 19). Anti-DC-SIGN antibodies inhibited adhesion of immature dendritic cells to SKBR3 completely and to SW948 partially, showing that DC-SIGN mediates interactions between immature dendritic cells and these cancer cell lines (Fig. 1B). In contrast, mature dendritic cells expressed lower levels of DC-SIGN and did not bind to SKBR3 and SW948 cells (Fig. 1A and B). Possibly, expression of DC-SIGN on mature dendritic cells is too low to support cellular adhesion with the tumor cell lines, which requires strong interactions and cytoskeletal rearrangements. We used stable K562 transfectants expressing DC-SIGN to further examine the role of DC-SIGN in cellular interactions with cancer cells. In contrast to mock-transfected K562 cells, K562 cells transfected with DC-SIGN bound to SKBR3 and SW948 cells (Fig. 1C). Adhesion of K562-DC-SIGN cells to the cancer cell lines was abrogated in the presence of anti-DC-SIGN antibodies. This shows that DC-SIGN is an important adhesion receptor that mediates cellular interactions with tumor cells independent of other receptors (Fig. 1C).

Dendritic cell–specific intercellular adhesion molecule-3–grabbing nonintegrin binds carcinoembryonic antigen and CEACAM1 on colorectal cancer cells. To identify the ligands of DC-SIGN on colorectal cancer cells, we used a chimera consisting of the extracellular domain of DC-SIGN and the Fc-domain of human IgG1 (DC-SIGN-Fc; 30). Immunoblotting of SW948 lysate with DC-SIGN-Fc showed that DC-SIGN binds a 160- and a 200-kDa ligand on the colorectal cancer cell line (Fig. 2A). Binding of DC-SIGN-Fc was specific, because staining with control-Fc was not observed on SW948 lysate or with DC-SIGN-Fc on a control cell lysate (Fig. 2A). SW948 cells express the tumor-associated antigen CEA and the closely related protein CEACAM1 of 200 and 160 kDa, respectively, as shown by immunoprecipitation with an antibody that recognizes both CEA and CEACAM1 (Fig. 2B). To investigate whether DC-SIGN binds these antigens, we examined binding of DC-SIGN-Fc to CEA and CEACAM1 immunoprecipitated from SW948 lysate. Indeed, DC-SIGN-Fc but not control-Fc bound to both CEA and CEACAM1 on SW948 cancer cells (Fig. 2B). Immunoprecipitation of DC-SIGN ligands from SW948 lysate using DC-SIGN-Fc and blotting with the antibody that recognizes both CEA and CEACAM1 confirmed that DC-SIGN binds to both CEA and CEACAM1 on SW948 colorectal cancer cells (Fig. 2C).

Lewis^x and Lewis^y carbohydrates mediate binding of carcinoembryonic antigen and CEACAM1 to dendritic cell–specific intercellular adhesion molecule-3–grabbing nonintegrin. We examined the carbohydrates that are recognized by DC-SIGN on CEA and CEACAM1. DC-SIGN not only displays high affinity for mannose in high-mannose moieties (13) but also for fucose in carbohydrates such as Lewis^x and Lewis^y (20). The cancer cell lines SKBR3 and SW948 express high levels of Lewis^x and Lewis^y compared with the cancer cell lines HT29 and WIDR (Fig. 3A). The four tumor cell lines bound to DC-SIGN-Fc, but SKBR3 and SW948 bound much stronger to DC-SIGN-Fc than HT29 and WIDR, which corresponds with the Lewis^x and Lewis^y expression levels on these cells (Fig. 3A and B). DC-SIGN binding to the cancer cell lines was specific, because it was completely inhibited in the presence of anti-DC-SIGN antibodies (Fig. 3B). Interestingly, WIDR and HT29 cells express CEA and CEACAM1 to a similar extent as SW948 and SKBR3, respectively (Fig. 3A), indicating that expression levels of Lewis^x and Lewis^y rather than expression levels of CEA and CEACAM1 regulate binding of DC-SIGN. CEA is a highly glycosylated protein that contains Lewis^x and Lewis^y moieties in liver metastases of colorectal cancer (24–26, 35). Indeed, the presence of Lewis^x was detected on CEA and/or CEACAM1 captured from SW948 and SKBR3 cells, but not from HT29 and WIDR cells (Fig. 3C). Binding of DC-SIGN to both CEA and CEACAM1 derived from the cancer cell lines was determined in a novel ELISA-based binding assay that allows the study of DC-SIGN binding to native ligands. DC-SIGN-Fc bound much stronger to CEA and/or CEACAM1 isolated from SKBR3 and SW948 cells, than from HT29 and WIDR cells (Fig. 3D), suggesting that the level of Lewis^x on CEA and/or CEACAM1 regulates binding to DC-SIGN. Block with the Ca²⁺ chelator EGTA showed that binding of DC-SIGN-Fc was specific (Fig. 3D). To determine whether Lewis^x and Lewis^y on CEA and CEACAM1 mediate interactions with DC-SIGN, these proteins were captured from SW948 cell lysate and incubated with α1-3,4-fucosidase, an enzyme that specifically removes the fucose moiety from Lewis^x and Lewis^y that is recognized by DC-SIGN. The enzyme did not disrupt integrity of CEA and CEACAM1 derived from SW948 but substantially inhibited binding of DC-SIGN to CEA and to a lesser degree to CEACAM1 (Fig. 3E). This shows that DC-SIGN establishes interactions with CEA and CEACAM1 from colorectal cancer cells through the Lewis^x and Lewis^y moieties present on these proteins. Binding of DC-SIGN to other carbohydrates on CEACAM1, such as high-mannose moieties, may explain the residual binding activity of DC-SIGN to CEACAM1 after removal of Lewis^x and Lewis^y.

Cellular dendritic cell–specific intercellular adhesion molecule-3–grabbing nonintegrin binds carcinoembryonic antigen and CEACAM1. DC-SIGN forms tetramers within the cell membrane (13) and is present in clusters on immature dendritic cells (36), which may influence its function compared with recombinant DC-SIGN. Therefore, we studied the adhesion of cellular DC-SIGN to CEA and CEACAM1 derived from SW948 cells. Beads coated with CEA and CEACAM1 or with CEA alone bound to K562-DC-SIGN cells, but not to K562 cells (Fig. 4A and B), showing that cellular DC-SIGN on K562-DC-SIGN mediates binding to colorectal cancer–derived CEA and CEACAM1. To investigate the involvement of the C-type lectin domain in binding to CEA and CEACAM1 in more detail, we used DC-SIGN mutants in which amino acids essential in positioning of the Ca²⁺ ions and ligand binding have been changed (30). Substitution of these essential amino acids results in complete loss of adhesion to SW948 CEA and CEACAM1, showing that the C-type lectin domain of DC-SIGN mediates adhesion (Fig. 4A and B). Monocyte-derived dendritic cells express high levels of endogenous DC-SIGN. On these cells, DC-SIGN is the major receptor that binds glycoconjugates of Lewis^x and Lewis^y (20, 21). Because Lewis^x and Lewis^y carbohydrates mediate binding of CEA and CEACAM1 to DC-SIGN, we investigated the adhesion of SW948-derived CEA and CEACAM1 to immature and mature dendritic cells. Immature dendritic cells strongly bound to beads coated with either CEA and CEACAM1 or CEA alone (Fig. 4C). Adhesion could be blocked with anti-DC-SIGN antibodies (Fig. 4C), showing that immature dendritic cells bind colorectal cancer–derived CEA and CEACAM1 through DC-SIGN. Dendritic cell maturation strongly reduced binding of CEA and CEACAM1 to dendritic cells (Fig. 4C and D), which may be explained by down-regulation of DC-SIGN on mature dendritic cells (Fig. 1A). The low level of DC-SIGN-dependent binding to CEA and CEACAM1 may be insufficient to support adhesion of mature dendritic cells to tumor cells, because cellular interactions require cytoskeletal rearrangements and therefore strong interactions (Fig. 1B).

Dendritic cell–specific intercellular adhesion molecule-3–grabbing nonintegrin binds carcinoembryonic antigen on primary colorectal cancer cells but not on normal colon epithelium. Glycosylation of CEA on colon epithelial cells drastically changes upon malignant transformation. Compared with CEA from normal colon epithelial cells, fucosylation increases on CEA from colorectal cancer cells, which results in up-regulation of Lewis^x moieties and de novo expression of Lewis^y moieties on tumor-associated CEA (20, 24–26). Here, we show in four of five patients with colorectal cancer that CEA and/or CEACAM1 contain high levels of Lewis^x on malignant colon tissue (Fig. 5A). Interestingly, of those four patients, three exhibited a strong increase of Lewis^x on CEA and/or CEACAM1 upon malignant transformation (Fig. 5A). Furthermore, in three of these patients Lewis^y carbohydrates were up-regulated on CEA and/or CEACAM1 from tumor colon epithelium compared with normal colon epithelium (Fig. 5B). To investigate binding of DC-SIGN to CEA and CEACAM1 independently, we immunoprecipitated CEA and CEACAM1 with an antibody that recognizes both proteins, separated them using SDS-PAGE, and immunoblotted with DC-SIGN-Fc. DC-SIGN did not bind to CEA from normal colon tissue but did bind to CEA from malignant colon tissue in these patients, whereas DC-SIGN interacted equally well with CEACAM1 derived from normal and malignant colon tissue (Fig. 5C). The expression levels of CEA on colorectal cancer cells compared with normal colon epithelial cells may influence binding of DC-SIGN. Therefore, we examined the concentration of CEA by sandwich ELISA and found equal levels of CEA in the lysates of normal and tumor colon epithelium (Fig. 5D). To examine whether immature dendritic cells use DC-SIGN to specifically recognize CEA on colorectal cancer cells, we analyzed binding of immature dendritic cells to CEA derived from either normal or tumor colon epithelium from the same patient. Immature dendritic cells bound CEA derived from tumor but not from normal colon epithelium, and binding was mediated by DC-SIGN (Fig. 5E). Thus, tumor-specific glycosylation of CEA enables DC-SIGN on immature dendritic cells to specifically interact with this antigen on colorectal cancer cells. In contrast, DC-SIGN binds CEACAM1 from both normal and tumor colon epithelium.

Immature DC-SIGN⁺ dendritic cells are present within primary colorectal cancer. DC-SIGN is expressed on immature dendritic cells within normal colonic mucosa (37). However, the presence of DC-SIGN⁺ cells has not been investigated in colorectal cancer; therefore, we analyzed expression of DC-SIGN on normal and malignant colon tissue from the same patient. We detected DC-SIGN expression on cells with dendritic cell–like morphology not only within the mucosa of normal colon tissue (Fig. 6A-E), as has been reported previously (37) but also within cancer colon tissue (Fig. 6F-J). To investigate whether these DC-SIGN⁺ cells may associate with the epithelial cells, we analyzed coexpression of DC-SIGN with CEA and Lewis^x in normal and cancer colon tissue. CEA and Lewis^x are expressed on both normal and tumor colon epithelium (Fig. 6A,, B, F, and G). Strikingly, DC-SIGN⁺ cells within tumor colon tissue were closely associated with CEA- and Lewis^x-expressing colorectal cancer cells, suggesting that interactions between DC-SIGN⁺ cells and colorectal cancer cells may occur in situ (Fig. 6F and G). As has been described previously (38), we found higher expression levels and less pronounced apical polarization of CEA and Lewis^x on malignant colon tissue compared with normal colon tissue (Fig. 6A,, B, F, and G). Thus, increased expression and loss of polarization of CEA on colorectal cancer cells in addition to altered glycosylation that results in increased levels of Lewis^x and Lewis^y on CEA of colorectal cancer cells may facilitate interactions with DC-SIGN⁺ cells. However, within normal colon tissue DC-SIGN⁺ cells are also in close proximity with the epithelial cells, indicating that interactions between DC-SIGN⁺ cells and normal colon epithelium may also occur, probably driven by interactions with CEACAM1 (Fig. 6A and B).

In situ, DC-SIGN is expressed on dendritic cell populations within dermis; mucosa; and secondary lymphoid tissue such as tonsil, lymph node, and spleen (11, 12). DC-SIGN is also present on dendritic cells in tumors, such as myxofibrosarcoma and melanoma (9, 10). To analyze the cell type that expressed DC-SIGN within colon cancer, we examined coexpression of DC-SIGN with dendritic cells and macrophage markers. DC-SIGN⁺ cells expressed CD68 intracellularly in a patch-wise pattern (Fig. 6H). The distribution pattern of CD68 resembled that of dendritic cells but contrasted with that of macrophages that have an overall cytoplasmic and surface expression of CD68 (39). This indicates that these DC-SIGN⁺ cells in colorectal cancer tissue are of the dendritic cell lineage. DC-SIGN⁺ cells also expressed high levels of MHC class II that is abundantly present on dendritic cells and other antigen presenting cells (data not shown). We examined expression of CD83 to analyze the phenotype of DC-SIGN⁺ dendritic cells within colorectal cancer. DC-SIGN did not colocalize with the dendritic cell maturation marker CD83, showing that colorectal cancer contains DC-SIGN⁺ dendritic cells with an immature phenotype (Fig. 6I). DC-SIGN⁺ cells of the surrounding normal colon tissue were of similar phenotype and also expressed CD68 and MHC class II but not CD83 (Fig. 6C-D; data not shown). Expression of DC-SIGN in both normal and cancer colon tissue partially overlapped with the macrophage marker CD163, indicating that part of the DC-SIGN⁺ cells are antigen-presenting cells that contain characteristics of both immature dendritic cells and macrophages (Fig. 6E and J).

Interactions of tumor cells with inflammatory cells affect tumor growth and metastasis formation. Tumor epithelial cells of breast carcinoma produce macrophage inflammatory protein-3α (MIP-3α), a chemokine that attracts immature dendritic cells, and indeed, immature dendritic cells in contrast to mature dendritic cells are present within tumor beds of breast cancer (2, 3). We have shown that immature dendritic cells bind to breast and colorectal cancer through glycosylation-dependent interactions of DC-SIGN with CEA and CEACAM1 that may enable cross-talk between immature dendritic cells and tumor cells.

Previously, we have described that ICAM-2 on endothelial cells (19) and ICAM-3 on T cells (11) are cellular ligands of DC-SIGN. Here, we have identified that DC-SIGN recognizes the immunoglobulins CEA and CEACAM1. CEA is present on most normal epithelia and has retained or enhanced expression after tumor development. Strikingly, DC-SIGN strongly bound CEA obtained from both colon cancer cell lines and primary colorectal cancer epithelium but did not bind to CEA derived from normal colon epithelium. Glycosylation is deregulated in experimental and human carcinomas resulting in expression of distinct carbohydrate profiles on normal and malignant tissue (40). Colorectal cancer tissue expresses higher levels of Lewis^x and Lewis^y antigens than normal colon tissue (40). These Lewis antigens are present on CEA isolated from colorectal cancer but not or reduced on CEA from normal colon epithelial cells. Increased levels of Lewis^x and Lewis^y on tumor CEA compared with normal CEA may explain the selective recognition of tumor CEA by DC-SIGN. Indeed, Lewis^x and/or Lewis^y carbohydrate moieties on CEA mediate binding to DC-SIGN, as shown by sensitivity of the DC-SIGN-CEA interaction to fucosidase treatment. Although Lewis antigens may also arise on other proteins during tumor development, DC-SIGN only binds CEA and CEACAM1 on colon cancer cells. This indicates that either the protein backbone of CEA and CEACAM1 or the abundance and/or spacing of Lewis^x and Lewis^y is important for binding to DC-SIGN. CEACAM1 is a close homologue of CEA that is also present on normal epithelia, that is overexpressed in gastric adenocarcinoma (41), and stably expressed in breast carcinoma (42) but down-regulated in endometrial and colorectal cancer (43, 44). In contrast to CEA, DC-SIGN binds equally well to CEACAM1 from normal and malignant colon epithelium. Binding of DC-SIGN to CEACAM1 is more resistant to fucosidase treatment that disrupts binding of DC-SIGN to Lewis antigens. This suggests that other carbohydrates that are not changed upon malignant transformation mediate interactions between DC-SIGN and CEACAM1. Future mass spectrometry analysis of CEA and CEACAM1 is required to analyze the DC-SIGN binding carbohydrates on these antigens in more detail. Thus, glycosylation rather than expression may regulate binding of DC-SIGN to ligands, and glycosylation may be a major way by which immature dendritic cells recognize colorectal cancer cells.

Previously, it has been shown that immature but not mature dendritic cells bind to breast cancer cells on tissue slides (2). We have shown in cell-cell adhesion assays that immature dendritic cells strongly bind to breast and colorectal cancer cells and that DC-SIGN is involved in the interaction. Adhesion of immature dendritic cells to colorectal cancer cells in contrast to breast cancer cells was not completely dependent on DC-SIGN. This indicates that besides the interaction of DC-SIGN with CEA and CEACAM1 other receptors may be involved in binding of immature dendritic cells to colorectal cancer cells. Dendritic cell maturation completely abolished dendritic cell-tumor cell adhesion and strongly reduced binding of colorectal cancer cell-derived CEA and CEACAM1. This may be due to the lower expression of DC-SIGN and possibly other tumor-binding receptors on mature dendritic cells. Interactions of DC-SIGN with CEA and CEACAM1 may enable DC-SIGN⁺ dendritic cells to associate with tumor cells in vivo. Indeed, high levels of DC-SIGN⁺ dendritic cells are present within colorectal cancer, and many of these dendritic cells are in close contact with tumor cells. We also observed high levels of DC-SIGN⁺ dendritic cells with similar phenotype in the mucosa of normal colon. This indicates that DC-SIGN⁺ dendritic cells may penetrate tumor beds of colorectal cancer from surrounding normal tissue. Indeed, it has been reported that tumor cells of breast cancer produce chemokines such as MIP-3α that attract immature dendritic cells (2, 45). In addition, CD1a⁺ dendritic cells have been detected at equal levels in normal colon epithelium and colorectal cancer (3). However, no coexpression of CD1a and DC-SIGN was detected, indicating that these markers stain distinct dendritic cell subsets in colorectal cancer (data not shown). Because DC-SIGN binds CEACAM1 derived from normal colon epithelium, DC-SIGN⁺ dendritic cells within normal colon mucosa may also interact with the epithelial cells. It has been reported that dendritic cells are able to send out dendrites through the epithelium to sample gut-derived bacteria (46). Possibly, binding of DC-SIGN to CEACAM1 contributes to the interactions of dendritic cells with epithelial cells that are required for the sampling of gut-derived bacteria.

DC-SIGN⁺ dendritic cells of colorectal cancer did not express the maturation marker CD83, indicating that these dendritic cells are of immature phenotype. Previously, it has been described that immature dendritic cells are present intratumorally, whereas mature dendritic cells are located peritumorally (2, 3). Possibly, expression of DC-SIGN specifically on immature dendritic cells enables these cells but not mature dendritic cells to associate with colorectal cancer cells. The presence of immature dendritic cells within breast cancer does not provide immune protection, because increased staining for the immature dendritic cells markers CD1a and S100 did not correlate with tumor grade nor with survival of breast cancer patients (47, 48). This may indicate that these immature dendritic cells are involved in homeostasis control of breast cancer tissue. Indeed, immature dendritic cells are incapable of priming vigorous T-cell responses and may induce T cell tolerance instead (8). The immature phenotype of the DC-SIGN⁺ dendritic cells in colorectal cancer may indicate that they are also involved in tolerance induction to the tumor cells. In contrast, peritumoral mature dendritic cells have been found to associate with proliferating CD4 and CD8 T cells (3, 49). The number of these mature dendritic cells (50) and tumor-infiltrating lymphocytes (51, 52) correlates with metastasis-free and overall survival of breast and colorectal cancer patients. This suggests that these mature dendritic cells are able to stimulate protective T cell responses against the tumor cells.

Tumor cells have developed strategies to down-regulate immune responses. In particular, tumor cells target dendritic cells that are pivotal in the activation of T cells and release VEGF, prostanoids, TGFβ, IL-10, and IL-6, which prevent differentiation of dendritic cell precursors and maturation of dendritic cells (5–7). We have shown that triggering of DC-SIGN with the cell wall component ManLAM of M. tuberculosis inhibits dendritic cell maturation and induces production of the immunosuppressive cytokine IL-10 (15). Similarly, colorectal cancer cells may employ CEA to prevent maturation of dendritic cells and to induce dendritic cells with a tolerogenic phenotype through interactions with DC-SIGN. In this context, it is interesting that CEA in contrast to CEACAM1 is a secreted product of tumor cells, and that CEA from colorectal cancer cells but not from normal colon cells is recognized by DC-SIGN. Thus, in parallel to M. tuberculosis that secrete large amounts of the tolerance-inducing factor ManLAM at the site of infection, colorectal cancer cells may provide a tolerogenic micromilieu at the tumor site through the release of CEA. Tumor-derived CEA that binds to DC-SIGN on immature dendritic cells and to a lesser extent on mature dendritic cells may inhibit these cells to properly activate antitumor T-cell responses.

We and others have shown that DC-SIGN functions as an internalization receptor on immature dendritic cells (29, 53). Antibodies that target DC-SIGN are taken up by immature dendritic cells and peptides of these antibodies are presented to IgG1-specific T cells (29). Because tumor cells are able to release CEA and soluble CEA can be detected in the serum of cancer patients, DC-SIGN may be involved in the internalization of CEA by immature dendritic cells for presentation to T cells. CEA-specific T cells have been detected in colorectal cancer patients in contrast to healthy controls (54), indicating that CEA is processed and presented to T cells by antigen-presenting cells such as dendritic cells. A better understanding of endogenous CEA targeting and processing may contribute to the development of more adequate dendritic cell-based vaccination strategies directed at CEA.

Grant support: Dutch Scientific Research NWO grant 901-07-220 (K.P.J.M. van Gisbergen) and Dutch Cancer Society Koningin Wilhelmina Fonds grant VU2002_2598 (C.A. Aarnoudse).

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.