Structural requirements for high affinity binding of complex ligands by the macrophage mannose receptor.

The mannose receptor of macrophage and hepatic endothelial cells discriminates between endogenous and exogenous sugar-bearing structures. Previous competition studies have indicated that the receptor binds the monosaccharides mannose, fucose, and N-acetylglucosamine but displays much higher affinity for multivalent oligosaccharides, such as those found on the surface of potentially pathogenic microorganisms. The hydrodynamic properties of the receptor have been examined, revealing that the receptor is a monomer. This result suggests that multiple carbohydrate recognition domains (CRDs) in the extracellular domain of a single receptor polypeptide cooperate to achieve high affinity binding of complex ligands. In order to determine the importance of individual CRDs, properties of receptor segments containing groups of CRDs expressed in insect cells have been examined. The results indicate that two of the CRDs (4 and 5) form a protease-resistant, ligand-binding core but that five CRDs in tandem (4-8) are required to match the affinity of the intact receptor for yeast mannan. A consequence of the organization of the receptor is that both valency and geometry of glycoconjugates are important determinants of binding affinity.

The mannose receptor is a type I transmembrane protein with an extracellular portion consisting of several cysteinerich domains (8). These include a fibronectin type I1 repeat (9) and eight domains that are related in sequence to the Ca2+-dependent carbohydrate-recognition domains (C-type CRDs)' of other animal lectins (10). C-type CRDs are approximately 130 amino acids long and are characterized by a pattern of 14 invariant residues and 18 residues that are conserved in character (11,12). C-type CRDs have been found in a large number of proteins, in association with a variety of effector domains (13, 14). They have specificities for various different saccharides. Examples of proteins containing C-type CRDs are the asialoglycoprotein receptor, which mediates endocytosis of serum glycoproteins (15), serum mannosebinding protein, which mediates antibody-independent defense against pathogens (16), and the selectin cell adhesion molecules (17). C-type CRDs can be found at either end of a polypeptide or at internal positions. The mannose receptor is the only protein known to contain more than one CRD within a single polypeptide (8).
Experiments in which truncated forms of the mannose receptor were expressed in fibroblasts have shown that the NHz-terminal cysteine-rich domain, the fibronectin type I1 repeat, and the first three CRDs are not essential for binding and endocytosis of glycoprotein ligands (18). A truncated receptor with an extracellular portion consisting of CRDs 4-8 is able to endocytose mannose-terminated glycoproteins as efficiently as the intact receptor and has the same affinity as the intact receptor for three different ligands. In an i n uitro translation assay, CRD 4 is the only domain that has carbohydrate binding activity when expressed alone (18). This domain, produced in a bacterial expression system, exhibits multispecificity and mimics the monosaccharide binding activity of the whole receptor. However, CRD 4 binds only poorly to natural glycoproteins and thus cannot account for binding of the receptor to such ligands (18).
Although CRD 5 has no demonstrable saccharide binding activity when expressed alone, a segment of the receptor consisting of CRDs 5-7 does show binding activity in the i n vitro translation assay, and a receptor consisting of CRDs 5-8 mediates endocytosis of glycoproteins, albeit at a much slower rate than the intact receptor. Since a segment consisting of CRDs 6-8 is not active in similar assays, it has been concluded that CRDs 4, 5, and at least one additional CRD nearer the membrane anchor are necessary for binding and endocytosis of multivalent ligands (18). Clustering of CRDs, each with weak affinity for single sugars, to achieve high affinity binding to oligosaccharides is a feature of other Ctype lectins (16,19,20). However, in the mannose receptor, multiple weak interactions appear to be achieved through The abbreviations used are: CRDs, carbohydrate-recognition domains; BSA, bovine serum albumin.

399
Ligand Binding by Macrophage Mannose Receptor several active CRDs in a single polypeptide rather than by oligomerization of multiple polypeptides each with a single CRD. This paper describes experiments designed to determine how CRDs in the mannose receptor are clustered to achieve high affinity binding to glycoproteins. It is demonstrated that a core consisting of CRDs 4 and 5, combined with accessory CRDs 6 to 8 in a single polypeptide, mediate binding of a natural glycoprotein ligand.

EXPERIMENTAL PROCEDURES
Materials-Restriction enzymes, linkers, polymerase, and ligase were from New England Biolabs. Sepharose 6B, D-mannOSe, invertase grade VII, mannan, and molecular weight markers were obtained from Sigma. Man23-BSA was a product of E. Y. Laboratories. Amersham Corp. was the source for N a 9 and Amplify fluorography solution. Immulon 96-well microtiter plates were obtained from Dynatech. Subtilisin was from Boehringer Mannheim. Immobilon membranes were obtained from Millipore Corporation. 36S-Labeled methionine-cysteine mixture was purchased from Du Pont-New England Nuclear. Methionine-free Dulbecco's modified Eagle's medium was purchased from GIBCO/BRL. Sf9 cells and baculovirus transfer vectors were kindly provided by Dr. Max Summers, Texas A&M University.
Hydrodynamic Experiments-Rat-6 fibroblasts (four 60-mm plates at half-confluence) expressing full-length receptor (18) were labeled for 4 h in methionine-free medium supplemented with [35S]methionine-cysteine mix at a concentration of 125 pCi/ml(2 ml/plate). Cells were harvested in phosphate-buffered saline by scraping and collected by low speed centrifugation. The cell pellet (0.25 ml) was solubilized by sonication for 5 s in 2 ml of Triton loading buffer (0.5 M NaCl, 25 mM Tris-C1, pH 7.8, 25 mM CaC12, and 0.05% Triton X-100). After addition of more Triton X-100 to bring the final concentration to 0.5%, the sample was again sonicated for 5 s and centrifuged at 150,000 X g for 30 min in a Beckman TLA100.3 rotor. The supernatant was applied directly to a l-ml column of mannose-Sepharose (21). Following rinsing with 5 ml of Triton loading buffer, the column was eluted with buffer in which CaC12 was replaced with 2.5 mM EDTA. Receptor eluting in the second 1-ml fraction was concentrated to 100 p1 in an Amicon Centricon 30 microconcentrator for centrifugation experiments.
Velocity sedimentation, equilibrium density gradient centrifugation, and gel filtration experiments were performed as described previously (19). The receptor content of fractions was assessed by immunoprecipitation with antibody prepared against CRD 4 of the mannose receptor expressed in bacteria (18), followed by gel electrophoresis and fluorography. Marker proteins were localized by enzymatic assays. Calculated i, values were derived from tabulated results for individual amino acid (22) and sugar (23) residues. The carbohydrate content of the receptor was estimated by assuming that all of the 7 potential N-linked sites are occupied by biantennary, complex oligosaccharides and that there are four sites of tetrasaccharide attachment to threonine residues (8).

Expression of Portions of the Mannose Receptor in a Bmulovirus
Expression System-The procedures used for the baculovirus expression system have been described in detail by Summers and Smith (24). Standard recombinant DNA techniques were used in the construction of plasmids (25). The starting plasmid was a derivative of pSP64 containing the 5"untranslated segment of Xenopus luevis pglobin mRNA and the region of the dog preproinsulin gene encoding the signal sequence (26). DNA coding for portions of the mannose receptor were cloned into this plasmid after the region encoding the insulin signal sequence. Stop codons were created with XbaI linkers. The insulin signal sequence-receptor fusions were then cloned into the baculovirus transfer vector pVL1392.
Expression vectors were cotransfected with wild type Autograph californica nuclear polyhedrosis-virus DNA into Sf9 cells. Recombinant plaques were identified by hybridization of filter lifts using appropriate fragments of the receptor cDNA as probes, subjected to three rounds of plaque purification, and used to produce stocks of recombinant virus. For protein production, 108-109 cells were infected with virus (multiplicity of infection approximately 10) and then resuspended to a concentration of 5 X lo6 ceIls/ml in TNM-FH medium (24) containing 10% fetal calf serum, in spinner flasks. After incubation at 27 'C for 4 days, medium was harvested by centrifugation at 10,000 X g for 10 min.
Expressed proteins were isolated from the medium by affinity chromatography. Medium was mixed with an equal volume of 20 mM Tris-HC1, pH 7.8, 0.5 M NaCl, 20 mM CaClZ (loading buffer) and passed through a 2-ml column of mannose-Sepharose. The columns were washed with loading buffer and eluted with 20 mM Tris-HC1, pH 7.8, 0.5 M NaC1, 2 mM EDTA.
Proteolysis-Limited proteolysis with subtilisin was performed directly on material eluted from the affinity columns, after adjustment of the Ca2+ concentration, following the protocol previously described (27). For sequence analysis, samples were separated on minigels that were blotted onto Immobilon membranes (28). Bands cut from the membrane were subjected to sequence analysis in an Applied Biosystems 471A sequencer.
Solid-phose Binding Assay-This assay has been described previously (18). Plastic microtiter plates were coated overnight at 4 "C with lOOpl/well of each baculovirus-expressedportion of the mannose receptor (approximately 100 pg/ml in loading buffer). After coating, the wells were washed three times with 25 mM Tris-HC1, pH 7.8,1.25 M NaCl, 25 mM CaC12 (washing buffer) and blocked with 5% (w/v) BSA in the same buffer for 2 h at 4 "C. The wells were again washed three times and '261-Man23-BSA (approximately lo6 cpm in 100 pl of washing buffer containing 5% BSA specific activity approximately IO7 cpm/pg) with varying amounts of inhibitors was added to each well. After incubation for 2 h at 4 "C, the wells were washed rapidly three times with cold washing buffer and shaken dry. Individual wells were counted in a Packard PRIAS gamma-counter. Data were analyzed using a nonlinear least squares fitting program (Sigmaplot, Jandel Scientific).

RESULTS
Oligomeric State of the Mannose Receptor-Since oligomerization of the mannose receptor could potentially result in clustering of very large numbers of CRDs, it was important to determine the native size of the receptor. This analysis was made possible by the availability of rat fibroblasts that express the intact receptor (18). Cells were labeled with 35S-containing methionine and cysteine, following which receptor was solubilized in Triton X-100 and isolated by a single step of affinity chromatography on mannose-Sepharose. Freshly isolated receptor was subjected to velocity gradient centrifugation on Triton X-100-containing gradients. As shown in Fig. 1, the protein sediments as a single peak, with an s20,w value of 6.6 S. The absence of receptor from fractions sedimenting lower in the gradient (i.e. at higher s values) indicates that this is the largest form of the solubilized protein. When compared with antiserum raised against intact receptor (30), anti-CRD 4 serum precipitates similar amounts of receptor from labeled cells. Therefore, it is unlikely that an oligomeric form, in which CRD 4 is inaccessible to the antibodies used, is undetected in this experiment.
In order to derive an estimate of the molecular weight of the protein portion of the receptor-detergent complex, the partial specific volume of the complex and its diffusion coefficient were determined by equilibrium density gradient centrifugation and gel filtration chromatography, respectively. In each case, a single peak of receptor was detected by immunoprecipitation, gel electrophoresis, and fluorography. Substitution of the resulting values (17 = 0.844 cm3/g and D z~, w = 2.33 x cm2/s) along with the sedimentation coefficient into the Svedberg equation yielded a molecular weight of 410,000 for the complex. The measured partial specific volume was combined with the partial specific volume of the glycosylated receptor polypeptide (0.718 cm3/g) and Triton X-100 (0.94 cm3/g) to estimate that the fraction of receptor in the complex is 43%. The result indicates that the non-detergent portion of the complex has a molecular weight of 180,000.
Since the calculated molecular weight of the glycosylated receptor polypeptide is estimated to be 183,000, the sedimentation results show that the receptor most likely is monomeric in detergent solution. Consistent with this conclusion, no  FRACTION  1 0 1 1 1 2 1 3 1 4 1 5 1 6 1 7 1 8 1 9  20 . . .

FRACTION NUMBER
evidence of oligomers was obtained when the solubilized protein was treated with the homobifunctional reagent bis(sulfosuccinimidy1)suberate (data not shown) under conditions that result in efficient cross-linking of the mannosebinding proteins (27). Similarly, treatment of cells with 1,5difluoro-2,4-dinitrobenzene produced no evidence of oligomer formation, although oligomers of the chicken hepatic lectin could be readily detected under the same conditions (19). These results suggest that the receptor is probably a monomer in the membrane as well as in detergent solution.
Expression of Internal Fragments of the Receptor-Since the mannose receptor polypeptide appears to be a monomer, the clustering of CRDs necessary to achieve high affinity binding of complex ligands must result solely from the activity of multiple CRDs within a single polypeptide. In previous studies designed to probe the role of various CRDs in the mannose receptor, truncated polypeptides lacking one or more NH2-terminal CRDs were expressed in rat fibroblasts and analyzed for ligand binding. Because it was necessary to retain the COOH-terminal membrane anchor in these constructs, it was not possible to analyze COOH-terminal truncations (18).
In this study, the baculovirus system has been used to produce internal fragments of the receptor. These fragments are water soluble and can be purified and analyzed more readily than those retained in the membrane. Portions of the receptor cDNA coding for CRDs 4-5,4-6, and 4-7 were fused t o codons specifying the dog insulin signal sequence (Fig. 2). The insulin signal sequence directs sequestration of the receptor fragments in the endoplasmic reticulum (18), leading to their eventual secretion. The fragments were purified from the medium by one step of affinity chromatography on mannose-Sepharose. In each case, gel electrophoresis revealed the presence of a single polypeptide band of the expected molecular weight (Fig. 3). The purified receptor fragments were used in proteolysis experiments and in assays to determine affinities for various ligands.
Domain Organization-A common property of C-type CRDs is resistance to proteolysis in the presence of Ca2+ (19, 27), reflecting their tightly folded structures. Therefore, it was of interest to determine whether the CRDs of the mannose receptor are similarly resistant to proteolysis. The receptor fragment consisting of CRDs 4-7 was subjected to limited proteolysis with subtilisin in the presence of Ca2+. As shown in Fig. 4, the amount of the CRD 4-7 fragment decreases, and three subfragments appear as the concentration of subtilisin is increased. The molecular weights of the protease-resistant subfragments, combined with NH2-terminal sequence analysis, indicate that they correspond to CRDs 4-6, 4-5, and 6. Similar digestion of the expressed fragment containing CRDs 4-6 also yielded subfragments corresponding to CRDs 4-5 and 6 (data not shown).
The proteolysis experiments provide information about the   1 0.2 0.4 1 .O 2.0 4.0  structures of the individual CRDs and their arrangement in clusters. CRD 7 is the most susceptible to proteolysis and is removed from the 4-7 fragment at the lowest concentration of subtilisin. No released CRD 7 was detected, indicating that it is probably degraded by subtilisin. Although the sequence of CRD 7 indicates that it is evolutionarily divergent from the other CRDs (31), its apparent susceptibility to proteolysis in the presence of Ca2+ is not correlated in a simple way with the absence of residues predicted from the structure of the mannose-binding protein to be involved in Ca2+ ligation (12).
CRD 6 is released intact from the CRD 4-6 fragment a t higher concentrations of subtilisin. Significant amounts of CRDs 4-5 and CRD 6 remain even after digestion for 60 min with 4 pg/ml subtilisin. These results indicate that CRDs 4, 5, and 6 each have compactly folded conformations that probably resemble the CRD of mannose-binding protein. The data also show that CRDs 4 and 5 form a tightly linked unit, suggesting that these CRDs form extensive contacts and are not simply linked by a flexible tether. Such an arrangement would be expected to fix the relative orientation of the saccharide-binding sites in these domains. Since CRD 4 has been shown to display the highest affinity for sugar of any of the CRDs (18), the CRD 4-5 pair probably forms a core for binding of multivalent ligands. This hypothesis is supported by results described below. Binding Affinities-A solid-phase binding assay previously used to study the binding of CRD 4 to saccharide ligands was used to measure binding to the larger fragments containing additional CRDs. In this assay, '2sII"an23-BSA binds to portions of the receptor immobilized on microtiter plates. Affinities for Man23-BSA, invertase, and mannan were determined based on their ability to compete with the labeled ligand. Typical binding curves for mannan are shown in Fig. 5, and the results for all three ligands are summarized in Table I. No differences in affinities for Man23-BSA or invertase are seen between the three fragments consisting of CRDs 4-5,4-6 , and 4-7. However, the affinity for mannan increases as additional CRDs are added.
The results obtained previously for CRD 4 alone and for CRDs 4-8 expressed in fibroblasts (18) are compared to the present results in Fig. 6, which shows a plot of affinities for the three ligands tested uersus the number of CRDs present. CRDs 4-8 have been shown to have the same affinities as the

TABLE I Inhibition of binding to mannose receptor CRDs
Competition binding assays were performed as described in Fig. 5 intact receptor for these three ligands (18). The greatest increase in affinity for all three ligands is seen when CRD 5 is combined with CRD 4. The increase is particularly marked for the two natural glycoproteins. For ManZ3-BSA and invertase, CRDs 4 and 5 are sufficient to reproduce the affinity of the intact receptor. The graph emphasizes that the affinity for mannan continues to increase as more CRDs are added.
The full affinity of the intact receptor for this ligand is only obtained when CRDs 4-8 are present.
In related studies, fragments were labeled and tested for binding to various neoglycoproteins and natural glycoproteins resolved by gel electrophoresis and immobilized on nitrocel-  (18) and other receptor fragments immobilized on polystyrene are compared. In all cases, KI was defined as described in Fig. 5. lulose.' As expected from the results just described, CRD-4 alone as well as CRDs 4-5,4-6, and 4-7 all bind Man23-BSA. However, CRD 4 alone does not bind to invertase or ribonuclease B, while the three larger fragments do. Thus, as in the assay using immobilized CRDs, the binding of larger receptor fragments to natural glycoproteins such as invertase and ribonuclease B reflects an increase in affinity due to the presence of multiple CRDs. Because it does not migrate as a discrete band on SDS gels, mannan could not be tested in this assay. Since these experiments were conducted with receptor fragments in solution, the fact that the results are completely consistent with those obtained with immobilized CRDs indicates that the densities of receptor fragments on the polystyrene surface used in the solid-phase assay do not affect the way multivalent ligands are bound.

DISCUSSION
It has been shown previously that, while CRD 4 alone can mimic the binding of monosaccharides to the intact mannose receptor, additional CRDs are needed for binding to complex ligands (18). In the present study, hydrodynamic experiments and expression of internal fragments of the receptor in insect cells have been used to define further which CRDs are required for binding ligand with high affinity.
The results confirm that the mannose receptor binds oligosaccharides through several CRDs in a single polypeptide. No evidence for oligomer formation was seen. The number of CRDs needed for high affinity binding is dependent on the ligand. CRDs 4 and 5 form the basic unit required for binding multivalent ligands with submicromolar affinities. While CRD 4 alone has weak affinity for natural glycoproteins such as invertase (KD = 1 PM), CRDs 4 and 5 together bind to this ligand almost as tightly as the whole receptor (KD = 20 nM).
Since these two domains cannot be separated by proteolysis, it is concluded that they form a tightly linked structural unit. ~~ M. E. Taylor, K. Bezouska, and K. Drickamer, unpublished observations. For other natural ligands, such as yeast mannan, CRDs 4-5 have a much higher affinity than does CRD 4 alone, but additional increases in affinity are seen as CRDs 6, 7, and 8 are added.
Clustering of CRDs has been shown to be important in several other C-type lectins. These other lectins contain only a single CRD in each polypeptide and thus must form homoor hetero-oligomers to achieve clustering of CRDs. The chicken hepatic lectin is a trimer formed through the association of transmembrane regions of identical polypeptide^.^ This oligomerization results in the intact receptor binding a multivalent glycoprotein ligand with 25-fold greater affinity than does the monomeric CRD (19). Two different polypeptides of the rat asialoglycoprotein receptor, each with a single CRD, form hetero-oligomers (20). The different subunits bind to different terminal galactose residues on a triantennary glycopeptide (32), indicating that clustering of CRDs determines the specificity of the interaction, as well as the affinity. It is probable that different forms of clustering of CRDs seen in various C-type lectins reflects differences in their modes of binding to multivalent ligands. The CRDs of each protein must be arranged spatially to match the geometric configurations of their particular oligosaccharide ligands. The structure of the hepatic lectins may provide optimal binding to clusters of terminal residues, while the linear arrangement of CRDs in the mannose receptor may be more suitable for high affinity binding to repeated polymers such as yeast mannan.
The observation that the core CRD 4-5 unit binds certain saccharides, such as those found in invertase and ribonuclease B, with high affinity, suggests that the binding sites in these CRDs are suitably disposed to interact with a structural motif common to the oligosaccharide portions of these glycoproteins. The role of accessory CRDs such as 6, 7, and 8, may be to provide alternative geometrical arrangements of binding sites that can interact with differently disposed sugars, such as those in mannan. Some C-type anima1 lectins, such as the asialoglycoprotein receptor, bind endogenous glycoproteins with a limited set of structures. In contrast, the mannose receptor must recognize a diversity of foreign oligosaccharides. The presence of multiple CRDs in addition to the CRD 4-5 core may provide the geometrical flexibility needed to interact with this diversity of ligands. Analysis of the crystal structure of the CRD of rat mannosebinding protein bound to a high mannose oligosaccharide (33) has highlighted why clustering of CRDs might be necessary for achieving high affinity binding and specificity. In the crystal, each CRD interacts with a single terminal sugar residue. Two hydroxyl residues of the sugar are ligated in a shallow indentation on the surface of the protein. The limited contact between the sugar and the CRD explains how the CRD can have specificity for several different monosaccharides, and why the interactions with monosaccharides are weak. The fact that two CRDs in the crystals simultaneously bind different terminal mannose residues in a single oligosaccharide indicates why more than one CRD may be required for recognition of branched oligosaccharides.
Each of the CRDs of the mannose receptor contains most of the conserved residues necessary for establishing the overall folded structure of the mannose-binding protein CRD. In CRDs 4 and 5, all 5 residues involved in ligation of the 2 sugar hydroxyl residues to the mannose-binding protein CRD are also conserved (33, 34). Thus, it is likely that the interaction of these two domains with monosaccharides is similar to that of the mannose-binding protein CRD. CRDs 6, 7, and 8 do F. Verrey and K. Drickamer, manuscript submitted. not contain all of the residues involved in ligation of sugar to mannose-binding protein and thus may bind sugars in a different manner. Further structural work will be required to