Glycosylation Influences the Lectin Activities of the Macrophage Mannose Receptor*

The mannose receptor (MR) is a heavily glycosylated endocytic receptor that recognizes both mannosylated and sulfated ligands through its C-type lectin domains and cysteine-rich (CR) domain, respectively. Differential binding properties have been described for MR isolated from different sources, and we hypothesized that this could be due to altered glycosylation. Using MR transductants and purified MR, we demonstrate that glycosylation differentially affects both MR lectin activities. MR transductants generated in glycosylation mutant cell lines lacked most mannose internalization activity, but could internalize sulfated glycans. Accordingly, purified MR bearing truncated Man5-GlcNAc2 glycans (Man5 -MR) or non-sialylated complex glycans (SA0-MR) did not bind mannosylated glycans, but could recognize SO4-3-Gal in vitro. Additional studies showed that, although mannose recognition was largely independent of the oligomerization state of the protein, recognition of sulfated carbohydrates was mostly mediated by self-associated MR and that, in SA0-MR, there was a higher proportion of oligomeric MR. These results suggest that self-association could lead to multiple presentation of CR domains and enhanced avidity for sulfated sugars and that non-sialylated MR is predisposed to oligomerize. Therefore, the glycosylation of MR, terminal sialylation in particular, could influence its binding properties at two levels. (i) It is required for mannose recognition; and (ii) it modulates the tendency of MR to self-associate, effectively regulating the avidity of the CR domain for sulfated sugar ligands.

The mannose receptor (MR) is a heavily glycosylated endocytic receptor that recognizes both mannosylated and sulfated ligands through its C-type lectin domains and cysteine-rich (CR) domain, respectively. Differential binding properties have been described for MR isolated from different sources, and we hypothesized that this could be due to altered glycosylation. Using MR transductants and purified MR, we demonstrate that glycosylation differentially affects both MR lectin activities. MR transductants generated in glycosylation mutant cell lines lacked most mannose internalization activity, but could internalize sulfated glycans. Accordingly, purified MR bearing truncated Man 5

-GlcNAc 2 glycans (Man 5 -MR) or non-sialylated complex glycans (SA 0 -MR) did not bind mannosylated glycans, but could recognize SO 4 -3-Gal in vitro.
Additional studies showed that, although mannose recognition was largely independent of the oligomerization state of the protein, recognition of sulfated carbohydrates was mostly mediated by self-associated MR and that, in SA 0 -MR, there was a higher proportion of oligomeric MR. These results suggest that self-association could lead to multiple presentation of CR domains and enhanced avidity for sulfated sugars and that non-sialylated MR is predisposed to oligomerize. Therefore, the glycosylation of MR, terminal sialylation in particular, could influence its binding properties at two levels. (i) It is required for mannose recognition; and (ii) it modulates the tendency of MR to self-associate, effectively regulating the avidity of the CR domain for sulfated sugar ligands.
The mannose receptor (MR) 6 was the first member of a family of four mammalian endocytic receptors to be discovered (1). These receptors share the same overall structure: an N-terminal cysteine-rich (CR) domain, followed by a fibronectin type II domain, several C-type lectin domains (CTLD; eight in the case of MR), and transmembrane and cytoplasmic regions (2)(3)(4). MR CTLD mediate binding to carbohy-drates terminating in Man, Fuc, or GlcNAc (5), and the MR CR domain recognizes sugars terminating in SO 4 -4/3-GalNAc or SO 4 -3-Gal (6,7). A natural cleavage product of MR can be found in supernatants of MR ϩ cells and in mouse serum (8 -10). In murine primary macrophages, this soluble form of MR (sMR) is generated from the cell-bound form by proteolytic cleavage by a metalloprotease (9). sMR comprises the whole extracellular region of the molecule and binds to both sulfated and mannosylated sugars (11). Our previous studies suggested that multimerization of the CR domain through the interaction of sMR with multivalent ligands for CTLD, such as mannan, leads to enhanced recognition of sulfated carbohydrates (11).
Natural endogenous ligands for MR CTLD include lysosomal hydrolases, tissue plasminogen activator, and myeloperoxidase (5) and thyroglobulin (12). Microbial ligands include endotoxin from selected strains of Klebsiella pneumoniae (11); capsular polysaccharide from selected strains of Streptococcus pneumoniae (11), Pneumocystis carinii (13,14), and Candida albicans (9,15); and human immunodeficiency virus (16,17). Only ligands of endogenous origin have been described for the CR domain, and they include the glycoprotein hormones produced by the anterior pituitary, lutropin and thyrotropin (7,18,19), and chondroitin sulfate and sulfated Le a and Le x (6). We have demonstrated binding of the MR CR domain to sulfated carbohydrate ligands expressed by subsets of macrophages and dendritic cells associated with B cell follicles in spleen and lymph nodes (20 -22). These CR domain ligands could act as molecular cues to direct sMR complexed with CTLD ligands to areas where B cell responses are generated (23). In agreement with these results, recombinant fusion proteins bearing the MR CR domain specifically target cells bearing CR domain ligands in vivo (24). This targeting depends on the lectin activity of the CR domain and, in agreement with the binding studies performed with sMR, is enhanced by multimerization (24).
Because of its binding properties and tissue distribution, MR can be placed at the interface between homeostasis and immunity. MR mediates efficient clearance of endogenous molecules (5,25) and could play a role in microbial recognition and antigen presentation (26 -29) and lymphocyte adhesion (30).
Even though MR is expressed by most tissue macrophages, it is not macrophage-restricted and liver sinusoidal cells and nonvascular endothelia, among other cell types, have been shown to express MR (31)(32)(33). In vitro, MR is expressed in human monocyte-derived and murine primary macrophages, but is absent in monocytes. In murine macrophages, MR is up-regulated in response to the Th2 cytokines interleukin (IL)-4, IL-13, and IL-10 (10,34,35) and is down-regulated by the Th1 cytokine interferon-␥ (3). Monocyte-derived human dendritic cells (26) and murine bone marrow-derived dendritic cells (36) express MR.
Previous studies have suggested that MR function is regulated in vivo. Kery et al. (37) observed large amounts of unbound MR after passing protein extracts from human placenta through a D-Man-coupled Sepharose column, suggesting that, in vivo, MR has functional diversity. Furthermore, MR purified from rat liver and lung have been shown to have different ligand binding activities, with the liver-specific form of MR displaying selective recognition of sulfated ligands (38). Additional studies suggested that this selectivity could be due to receptor dimerization in the liver (39). Intriguingly, an interaction has been detected between the lectin domain of L-selectin and MR expressed by lymphatic endothelia in human lymph nodes (30). These results indicate that MR function might be modified in a cell-or tissue-dependent manner.
Increasing evidence has shown that glycosylation is directly involved in the regulation of protein function, orientation, and organization during immune responses (40 -43). Therefore, we reasoned that carbohydrate modifications might play a role in modulating MR function. In this study, we investigated the relationship between the lectin activities of MR and its glycosylation and show, for the first time, that glycosylation, especially terminal sialylation, has a major effect on the functional specialization of MR.

EXPERIMENTAL PROCEDURES
Cells-Chinese hamster ovary (CHO) cells were obtained from frozen stocks stored at the Sir William Dunn School of Pathology and cultured in Ham's F-12K medium with 2 mM L-glutamine. LEC1 and LEC2 cells were from American Type Culture Collection (Manassas, VA) and cultured with ␣-minimal essential medium with ribonucleosides and deoxyribonucleosides (Invitrogen).
Retroviral Transduction of MR in CHO, LEC1, and LEC2 Cells-The CHO MR , LEC1 MR , and LEC2 MR cells were generated as described (10).
Detection of MR in Cell Lysates and Supernatants by Western Blotting-The cell lysates and supernatants were prepared and tested by Western blotting with anti-MR monoclonal antibody (mAb) 5D3 as described (10).
Preparation of Total Tissue Lysates-Tissues were collected and kept frozen at Ϫ70°C until used. Ice-cold lysis buffer (2% Triton X-100, 10 mM Tris-HCl (pH 8), 150 mM NaCl, 10 mM NaN 3 , and 10 mM EDTA) containing protease inhibitors (Roche Applied Science) was added to tissues (100 mg of tissue/ml of lysis buffer). Tissues were homogenized using a Polytron and maintained on ice for 30 min. Soluble material was selected by centrifugation at 500 ϫ g for 10 min at 4°C and further centrifugation at 35,000 rpm for 60 min at 4°C using a Ti-60 rotor.
Preparation of Membrane Lysates-Organs were homogenized using a Polytron in 10 mM Tris-HCl (pH 8), 2.5% (v/v) Tween 40, 150 mM NaCl, 10 mM NaN 3 , and 2 mM EDTA containing protease inhibitors. Nuclei were removed by centrifugation at 500 ϫ g for 15 min at 4°C. The supernatant was collected and centrifuged at 35,000 rpm for 60 min at 4°C using a Ti-60 rotor. The membrane pellet was washed twice with 10 mM Tris-HCl (pH 8), 10 mM NaN 3 , and 150 mM NaCl containing protease inhibitors, and proteins were solubilized in lysis buffer containing protease inhibitors. The lysate was clarified by centrifugation at 35,000 rpm for 60 min at 4°C in a Ti-60 rotor.
Protein Quantification-Protein was quantified using the BCA protein assay kit (Pierce).
Confocal Assays-To assess MR trafficking in CHO cells, CHO MR cells were incubated with Alexa 647-labeled dextran (10 g/ml) and Man-PAA-FITC (10 g/ml) in serum-free medium at 37°C for 15 min. One set of cells was fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS) for 20 min at room temperature, whereas another set was washed twice with the medium and incubated further for 75 min in medium with 10% fetal calf serum before fixation. Fixed cells were blocked and permeabilized in 0.1% Triton X-100 with 1% goat serum and 1% BSA in PBS for 30 min at room temperature. The permeabilized cells were incubated with anti-MR mAb 5D3 (10 g/ml) in blocking/ permeabilization buffer for 60 min at room temperature, washed five times with PBS, and incubated with Alexa 568-conjugated goat anti-rat IgG (10 g/ml) in PBS for an additional 60 min at room temperature.
To assess the subcellular distribution of MR, CHO MR , LEC1 MR , and LEC2 MR were incubated with Alexa 647-labeled transferrin (5 g/ml) for 15 min and fixed. MR was detected using mAb 5D3 and Alexa 488-conjugated goat anti-rat IgG (5 g/ml). Coverslips were mounted with fluorescent mounting medium. Immunofluorescence was analyzed on a Bio-Rad Radiance 2000 confocal laser scanning microscope with Lasersharp 2000 software, and images were processed using Adobe Photoshop.
Endocytosis Analysis-MR-mediated endocytosis was assessed by incubating cells in the presence of conjugated ligands (5 g/ml) for 2 h as described (10). At the end of the incubation, cells were harvested using trypsin/EDTA, fixed in 4% paraformaldehyde in PBS, and analyzed using a BD Biosciences FACSCalibur with CellQuest software.
Selection of Cells with Similar MR-mediated Endocytic Activity by Cell Sorting-CHO MR , LEC1 MR , and LEC2 MR cells (5 ϫ 10 6 ) were incubated with Alexa 488-labeled mAb 5D3 (10 g/ml) for 2 h at 37°C. Cells were harvested using trypsin/EDTA prior to analysis to ensure that only ligand internalization was assessed. Homogeneous populations of CHO MR , LEC1 MR , and LEC2 MR cells were obtained by cell sorting using a DakoCytomation MoFlo cell sorter with an argon ion laser.
MR Purification-A 0.5 ϫ 5-cm affinity column and a GammaBind Plus-Sepharose column (Amersham Biosciences) of the same size were prepared by packing 5 ml of mAb 5D3-cross-linked GammaBind Plus-Sepharose (Amersham Biosciences) and GammaBind Plus-Sepharose, respectively. After washing with 10 mM Tris-HCl (pH 8.0), 0.5% Triton X-100, 150 mM NaCl, and 10 mM NaN 3 , the cell culture supernatant was loaded onto the column (1 ml/min). After washing with the loading buffer, bound proteins were eluted using 0.5% diethylamine in 1% Triton X-100 (pH 11). 1-ml fractions were collected in 100 l of 1 M Tris-HCl (pH 7.0) for neutralization. All procedures were performed at 4°C. A similar procedure was used to purify MR from spleen and lung membrane lysates.
Ligand Binding Assays-96-Well Maxisorp plates (Nalge Nunc International, Naperville, IL) were coated overnight at 37°C with 50 l of 100 g/ml mannan, 10 g/ml Gal-BSA, or 10 g/ml SO 4 -3-Gal-PAA in PBS and washed twice with binding buffer (10 mM Tris-HCl (pH 7.5), 15 mM CaCl 2 , 150 mM NaCl, and 0.1% Tween 20) for 5 times; in the last wash, the plate was incubated at 25°C for 5 min for blocking. Serial dilutions of MR (50 l/well) were added to the plates and incubated at 25°C for 2 h to allow binding to occur. Plates were washed eight times with binding buffer. mAb 5D3 (10 g/ml) and alkaline phosphatase-conjugated goat anti-rat IgG, both diluted in binding buffer and incubated for 1 h at 25°C, were subsequently used to detect MR binding. Enzyme activity was developed with 50 l of p-nitrophenyl phosphate (1 mg/ml; Sigma) and read at 405 nm after 30 min at 37°C.
Gel Filtration-Gel filtration chromatography was carried out on an AKTA Superose 12 FPLC column (10 ϫ 300 mm) in binding buffer. The flow rate was 0.5 ml/min, and the absorbance of the eluant was monitored at 280 nm; 34 fractions (0.5 ml each) from each sample were collected at V 0 ϭ 7.7 ml.
Surface Plasmon Resonance-Binding experiments were performed by surface plasmon resonance on a Biacore 2000 instrument. All experiments were performed at 25°C. SO 4 -3-Gal-PAA-biotin and Gal-PAAbiotin were covalently coupled to the carboxymethylated dextran matrix on a research-grade CM5 sensor chip (Biacore) precoated with streptavidin using the Biacore amine coupling kit as directed by the manufacturer with the following modifications. After an activation step of 300 -600 s, MR in binding buffer without Tween 20 was injected for 300 -900 s at 10 l/min. The basic amine coupling buffer was 10 mM Tris-HCl (pH 7.5), 0.015 M CaCl 2 , and 150 mM NaCl, and the flow rate was set at 10 l/min. 70 l of 1ethyl-3(3-dimethylaminopropyl)carbodiimide hydrochloride/N-hydroxysuccinimide mixture was injected to activate the chip surface. Immediately after activation, 70 l of streptavidin was covalently coupled to the research-grade CM5 sensor chip via primary amine groups using the amine coupling kit. Then, 70 l of ethanolamine was applied to block the sensor chip surface. Finally, the "dips" were checked to confirm that the immobilization was homogeneous.
N-Linked Glycan Analysis-sMR bearing sialylated (SA-sMR) and non-sialylated (SA 0 -sMR) complex glycans were resolved by 6% SDS-PAGE, and the protein bands were excised. In situ release of N-glycans with peptide N-glycanase F (Oxford GlycoSciences, Abingdon, Oxon, UK) was carried out. The extracted glycans were labeled with the fluorophore 2-aminobenzamide (Oxford GlycoSciences) by reductive amination and processed by normal-phase HPLC using a low salt buffer system (45). The system was calibrated using an external standard of hydrolyzed and 2-aminobenzamide-labeled glucose oligomers to provide a dextran ladder from which the retention times for the individual glycans were converted to glucose units. These glucose units were compared with a data base of experimental values to obtain preliminary assignments for the glycans that were confirmed by digestion with exoglycosidases and by mass spectrometry. The following exoglycosidases were purchased from Glyko (Novato, CA): Arthrobacter ureafaciens sialidase (1-2 units/ml), bovine testis ␤-galactosidase (1 unit/ml), bovine kidney ␣-fucosidase (1 unit/ml), and S. pneumoniae hexosaminidase (1 unit/ml).
Assessment of Glycans Associated with MR by Lectin Blotting-To determine the sugars associated with MR purified from spleen and lung, a digoxigenin glycan differentiation kit (Roche Applied Science) was used following the manufacturer's instructions. Briefly, MR from spleen and lung were electrophoresed and transferred to nitrocellulose filters as described above. After blocking, filters were washed twice with Trisbuffered saline (50 mM Tris-HCl (pH 7.5) and 150 mM NaCl), and once with buffer A (50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM MgCl 2 , 1 mM MnCl 2 , and 1 mM CaCl 2 ). Filters were incubated with digoxigenin- conjugated lectins in buffer A. After washes with Tris-buffered saline, filters were incubated with an alkaline phosphatase-conjugated sheep anti-digoxigenin F(abЈ) 2 fragment (0.75 unit/ml). Bound enzyme was detected using 4-nitro blue tetrazolium chloride/5-bromo-4-chloro-3indolyl phosphate. Carboxypeptidase Y was used as a positive control for the lectins peanut agglutinin and Datura stramonium agglutinin.

Generation of MR Transductants in Glycosylation Mutant Cells:
LEC1 MR and LEC2 MR -To determine whether glycosylation influences MR function, three stable cell lines expressing MR (CHO MR , LEC1 MR , and LEC2 MR ) were generated by transducing CHO, LEC1, and LEC2 cells (44), respectively, using a retroviral vector encoding murine fulllength MR as described (10). LEC1 cells lack GlcNAc-glycosyltransferase activity, so N-linked carbohydrates are blocked at the Man 5 -Gl-cNAc 2 -Asn intermediate. LEC2 cells exhibit a drastic reduction in the transport of CMP-sialic acid into the Golgi compartment, and all carbohydrates lack terminal sialylation. Flow cytometry analysis demonstrated that MR expression in LEC1 MR cells was very low and that LEC1 MR and LEC2 MR cells were not homogeneous, as two different populations could be observed (data not shown). As most of the MR localizes in the early endosomal compartment and not at the plasma membrane, to normalize for MR expression, we selected for CHO MR , LEC1 MR , and LEC2 MR cells with similar ability to internalize labeled anti-MR mAb 5D3 (10) using cell sorting as described under "Experimental Procedures." Harvesting the cells using trypsin/EDTA prior to analysis ensured that ligand internalization and not cell association was being assessed in these assays because MR is trypsin-sensitive. After sorting, we obtained homogeneous cell populations suitable for our studies, even though the internalization of anti-MR mAb by LEC1 MR and LEC2 MR cells was still reduced compared with CHO MR cells (CHO MR /LEC1 MR /LEC2 MR ϭ 100:64:80) (Fig. 1). The mAb internalization data seemed to give a good indication of the levels of MR expressed by the transductants, as they correlated with the results obtained by Western blot analysis (see Fig. 3A).
Internalization of MR-specific Carbohydrates Is Affected by Glycosylation-To determine whether the specificity of MR can be altered by its expression in glycosylation mutant cells, the ability of CHO MR , LEC1 MR , and LEC2 MR cells to internalize the CTLD-specific endocytic tracer Man-PAA-FITC and the CR domain-specific endocytic tracer SO 4 -3-Gal-PAA-FITC was analyzed (Fig. 1). The multimeric nature of these synthetic tracers, in which several monosaccharides are conjugated to a high molecular mass carrier (PAA), enables these probes to be used as surrogate ligands for MR. None of the MR transductants endocytosed nonspecific ligands, such as Gal-PAA-FITC and rat IgG2a (data not shown), and no internalization of any of the endocytic tracers was observed in the case of CHO, LEC1, and LEC2 cells (data not shown). In all instances, the internalization of the glycoconjugate tracers was compared with that of anti-MR mAb 5D3, which provided an internal control for MR-dependent internalization independent of lectin activity. With respect to Man-PAA-FITC, LEC1 MR cells lost most of their endocytic activity (only 9% of mAb 5D3 internalization activity compared with 85% for CHO MR cells), whereas the sialylation-deficient LEC2 MR cells retained some (20% of mAb 5D3 internalization activity). Similar results were obtained with Man-BSA, which is another MR CTLD ligand (data not shown). Endocytosis of SO 4 -3-Gal-PAA-FITC was less affected by the altered pattern of glycosylation present in LEC1 MR and LEC2 MR cells. In these cells, internalization of SO 4 -3-Gal-PAA-FITC  15 g) from CHO, LEC1, LEC2, CHO MR , LEC1 MR , and LEC2 MR cells and a fraction of supernatants equivalent to the fraction of cell lysate used were electrophoresed on 6% SDS-polyacrylamide gel and transferred to filters. The presence of MR was detected using mAb 5D3 (2 g/ml) and horseradish peroxidase-conjugated goat-anti-rat IgG. B, purified SA-sMR (0.2 g), Man 5 -sMR (0.13 g), and SA 0 -sMR (0.29 g) were electrophoresed on a 6% SDS-polyacrylamide and detected by silver staining. Staining showed that these preparations contained a single protein with the expected relative molecular mass. C, the N-linked glycans attached to purified SA-sMR and SA 0 -sMR are consistent with the characteristics of host cell from which they are derived. The N-linked glycans of SA-sMR and SA 0 -sMR were released in-gel as described under "Experimental Procedures." After 2-aminobenzamide labeling, glycans from each sample were analyzed by normal-phase HPLC. Other aliquots were subjected to simultaneous digestion with the exoglycosidase arrays indicated. A. ureafaciens sialidase (ABS) releases ␣2-6-, ␣2-3-, and ␣2-8-linked nonreducing terminal sialic acids. Bovine testis ␤-galactosidase (BTG) hydrolyzes nonreducing terminal Gal␤1-3 and Gal␤1-4 linkages. Bovine kidney ␣-fucosidase (BFK) releases ␣1-2-linked nonreducing terminal fucose residues more efficiently than ␣1-3and ␣1-4-linked fucose residues. S. pneumoniae hexosaminidase (SPH) digests GlcNAc␤1-3Gal and GlcNAc␤1-6Gal. The schematics show the representation of glycan structures. SEPTEMBER 23, 2005 • VOLUME 280 • NUMBER 38 was quantified as 41% (LEC1 MR cells) and 62% (LEC2 MR cells) of mAb 5D3 internalization. These results indicate that it is the ability of MR to recognize carbohydrates and not its endocytic function that is altered in LEC1 MR and LEC2 MR cells.

Changes in Glycosylation Do Not Alter the Intracellular Distribution of MR-Expression of a lectin in glycosylation-mutant cells could lead
to changes in protein trafficking and/or distribution caused by protein misfolding, recognition of intracellular glycoproteins, or alterations in their interaction with the protein sorting machinery. To determine whether these processes could explain the altered recognition pattern observed in CHO MR , LEC1 MR , and LEC2 MR cells, we analyzed the subcellular distribution of MR in these cells by confocal microscopy. First, CHO MR cells were assessed for suitability as a system to investigate MR function by determining the fate of a mannosylated ligand after uptake. After a 15-min incubation of the cells with Man-PAA-FITC and dextran, co-localization of MR with both tracers was observed ( Fig. 2A). After a 75-min chase, Man-PAA-FITC was shown to co-localize with dextran in the lysosomes, whereas the distribution of MR was similar to that at 15 min (Fig. 2A). These results indicate that, as expected (5), in CHO MR cells, MR ligands dissociate from the receptor and are delivered to the lysosomal compartment for degradation.
When sorted CHO MR , LEC1 MR , and LEC2 MR cells were analyzed by immunofluorescence staining using mAb 5D3, the three cell populations were found to be relatively homogeneous. As MR has been described as a marker of the recycling endocytic compartment, its distribution in CHO MR , LEC1 MR , and LEC2 MR cells was compared with that of the endocytic tracer transferrin (Fig. 2B). After a 15-min incubation, MR co-localized with transferrin in the recycling compartment with a perinuclear localization in all three cell lines. These results indi-cate that glycosylation does not affect the subcellular distribution of MR.
Production of sMR Is Maintained in LEC1 MR and LEC2 MR Cells-MR is produced both as a cell-associated (cMR) and a soluble (sMR) form. sMR displayed similar carbohydrate binding properties and specificity compared with cMR (9). Western blotting was used to determine the relative molecular mass of MR and to test for sMR production in the glycosylation mutant cells. As shown in Fig. 3A, both cMR and sMR with the expected relative molecular masses could be detected in cell lysates and supernatants of CHO MR , LEC1 MR , and LEC2 MR cells, indicating that cleavage takes place regardless of the altered glycosylation. A direct correlation between the amount of cMR and sMR was observed, suggesting that the proteolytic processing of the receptor per se is not affected by glycosylation. In agreement with the internalization assays shown in Fig. 1, lower amounts of MR bearing truncated Man 5 -Glc-NAc 2 glycans (Man 5 -MR) and, to a lesser extent, SA 0 -MR were detected in both cell lysates and supernatants in this investigation (Fig. 3A), suggesting that lack of N-linked glycan processing might affect MR synthesis and/or stability.
Analysis of N-Linked Glycans Associated with sMR Produced by CHO MR and LEC2 MR Cells-To confirm that the changes in relative molecular mass observed by Western blotting were due to changes in glycosylation, sMR present in the supernatants of CHO MR , LEC1 MR , and LEC2 MR cells (SA-sMR, Man 5 -sMR, and SA 0 -sMR, respectively) were purified by mAb 5D3-Sepharose affinity chromatography as N-Glycans associated with SA-sMR and SA 0 -sMR FIGURE 4. SA 0 -sMR and Man 5 -sMR show impaired recognition of mannan and altered avidity for sulfated ligand in vitro. Binding of purified SA-sMR, Man 5 -sMR, and SA 0 -sMR to Gal-BSA (A), mannan (B), and SO 4 -3-Gal-PAA (C) was assessed using the binding assay described under "Experimental Procedures." Purified SA-sMR, Man 5 -sMR, and SA 0 -sMR were used at 0, 1, 2.5, 5, and 8 g/ml (0, 0.1, 1, 5, and 10 g/ml for Gal-BSA), and binding was detected using mAb 5D3 and alkaline phosphatase-conjugated goat antirat IgG.
described (11). SDS-PAGE analysis of these preparations showed that purified products consisted only of a single component (Fig. 3B), which was proven to be MR by dot blot analysis using mAb 5D3 (data not shown). Only CHO MR and LEC2 MR cells yielded enough sMR for glycan analysis. N-Linked glycans released from SA-sMR and SA 0 -sMR were analyzed by HPLC in combination with a series of exoglycosidase digestions ( Fig. 3C and TABLE ONE) as described (45). 75-80% of the N-linked glycans attached to SA-sMR terminating with sialic acid. In contrast, the dominant terminal monosaccharide on sugars released from SA 0 -sMR was demonstrated to be Gal (Fig. 3C and TABLE ONE). By comparison, SA-sMR had very similar glycan structures after removal of the sialic acids compared with those in the released glycan pool of SA 0 -sMR, indicating that lack of sialylation on the glycans attached to SA 0 -sMR is the only difference in glycan processing between SA 0 -sMR and SA-sMR.
sMR Produced by LEC1 MR and LEC2 MR Cells Display Altered Carbohydrate Binding Activity in Vitro-The endocytosis analysis could not distinguish whether carbohydrate modifications of MR itself or of other glycoproteins influenced the internalization of MR carbohydrate ligands. This is particularly important in LEC1 cells, where all cell-surface glycoproteins are associated with the carbohydrate structure Man 5 -GlcNAc 2 -, a putative ligand for the CTLD of MR. Therefore, we assessed the binding properties of the different glycoforms of MR using purified SA-sMR, Man 5 -sMR, and SA 0 -sMR in solid-phase binding assays as described (11). Increasing amounts of purified sMR were absorbed to carbohydrate-precoated plates and detected using mAb 5D3. As expected, none of the three sMR bound to Gal-BSA (Fig. 4A). Although SA-sMR showed dose-dependent binding to mannan consistent with published results, we were unable to detect binding of Man 5 -sMR and SA 0 -sMR to this CTLD ligand (Fig. 4B). With respect to the CR domain ligand SO 4 -3-Gal-PAA, the three preparations of purified sMR (SA-sMR, Man 5 -sMR, and SA 0 -sMR) showed efficient binding activities. Data obtained at low protein concentration suggest that there might be differences between these proteins regarding sulfated sugar recognition; and although reduced binding was obtained with Man 5 -sMR, enhanced binding was detected in the case of SA 0 -sMR compared with SA-sMR (EC 50 ϭ 4.9 M (SA 0 -sMR), 7.3 M (SA-sMR), and 9.9 M (Man 5 -sMR)) (Fig. 4C).
Binding of Sulfated Carbohydrate Is Mediated by Oligomerized Forms of sMR-Because multimerization of the CR domain seems to play a role in efficient ligand binding (24,39), we determined the aggregation levels of the SA-sMR and SA 0 -sMR preparations used in the solid-phase binding assays shown in Fig. 4. Gel filtration analysis demonstrated that the amount of multimerized protein was substantially increased in the absence of terminal sialic acid (SA 0 -sMR) (data not shown). In view of these results and to determine how multimerization could influence MR binding properties, we analyzed the lectin activities of monomeric and multimeric forms of newly purified SA-sMR and SA 0 -sMR (Fig. 5A) resolved by gel filtration chromatography (Fig. 5B). SA 0 -sMR resolved into three forms: a slow eluting form (fraction B1ϩC1) and two fast eluting forms (fractions B6 and B9). Under similar conditions, the majority of SA-sMR moved as a single entity with an elution profile similar to that of the slow moving component of SA 0 -sMR. The three different eluates obtained from SA 0 -sMR probably represent different geometrical arrangements of the same protein because a single MR band was identified by electrophoresis under reducing conditions. According to the size markers, fraction B1ϩC1 is likely to contain monomeric MR, whereas fractions B6 and B9 probably contain multimeric structures. Therefore, by comparison and in agreement with the results obtained with earlier preparations, SA 0 -sMR contained a higher proportion of multimeric MR compared with SA-sMR (Fig. 5B).
We investigated the ability of the different fractions of sMR obtained by gel filtration to bind Man-BSA (data not shown) and SO 4 -3-Gal-PAA using a Biacore assay (Fig. 5C). Although both fractions B1ϩC1 (monomeric) and B6 and B9 (multimeric) of SA-sMR displayed Man-BSA binding activity (data not shown), no activity was observed in the case of monomeric SA 0 -sMR. Intriguingly, a minor binding activity was observed in the case of oligomeric SA 0 -sMR (data not shown). In contrast to mannose recognition, a clear correlation between the oligomerization level of sMR and its ability to bind sulfated carbohydrates was observed for SA-sMR and SA 0 -sMR. In both cases, fractions B6 and B9 showed at least a 100-fold increase in response units compared with their corresponding monomer regardless of reduced amount of SA 0 -sMR available for these assays (Fig. 5C). It is noteworthy that monomeric SA-sMR (fraction B1ϩC1) still bound to the sulfated carbohydrate, but monomeric SA 0 -sMR did not, suggesting that, besides the global role of terminal sialic acids in the geometric arrangement of MR, the single sialylated N-linked glycan associated with the CR domain might also contribute to some extent to its recognition of sulfated glycans.
MR Is Differentially Glycosylated in Vivo-Despite the strong evidence provided by our results for a role for glycosylation in controlling MR function, no information is available regarding MR glycosylation in vivo. To determine whether MR could be differentially glycosylated in different tissues, we analyzed the relative molecular mass of MR in total tissue lysates of liver, lung, thymus, and spleen, organs in which MR expression has been described previously (32). As shown in Fig. 6A, although MR from liver and spleen showed a very similar relative molecular mass, a slightly smaller relative molecular mass was observed in the case of MR from lung, and the opposite was true for MR from thymus. As no differentially spliced forms of MR have been described, we considered this observation suggestive of differential glycosylation and performed lectin blot analysis of MR purified from membrane preparations of lung and spleen (Fig. 6B). The lectins used were Galanthus nivalis agglutinin (binds Man␣1-2, Man␣1-3, and Man␣1-6), D. stramonium agglutinin (binds Gal␤1-4GlcNAc), Maackia amurensis agglutinin (binds NeuNAc␣2-3Gal), Sambucus nigra agglutinin (binds Neu-NAc␣2-6Gal/GalNAc), and peanut agglutinin (binds Gal␤1-3GalNAc). This analysis showed that glycans attached to spleen MR contained mostly ␣2-6-linked terminal sialic acids and less ␣2-3-linkage and that there was no terminal ␣1-3-, ␣1-6-, or ␣1-2-linked terminal Man as shown by the negative G. nivalis agglutinin reactivity. These data are consistent with the results from the analysis of glycans associated with spleen MR we performed by exoglycosidase digestions and HPLC. 7 MR isolated from lung had more terminal sialic acid in the ␣2-3-configuration than in the ␣2-6-configuration. Furthermore, terminal mannose and probably Gal␤1-4GlcNAc were also present in lung MR. These results indicate that MR can undergo differential glycosylation in vivo.

DISCUSSION
In this study, we have demonstrated that glycosylation, especially terminal sialylation, can differentially regulate the dual lectin activities of MR. Our observations can be summarized as follows. (i) Glycosylation does not substantially affect MR biosynthesis, proteolytic processing, and trafficking; (ii) lack of terminal sialic acid leads to reduction of mannosylated ligand binding activity; and (iii) the presence of neutral glycans promotes the formation of multimeric structures of MR and indirectly increases avidity for sulfated carbohydrate ligands (TABLE TWO).
In MR glycans contribute ϳ15 kDa to its relative molecular mass, and there is conservation of seven potential N-linked glycosylation sites between human and mouse (Fig. 7, A and B). Apart from one N-linked oligosaccharide attached to the CR domain, the rest are distributed along CTLD, especially CTLD1-3 and CTLD6 -8. This conservation of N-linked glycosylation sites is suggestive of an important contribution of N-linked glycans in MR biology. Our results support this hypothesis, as lack of terminal sialic acid deprived the molecule of its most characteristic feature, mannose recognition. This conclusion is based on two distinct types of assays: endocytosis assays using glycosylation mutant cell lines and solid-phase binding assays (enzyme-linked immunosorbent assay and Biacore assay). The proteins used for the in vitro studies were purified using an affinity chromatography procedure that does not depend on the lectin binding properties of the molecule (11). This procedure involves the use of anti-MR mAb 5D3 (10) and has been successfully used previously to purify sMR from supernatants of MR transductants. sMR purified using this method shows the characteristic MR binding properties and does not display any major intrinsic tendency to self-association (11). The minor disagreement between the cell-based and in vitro assays observed in the case of LEC2 MR cells and purified SA 0 -sMR (we detected residual internalization activity for Man-PAA-FITC in the case of LEC2 MR cells) indicates that, unlike purified SA 0 -sMR, non-sialylated MR at the cell surface partially retains its ability to FIGURE 6. Tissue-specific glycosylation of MR. A, the relative molecular mass of MR in different mouse tissues is suggestive of altered glycosylation patterns in vivo. Total tissue lysates from liver (100 g), lung (200 g), thymus (100 g), and spleen (50 g) were electrophoresed under nonreducing conditions and transferred to nitrocellulose. The presence of MR was analyzed using anti-MR mAb 6F3 (10 g/ml) (10) and horseradish peroxidase-conjugated goat anti-rat IgG. The same pattern was obtained with several anti-MR mAbs using individual gels (data not shown). The relative molecular mass of MR could be graded as follows: lung Ͻ liver ϭ spleen Ͻ thymus. B, differential lectin reactivity of MR purified from lung and spleen. Equivalent amounts of MR purified from lung (Lu) and spleen (Sp), determined empirically through their reactivity with digoxigeninlabeled mAb 5D3, were electrophoresed; transferred to nitrocellulose; and incubated with several lectins as described under "Experimental Procedures." Although lung MR readily bound G. nivalis agglutinin (GNA), M. amurensis agglutinin (MAA), and (weakly) D. stramonium agglutinin (DSA), spleen MR reacted with S. nigra agglutinin (SNA). PNA, peanut agglutinin. bind mannosylated ligands. Intriguingly, a small amount of binding to Man-BSA was observed in the case of oligomeric SA 0 -sMR using the Biacore assay (data not shown). Therefore, it is possible that the rapid internalization of MR clustered at the plasma membrane could mediate endocytosis of weak ligands, mannose in this instance. A requirement of terminal sialic acid for mannose recognition is consistent with a report of an inactive precursor for MR (46). Pontow et al. (46) showed that MR is initially synthesized as an inactive precursor form and that mannose binding activity develops following exit from the endoplasmic reticulum.
It is difficult to envisage the mechanism behind a sialic acid dependence for mannose binding for MR because no N-linked glycosylation sites were found in the CTLD mainly responsible for sugar recognition: CTLD4 and CTLD5 (Fig. 7) (47). Nevertheless, the same study found that CTLD6 -8 are required for high affinity binding, and it is possible that, in the context of the whole extracellular region of MR, the glycosylation of these "supporting" domains could affect the function of the prime actors.
By comparison, the glycosylation of MR did not substantially affect the final outcome of its binding to sulfated carbohydrates. As shown in Fig. 4C, equivalent binding levels for all three purified sMR were observed under saturating conditions. However, glycosylation seemed to affect the kinetics of binding, as reduced and enhanced association constants were observed for Man 5 -sMR and SA 0 -sMR, respectively. In the case of Man 5 -sMR, these results correlated with the reduced uptake of sulfated carbohydrates observed in LEC1 MR cells (Fig. 1A). With respect to SA 0 -sMR, the deleterious effect of lack of sialylation on binding of monomeric sMR observed by Biacore assay (Fig. 5C) seemed to be compensated for by the increased avidity obtained by oligomerization and multipresentation of CR domains (11,24). Indeed, our modeling of the interaction between two CR domains and the glycans attached to human chorionic gonadotropin (Fig. 7C) indicates that the two sulfated glycans, spaced ϳ35 Å apart, would optimally fit into two neighboring CR domain binding pockets. This specific geometric arrangement might therefore be physiologically advantageous to the interaction between glycosylated hormones and MR, as suggested for MR isolated from rat liver (39). At the cellular level, SA 0 -MR retained effective internalization of SO 4 -3-Gal, although slightly lower compared with SA-MR (Fig. 1). This reduced efficiency seems contradictory because SA 0 -MR appeared to have a higher affinity for sulfated ligands in vitro (compare Figs. 1 and 4C). However, this result could be explained by a slower dissociation rate of the SA 0 -MR⅐SO 4 -3-Gal complex, leading to a slower MR-ligand separation, lower availability of free MR on the cell surface, and therefore a lower internalization of sulfated ligands.
Previous studies have suggested that sialic acid, by virtue of both its location at the terminal position on glycans and its net negative charge a Ϫ(ϩ), weak binding; ϩ(ϩ), increased binding at low protein, concentrations; ϩϩ(ϩ), level of multimerization varies among preparations. at physiological pH, serves as a potentially important regulator of molecular and cellular interactions (48,49). Our data suggest that the negative charge inherent to the carboxylate group at C-1 of sialic acid (at physiological pH) may minimize spontaneous self-association of MR simply by electrostatic repulsion or steric hindrance, as has been demonstrated for the neural cell adhesion molecule (50) and the CD8␣␤ coreceptor (41) or that it might have a role in preventing nonspecific interaction with other molecules (51)(52)(53). Sialic acids form one of the most important families in biology. N-Linked glycan sialylation regulates the ability of at least three I-type lectins (CD22, CD33, and sialoadhesin) to interact with their ligands (40,54,55). Physiologically, our results suggest that alteration of the expression/function of sialyltransferases or sialidases in vivo can provide a means for the cell to regulate MR-mediated functions as described in other cell systems. For instance, in B cells, there is a cell cycle-dependent ␣2-6-sialyltransferase that is induced upon cell activation and an endogenous sialidase activity that could affect interaction with T cells (56). There is no direct evidence of modulation of glycosyltransferase and glycosidase activities upon macrophage activation. However, the reduced presence of sialic acid upon activation of mouse peritoneal macrophages by trehalose dimycolate and lipopolysaccharide has been reported (57). Many macrophage glycoproteins, such as ferritin, scavenger receptor class A, HLA-DR molecules, and macrosialin, have been shown to undergo differential glycosylation in response to various stimuli, such as Th1 (interferon-␥) and Th2 (IL-4) cytokines and phagocytic stimuli (58 -61). We have not observed obvious changes in the relative molecular mass of MR in murine macrophages cultured in the presence of interferon-␥, IL-4, lipopolysaccharide, or phorbol 12-myristate 13-acetate (9) (data not shown), but the altered glycosylation pattern observed in MR from murine lung argues in favor of tissue-dependent glycosylation. Thus, a major challenge in the future is to understand the tissue-and/or cell-dependent glycosylation of MR and how it relates to MR-ligand interactions.