Mechanism of Ca2+ and monosaccharide binding to a C-type carbohydrate-recognition domain of the macrophage mannose receptor.

Site-directed mutagenesis has been used to identify residues that ligate Ca2+ and sugar to the fourth C-type carbohydrate-recognition domain (CRD) of the macrophage mannose receptor. CRD-4 is the only one of the eight CRDs of the mannose receptor to exhibit detectable monosaccharide binding when expressed in isolation, and it is central to ligand binding by the receptor. CRD-4 requires two Ca2+ for sugar binding, like the CRD of rat serum mannose-binding protein (MBP-A). Sequence comparisons between the two CRDs suggest that the binding site for one Ca2+, which ligates directly to the bound sugar in MBP-A, is conserved in CRD-4 but that the auxiliary Ca2+ binding site is not. Mutation of the four residues at positions in CRD-4 equivalent to the auxiliary Ca2+ binding site in MBP-A indicates that only one, Asn728, is involved in ligation of Ca2+. Alanine-scanning mutagenesis was used to identify two other asparagine residues and one glutamic acid residue that are probably involved in ligation of the auxiliary Ca2+ to CRD-4. Sequence comparisons with other C-type CRDs suggest that the proposed binding site for the auxiliary Ca2+ in CRD-4 of the mannose receptor is unique. Evidence that the conserved Ca2+ in CRD-4 bridges between the protein and bound sugar in a manner analogous to MBP-A was obtained by mutation of one of the amino acid side chains at this site. Ring current shifts seen in the 1H NMR spectra of methyl glycosides of mannose, GlcNAc, and fucose in the presence of CRD-4 and site-directed mutagenesis indicate that a stacking interaction with Tyr729 is also involved in binding of sugars to CRD-4. This interaction contributes about 25% of the total free energy of binding to mannose. C-5 and C-6 of mannose interact with Tyr729, whereas C-2 of GlcNAc is closest to this residue, indicating that these two sugars bind to CRD-4 in opposite orientations. Sequence comparisons with other mannose/GlcNAc-specific C-type CRDs suggest that use of a stacking interaction in the binding of these sugars is probably unique to CRD-4 of the mannose receptor.

Many proteins of both plants and animals are involved in recognition of complex carbohydrates attached to proteins or lipids (1,2). One such protein, the macrophage mannose receptor, binds terminal mannose, fucose, or N-acetylglucosamine residues of glycoconjugates in a Ca 2ϩ -dependent manner. The receptor acts as a molecular scavenger by clearing endogenous glycoproteins bearing high mannose oligosaccharides as well as pathogenic microorganisms (1). The extracellular region of the mannose receptor contains an N-terminal cysteine-rich domain and a fibronectin type II repeat as well as eight C-type carbohydrate-recognition domains (CRDs), 1 which make it a member of the C-type lectin family (3). The C-type lectin family is a large and diverse group of proteins characterized by homologous CRDs that generally mediate Ca 2ϩ -dependent sugar recognition (4). Well characterized members of the C-type lectin family include the mammalian mannose-binding proteins (MBPs), the hepatic asialoglycoprotein receptor, and the selectins (1,2).
Details of how different animal and plant proteins recognize carbohydrates are becoming clearer, due to the availability of crystal structures of the proteins in complex with sugar ligands (2). Crystal structures of two C-type CRDs, those of rat serum MBP (MBP-A), and rat liver MBP (MBP-C) have been solved in complex with sugar ligands (5,6). The C-type CRD of E-selectin has also been crystallized, but without bound sugar (7). Examination of these crystal structures, combined with other physical techniques and mutagenesis have established some of the molecular mechanisms involved in sugar recognition by C-type CRDs (2). However, little is known about the molecular mechanisms of carbohydrate recognition by the C-type CRDs of the mannose receptor.
Analysis of sugar recognition by the mannose receptor is complicated by the presence of eight different C-type CRDs in a single polypeptide. Apart from the mannose receptor, two other proteins, a phospholipase A 2 receptor of muscle and an endocytic receptor called DEC-205, located on dendritic cells, have multiple CRDs in a single polypeptide (8 -10). These two proteins are very divergent from other groups of C-type lectins and probably do not bind carbohydrates. The CRDs of the mannose receptor are also quite divergent from the prototypes of the C-type lectin family, but most of them contain at least some of the residues shown to be important for sugar recognition. Of the eight CRDs, only CRD-4 has been shown to bind monosaccharides when expressed in isolation, although other CRDs must contribute to binding of oligosaccharides, since CRDs 4 -8 must be present to achieve the affinity of the intact receptor for natural ligands (11,12).
Since CRD-4 is the smallest piece of the receptor that retains the ability to interact with sugars, an understanding of the molecular mechanism of sugar binding by this CRD would be a first step toward understanding how the whole receptor recog-MBP-C, two Ca 2ϩ are ligated at the same sites as in MBP-A (6). In contrast, the crystal structure of the CRD of E-selectin reveals bound Ca 2ϩ only at the site corresponding to the conserved Ca 2ϩ in MBP-A, and binding studies in solution confirm that there is a single Ca 2ϩ binding site in this CRD (7,14). Like CRD-4, the selectin CRD lacks the residues involved in Ca 2ϩ coordination at the auxiliary site in MBP-A. Thus, the mode of binding of the second Ca 2ϩ to CRD-4 must be different from that of many other C-type CRDs.
Although it is probable that the major interaction between sugars and C-type CRDs, including CRD-4 of the mannose receptor, is via direct ligation to a conserved Ca 2ϩ , there are likely to be other contacts between the protein and sugar. In MBP-A, interaction between bound sugar and the ␤-carbon of a histidine residue, His 189 , contributes significantly to the overall binding energy (15). It is probable that additional contacts between CRD-4 and bound sugar will be different from those seen in MBP-A. In MBP-A, binding of the auxiliary Ca 2ϩ is responsible for correct positioning of loops around the other Ca 2ϩ to which the sugar ligates (5). If, as suggested by the sequence comparisons, the mode of ligation of the auxiliary Ca 2ϩ in CRD-4 is different from that seen in MBP-A, the position of the loops near the sugar binding site is likely to be different.
In order to clarify how CRD-4 of the mannose receptor binds two Ca 2ϩ in the absence of amino acid residues corresponding to one of the sites in MBP-A, site-directed mutagenesis experiments have been undertaken to identify residues involved in ligating each of the two Ca 2ϩ to CRD-4. NMR studies designed to probe the nature of the molecular interactions between CRD-4 and monosaccharides have also been undertaken. The results suggest that one Ca 2ϩ is ligated to CRD-4 in the same way as in MBP-A but that CRD-4 has a unique binding site for the auxiliary Ca 2ϩ . In addition to direct ligation of sugar at the conserved Ca 2ϩ site, a hydrophobic stacking interaction between the bound sugar and a tyrosine residue contributes about 25% of the overall binding energy.

EXPERIMENTAL PROCEDURES
Materials-Bovine serum albumin, N-tosylphenylethyl chloromethyl ketone-treated trypsin, monosaccharides, D 2 O, and deuterated buffer components were from Sigma. Deuterated imidazole was from Cambridge Isotope Laboratories. Restriction enzymes and other DNA modification enzymes were from New England Biolabs. The Sequenase II kit for DNA sequencing and Na 125 I were obtained from Amersham Corp. Mannose 31 -BSA was purchased from EY Laboratories. Volumetric CaCl 2 solution was from BDH, and Immulon 96-well microtiter plates were from Dynatech. Mannose-Sepharose was prepared using the divinylsulfone method (16).
Mutagenesis Procedures-Mutagenesis was performed on a fragment corresponding to bases 2000 -2467 of a cDNA for the human macrophage mannose receptor (3) inserted into the vector pUC19 grown in Escherichia coli strain HB101. Synthetic oligonucleotides were inserted at appropriate restriction sites using standard recombinant DNA procedures. Modified XhoI-BamHI fragments were transferred into a CRD-4 expression plasmid derived from pINIIIompA3 (11) and transformed into E. coli strain JA221. Plasmids were sequenced to verify the mutations made.
Protein Expression and Purification-Mutant proteins were prepared using the procedure developed for wild type CRD-4 (13). Growth, induction, and harvesting of transformed bacteria were exactly as described for wild type CRD-4 (13). The procedure for purification of expressed mutant domains from the bacterial cell lysates was modified slightly to allow for any decrease in sugar affinity due to the mutations. Pelleted bacteria from 1 liter of culture were lysed by sonication in 50 ml of loading buffer (25 mM Tris-HCl, pH 7.8, 1.25 M NaCl, 3 mM CaCl 2 ). The lysate was spun at 45,000 rpm for 1 h in a Beckman 55.2Ti rotor. The supernatant (50 ml) was passed over a 10-ml column of mannose-Sepharose equilibrated with loading buffer. The column was washed with 10 ml of wash buffer (25 mM Tris-HCl, pH 7.8, 1.25 M NaCl, 20 mM CaCl 2 ), followed by a further 15 fractions (2 ml) of wash buffer. The column was eluted with 10 aliquots (2 ml) of elution buffer (25 mM Tris-HCl (pH 7.8), 1.25 M NaCl, 2.5 mM EDTA). Both wash fractions and elution fractions were analyzed by SDS-polyacrylamide gel electrophoresis, and fractions containing pure domains were pooled for analysis. Under these conditions, wild type CRD-4 binds to the column and only elutes upon the addition of EDTA-containing buffer. Some mutant domains eluted from the column during the wash before the addition of EDTA. Mutant CRDs were classified into four groups with phenotypes characterized by their behavior on the mannose-Sepharose column: (a) wild type or tight binding proteins eluted only after 10 -12 ml of elution buffer; (b) intermediate binding proteins started to elute after 24 -30 ml of wash buffer; (c) weak binding proteins were seen in all wash fractions; and (d) nonbinding proteins were only detected in the flowthrough and the first 10 ml of wash buffer.
Protein for NMR experiments was further purified by reverse phase high performance liquid chromatography on a C3 column, as described previously (13).
Proteolytic Digestion-Bacterial lysates containing expressed proteins were prepared as described above. Lysates were dialyzed extensively against water and freeze-dried. Dried lysates were resuspended to a concentration of 65 mg/ml in 25 mM Tris-HCl (pH 7.8), 1.25 M NaCl, and either 1 or 25 mM CaCl 2 . Trypsin was added to give a concentration of 1 mg/ml, and reactions were incubated at 37°C. Reactions were stopped at various time intervals by the addition of an equal volume of 2 ϫ SDS gel sample buffer containing 20 mM EDTA and 2% 2-mercaptoethanol followed by boiling at 100°C for 5 min. The products of the reaction were analyzed by SDS-gel electrophoresis followed by immunoblotting. Protease-resistant CRD-4 was detected by autoradiography after incubation of the blot with a rabbit polyclonal anti-mannose receptor antibody, raised against a fragment of the receptor consisting of CRDs 4 -7 produced in insect cells (12), followed by 125 I-labeled protein A. The resistant material was quantified using the Macintosh Image 1.49 application.
Ca 2ϩ Binding Assay-Ca 2ϩ -dependence of mutant CRD-4 binding to 125 I-Man-BSA was determined using a solid phase binding assay described previously for wild type CRD-4 (13). Serial 1:1.5 dilutions of CaCl 2 in 2 ϫ blocking buffer were prepared starting at 150 mM or 50 mM. To these, 125 I-Man-BSA was added to give a final ligand concentration of 0.75 g/ml. Aliquots (75 l) of these mixtures were then added to duplicate wells of CRD-4-coated plates. At least two assays were performed for each mutant protein. The resulting data were fitted to equations for both first order (Equation 1) and second order (Equation 2) processes using a nonlinear least squares fitting program (Sigmaplot, Jandel Scientific).
Competition Binding Assays-Competition assays using 125 I-Man 31 -BSA as test ligand were performed as described previously (11). Values for K i , corresponding to the concentration of monosaccharide that gives 50% inhibition of test ligand binding, were calculated using the nonlinear, least squares fitting program. Mean Ϯ S.D. values of three independent assays performed in duplicate were used to calculate K i values relative to the K i for mannose.
NMR Binding Assays-NMR binding assays were performed basically as described for the CRD of MBP-A (15), except that the protein was not exchanged into D 2 O. Protein was dissolved in D 2 O and adjusted to 25 mM deuterated imidazole-HCl, pH 7.2, 0.5 M NaCl, and 5 mM CaCl 2 , using concentrated stock solutions. The final volume was 0.55 ml at a protein concentration of approximately 5-10 mg/ml. Titrations were performed by adding concentrated stock solutions of sugars (0.137 M and 1.1 M) in D 2 O. One-dimensional 1 H NMR spectra were recorded on a Varian Unity 500 MHz spectrometer at a temperature of 30°C. Chemical shifts were measured relative to an internal HDO standard, referenced 4.78 ppm from sodium 4,4-dimethyl-4-silapentane-1-sulfonate. Peak shifts that occurred upon the addition of sugar were analyzed using the following equation.
⌬␦ is the shift in the peak position at a given glycoside concentration relative to the peak position in the protein spectrum in the absence of sugar, ⌬␦ B is the difference in peak positions between the glycosideprotein complex and the free protein, and K D is the concentration of glycoside that gives half-maximal change in chemical shift. ⌬␦ B and K D were calculated using the nonlinear least squares fitting program.

Localization of Ca 2ϩ Binding Sites in CRD-4
Interaction of CRD-4 with Ca 2ϩ and Mannose at the Conserved Site-To confirm the proposed role of the conserved Ca 2ϩ binding site in CRD-4 ( Fig. 1), site-directed mutagenesis was used to assess the role of Glu 733 in sugar and Ca 2ϩ binding. If the pattern of ligation in CRD-4 is the same as in MBP-A, the side chain of Glu 733 should form a coordinate bond to Ca 2ϩ and a hydrogen bond to hydroxyl 3 or 4 of mannose. Mutation of Glu 733 to Gln should produce a domain that does not bind sugar, since the hydrogen bond donor for the sugar hydroxyl has been removed. However, the carbonyl oxygen of Gln can form a coordinate bond with Ca 2ϩ , so Ca 2ϩ binding should be unaffected. A similar phenotype has been seen in MBP-A when Asn 187 is mutated to Asp or Glu 185 is changed to Gln (5,15). However, mutation of Glu 733 to Ala would perturb Ca 2ϩ binding as well as sugar binding.
Mutation of Glu 733 in CRD-4 to either Gln or Ala resulted in domains that do not bind to mannose-Sepharose, indicating that sugar binding activity is lost. The Ca 2ϩ binding properties of these domains were investigated using proteolysis (Fig. 2). Since the lack of sugar binding activity made it difficult to isolate the two mutant domains, protease resistance experiments were carried out on bacterial lysates containing the expressed domains. CRD-4, like MBP-A, is resistant to proteolysis in the presence of Ca 2ϩ , and two Ca 2ϩ must bind to the CRD to render it fully resistant to protease (13). Thus, a mu-tation affecting Ca 2ϩ binding would be expected to produce a domain that is less resistant to proteolysis.
In the presence of 25 mM Ca 2ϩ , both wild type CRD-4 and the Glu 733 3 Gln mutant are resistant to proteolysis by trypsin ( Fig. 2A). However, the Glu 733 3 Ala mutant shows susceptibility to proteolysis, with approximately 50% of the protein being digested over the time course of the experiment. In the presence of 1 mM Ca 2ϩ all of the domains show susceptibility to proteolysis, but the degree of susceptibility differs (Fig. 2B). For wild type CRD-4 and the Glu 733 3 Gln mutant, approximately 60% of the protein is still intact after 80 min, but for the Glu 733 3 Ala mutant the amount of CRD-4 remaining drops to below 50% in only 20 min. Since the Glu 733 3 Gln mutant shows the same degree of resistance to proteolysis as does wild type CRD-4, at both Ca 2ϩ concentrations, this domain must still bind two Ca 2ϩ . Loss of resistance to protease in the presence of Ca 2ϩ by the Glu 733 3 Ala mutant indicates that this domain does not bind two Ca 2ϩ .
The phenotypes produced by the two mutations at position 733 suggest that the pattern of ligation at the conserved Ca 2ϩ site in CRD-4 is like that of MBP-A rather than E-selectin, in which the equivalent glutamate (Glu 88 ) is replaced as a Ca 2ϩ ligand by a water molecule ligated to an adjacent asparagine side chain (7). Glu 733 in CRD-4 must act as a ligand for both Ca 2ϩ and a hydroxyl group of mannose.
Interaction with Ca 2ϩ at the Auxiliary Site-Residues ligating Ca 2ϩ 1 in the crystal structure of MBP-A and residues at the equivalent positions in a molecular model of CRD-4 are shown in Fig. 1. Only the glutamate residue of MBP-A is absolutely conserved in CRD-4. The two tyrosine residues at this site in CRD-4 would not be predicted to be involved in Ca 2ϩ binding, because OH groups of tyrosines have not been seen to ligate Ca 2ϩ in any other Ca 2ϩ -binding proteins (17,18). Substitution of Asp 188 of MBP-A by asparagine in CRD-4 is conservative, because the carbonyl oxygen atom of asparagine can ligate Ca 2ϩ . The auxiliary Ca 2ϩ binding site in MBP-C is identical to the site in MBP-A, except that in this case Asp 188 is also replaced by asparagine (6). Thus, at the site equivalent to the Ca 2ϩ 1 binding site in MBP-A, the only residues of CRD-4 with side chains that could ligate Ca 2ϩ are Asn 728 and Glu 706 .
To identify amino acid residues involved in ligation of Ca 2ϩ at the auxiliary site in CRD-4, potential Ca 2ϩ ligands were mutated either to alanine or phenylalanine. An initial assessment of whether a mutation alters the Ca 2ϩ binding properties of the domain was made by observing the behavior of the protein on the mannose-Sepharose column used to purify the domains, as described under "Experimental Procedures." Mutation of Glu 706 to Ala or mutation of either Tyr 701 or Tyr 734 to Phe had no effect on the ability of CRD-4 to bind to mannose-Sepharose; each of these three mutant domains was eluted from the column only in the presence of EDTA. If, as in MBP-A, the role of Ca 2ϩ at the auxiliary site of CRD-4 is in orientation of the loops forming the sugar binding site rather than in direct ligation of sugar, mutations leading to loss of Ca 2ϩ at this site should not abolish sugar binding altogether. Loss of Ca 2ϩ from the auxiliary site would, however, be expected to decrease the affinity of the CRD for Ca 2ϩ at the conserved site, so that sugar binding would be affected at low Ca 2ϩ concentrations. Since these three mutants do not show reduced binding to mannose-Sepharose under the conditions used, it seems unlikely that these residues play a role in Ca 2ϩ binding.
In order to confirm this conclusion, the Ca 2ϩ binding properties of the domains were further investigated using a solid phase binding assay that measured Ca 2ϩ -dependent binding of CRD-4 to 125 I-Man-BSA. By fitting the data to equations for either first or second order dependence on Ca 2ϩ , it was possible to determine whether the mutant CRDs bound one or two Ca 2ϩ and to obtain values for the apparent K Ca . A mutation that eliminated the auxiliary Ca 2ϩ binding site would be predicted to change the dependence on Ca 2ϩ from second order to first order and to increase the apparent K Ca . Ca 2ϩ dependence of binding to 125 I-Man-BSA is shown in Fig. 3. Like wild type CRD-4 ( Fig. 3A), each of these domains show second order dependence on Ca 2ϩ for binding to 125 I-Man-BSA, with values for K Ca almost identical to wild type (Table I ligation of Ca 2ϩ to CRD-4. The change Asn 728 3 Ala did affect the ligand binding properties of CRD-4. A domain with this mutation shows weak binding to mannose-Sepharose, with protein appearing in Ca 2ϩ -containing wash fractions before the addition of EDTA. However, the domain retains sufficient sugar-binding activity to allow purification and assaying of binding to 125 I-Man-BSA. This domain shows first order dependence on Ca 2ϩ for binding to 125 I-Man-BSA ( Fig. 4B) with an apparent K Ca approximately 10-fold higher than that of wild type (Table I). Since K Ca is measured indirectly, through formation of a ternary complex of Ca 2ϩ , protein, and sugar ligand, the concentration of the sugar ligand or the affinity of the protein for sugar could affect the apparent K Ca (13). However, the concentration of neoglycoprotein reporter ligand is constant, and a change in a direct interaction between the sugar and protein would not be expected to change the second order dependence on Ca 2ϩ and can thus be distinguished from a change in a Ca 2ϩ ligand. Therefore, the results in Fig. 4B, which indicate that the Asn 728 3 Ala mutant binds only one Ca 2ϩ , suggest that Asn 728 directly ligates the auxiliary Ca 2ϩ in CRD-4. Thus, only one of the four residues of CRD-4 at positions equivalent to the Ca 2ϩ 1 ligands in MBP-A is involved in ligation of Ca 2ϩ to CRD-4.
Identification of Other Potential Ca 2ϩ Ligands-Further mutagenesis was carried out to identify other amino acid residues involved in formation of the auxiliary Ca 2ϩ binding site in CRD-4. Since examination of crystal structures of Ca 2ϩ -binding proteins shows that over 98% of oxygen atoms ligated to Ca 2ϩ come from carboxyl and carbonyl groups or water molecules (17,18), only amino acids with side chain carboxyl or amide groups were investigated as potential Ca 2ϩ ligands. The position of residues with acid or amide side chains was also taken into account. Since CRD-4 shows second order dependence on Ca 2ϩ for sugar binding, binding of one Ca 2ϩ must be influenced by binding of the other. Thus, it is reasonable to suppose that the two Ca 2ϩ binding sites in CRD-4 will be close together. For instance, the two Ca 2ϩ bound to the CRD of MBP-A are 8.45 Å apart (5). Furthermore, identification of Asn 728 as a ligand for the auxiliary Ca 2ϩ means that other ligands must be close to Asn 728 . The molecular model of CRD-4 predicts Asn 728 to be 9 Å from the conserved Ca 2ϩ .
In the molecular model of CRD-4, seven residues with acid or amide side chains are found within 15 Å of the conserved Ca 2ϩ . These seven residues, together with another two located 18 Å from the conserved Ca 2ϩ , were selected for mutagenesis. Apart from Asp 741 , all of these residues are also within 18 Å of Asn 728 . Residues as far from the conserved Ca 2ϩ as 18 Å were considered to make allowances for possible differences between the structure predicted by the model and the actual structure. Although the ␤-strands and ␣-helices predicted by the molecular modeling are likely to be present in the real structure, the positions of loops cannot be predicted with the same level of confidence.
Each of the nine residues described above was mutated to alanine. The results obtained from the solid phase binding assay for the mutated domains are summarized in Table I. Of the nine mutations, only one, Glu 737 3 Ala, gives rise to a phenotype similar to that of the Asn 728 mutant. Like the Asn 728 mutant, the Glu 737 mutant shows weak binding to mannose-Sepharose. This domain also shows first order dependence on Ca 2ϩ for binding to 125 I-Man-BSA (Fig. 4C), indicating that it binds only one Ca 2ϩ . The apparent K Ca is increased approximately 10-fold compared with that of wild type CRD-4 (Table I). These results suggest that like Asn 728 , Glu 737 is involved in ligation of the auxiliary Ca 2ϩ . The data for the two mutant domains are shown fitted to the equation for first order dependence on Ca 2ϩ concentration. For the Asn 728 mutant, the sum of squared residuals for the first order fit is 0.04 compared with 0.24 for the second order fit; for the Glu 737 mutant, the sum of squared residuals for the first order fit is 0.12 compared with 0.28 for the second order fit.
Mutation of six residues (Glu 719 , Asn 720 , Gln 730 , Asp 741 , Glu 752 , and Asn 755 ) has no effect on binding of Ca 2ϩ or sugar. Domains with these residues mutated bind tightly to mannose-Sepharose and show second order dependence on Ca 2ϩ for binding to 125 I-Man-BSA, with values for apparent K Ca very similar to wild type CRD-4 ( Table I). The results indicate that side chains of these six residues are not acting as ligands for Ca 2ϩ .
Mutation of the remaining two potential ligands for the auxiliary Ca 2ϩ , Asn 731 and Asn 750 , gave rise to domains that still show second order dependence on Ca 2ϩ for binding to 125 I-Man-BSA (Fig. 5, A and B). However, the apparent K Ca for each of these mutant CRDs is increased 2-3-fold when compared with wild type CRD-4 (Table I). In addition, these CRDs showed intermediate binding to mannose-Sepharose, with protein starting to appear in the last few wash fractions. One explanation for the phenotypes of these mutants is that these residues are involved in ligation of Ca 2ϩ but that their contribution to the overall binding energy is small, so that removal of one of these side chains is not enough to cause loss of Ca 2ϩ from the auxiliary site but does decrease the affinity of the domain for Ca 2ϩ .
The hypothesis that Asn 731 and Asn 750 contribute to binding the auxiliary Ca 2ϩ was investigated further by making a double mutant in which both of these residues were changed to alanine. Ca 2ϩ dependence of binding to 125 I-Man-BSA by this domain is shown in Fig. 5C. The double mutation causes a phenotype very similar to that of the Asn 728 or Glu 737 single mutations. Thus, the double mutant shows first order dependence on Ca 2ϩ for sugar binding, with a greatly increased apparent K Ca (Table I) and shows weak binding to mannose-Sepharose. The binding curve for a double mutant containing Asn 731 and Asn 755 both changed to alanine, prepared as a control, is also shown in Fig. 5D. This double mutant still binds to the mannose-Sepharose column in the same way as the domain with the Asn 731 3 Ala single mutation and displays second order dependence on Ca 2ϩ for binding to 125 I-Man-BSA, although there is a small rise in apparent K Ca when compared with to the Asn 731 3 Ala mutant (Table I). These results support the hypothesis that side chains of Asn 731 and Asn 750 are involved in ligation of the auxiliary Ca 2ϩ but that they contribute less to the overall binding energy than the side chains of Asn 728 and Glu 737 .

Monosaccharide-Protein Interactions
Determination of Sugar Dissociation Constants by NMR-An NMR-based assay has been used to study the interaction of CRD-4 with monosaccharides. This type of assay allows the determination of direct binding constants for relatively weak interactions and has been used previously to study sugar binding by the CRD of MBP-A (15) as well as by various plant lectins (19) and influenza hemagglutinin (20). Inhibition assays, which measured inhibition of binding to CRD-4 by monosaccharides in a solid phase assay, indicate that the affinity of CRD-4 for monosaccharides is in the 2-8 mM range (11), but this type of assay cannot be used to determine actual dissociation constants. In the NMR assay, binding of a sugar to the protein causes a change in the chemical environment of some protons in both the sugar and the protein, and the resultant changes in chemical shifts of these resonances can be quantified and used to derive binding constants for the interaction in solution. Since sufficient amounts of both wild type and mutant CRD-4 could be produced to give samples of 0.2-1 mM, shifts in the protein spectrum were used to derive binding constants in this study.
The 1 H NMR spectrum of CRD-4 has good dispersion, characteristic of a folded protein (not shown). When ␣-methyl mannoside (␣-Me-Man) was titrated in, shifts of several peaks were observed, but the greatest change occurred in a resonance in the aromatic region of the spectrum at about 8.2 ppm and in the  Fig. 6. Although it is not possible to assign these peaks to particular residues of the protein, it is possible to say what type of residues they represent. The peak in the aromatic region (Fig. 6A) is likely to be from a histidine proton, since it shows a characteristic shift on change of pH (data not shown). Since the most upfield protons in protein spectra are the methyl protons of valine, leucine, and isoleucine residues (21), it is likely to be one of these residues in CRD-4 that is affected by binding of sugar. Examples of the binding curves obtained by quantifying the shifts in the two resonances are shown in Fig. 7.
Values for binding constants derived from titrations with methyl glycosides of mannose, L-fucose, and GlcNAc, the three sugars for which CRD-4 is specific are shown in Table II. ␣-Me-Man, ␣and ␤-Me-Fuc, and ␣-Me-GlcNAc all produced similar shifts in the methyl proton resonance (⌬␦ B methyl). Similar shifts in the histidine proton resonance (⌬␦ B histidine) were produced by ␣-Me-Man and ␣-Me-Fuc, but binding of ␤-Me-Fuc caused a smaller shift in this peak. Shifts in the histidine residue during the titration of ␣-Me-GlcNAc could not be quantified, since an alkaline contaminant in the stock solu-tion of this sugar caused the pH to rise from 7.2 to 7.4 over the course of the experiment, which in itself caused large shifts in this resonance. However, experiments with free GlcNAc showed that this sugar also causes the histidine peak to shift to the same extent as ␤-Me-Fuc, although with GlcNAc it shifted upfield rather than downfield. The fact that each of the sugars bound by the domain produces shifts in the same resonances in the protein spectrum confirms that they are all binding at the same site.
The dissociation constants obtained for the methyl glycosides show that CRD-4 binds ␣-Me-Man more tightly than ␣-Me-GlcNAc and ␤-Me-Fuc. These results are in good agreement with inhibition studies with free monosaccharides in a solid phase assay, where mannose and fucose showed approximately equal inhibition of neoglycoprotein binding to CRD-4, with GlcNAc being a less potent inhibitor (11). The results are also similar to the dissociation constants obtained by NMR for binding of the same methyl glycosides to the CRD of MBP-A (15). However, CRD-4 shows a strong preference for ␣-Me-Fuc, which binds 10 times more tightly than ␤-Me-Fuc.
Effects of Binding on Sugar Spectra-Although shifts in the protein spectrum were used to derive dissociation constants, analysis of the sugar spectra during the titrations also proved to be informative about the interaction between sugar and CRD The data suggest that when bound to CRD-4, C-5 and C-6 of mannose are located close to an aromatic residue.
The ring current effect is also seen in the spectra of the other glycosides tested. When ␣-Me-GlcNAc interacts with CRD-4, the most marked effect is on the H-2 resonances, which are suppressed relative to H-6, H-6Ј, and H-3 (Fig. 8B). The Omethyl group resonance is also suppressed when ␣-Me-GlcNAc binds to CRD-4 (not shown). Therefore, ␣-Me-GlcNAc bound to CRD-4 must be orientated such that C-2 and the O-methyl group are close to an aromatic residue. In the spectrum of ␣-Me-Fuc, the greatest ring current effect is seen on the Omethyl resonance, with smaller effects on H-1 and H-5, whereas the ring current effect is not apparent in the spectrum of ␤-Me-Fuc (data not shown).
Role of Tyr 729 in Sugar Binding by CRD-4 -Comparison of the sequence of CRD-4 and MBP-A suggests that the residue most likely to be causing the ring current effect is Tyr 729 . This residue is at the position corresponding to His 189 of MBP-A (Fig. 1). In the crystal structure of MBP-A, both the ␤-carbon and the imidazole ring of His 189 are in van der Waals contact with mannose bound to Ca 2ϩ 2, but mutagenesis combined with NMR have shown that only the ␤-carbon contributes significantly to the net binding energy (15). No ring current shifts are seen in the spectrum of ␣-Me-Man in the presence of MBP-A (15). However, in most galactose-binding C-type CRDs, tryptophan is present at the position corresponding to His 189 of MBP-A (4). NMR analysis of a version of MBP-A that has been mutated to incorporate tryptophan at position 189 so that it binds galactose does show a ring current shift in the spectrum of galactose, indicative of an interaction between galactose and this tryptophan (23). Sequence comparisons and molecular modeling show that the Ca 2ϩ 2 binding site of MBP-A is conserved in CRD-4, and data presented above provide evidence that mannose binds at this site. Thus, it is likely that Tyr 729 will be close enough to interact with sugars bound at the conserved Ca 2ϩ site.
Site-directed mutagenesis was used to determine the role of Tyr 729 in sugar binding by CRD-4. Domains with Tyr 729 changed to alanine, leucine, or histidine showed reduced binding to mannose-Sepharose but bound well enough that they could be purified for analysis by NMR. In contrast, a domain with glycine at position 729 did not bind to mannose-Sepharose, so it could not be purified for further analysis. CRD-4 with phenylalanine at position 729 showed wild type binding to mannose-Sepharose and was also analyzed by NMR.
Binding curves obtained from titrations of mutated domains with ␣-Me-Man are shown in Fig. 9. The binding constants derived from these data are summarized in Table III. For each mutant domain, the maximum shifts in the histidine resonance and in the methyl proton resonance were similar to those for wild type CRD-4, except that the methyl proton resonance shifted less in the Tyr 729 3 Leu mutant. These data suggest that the sugar binding site is not grossly rearranged by any of the mutations.
The change Tyr 729 3 Ala decreases the affinity of CRD-4 for ␣-Me-Man by a factor of 4 (Table III), indicating that interactions with the side chain of Tyr 729 must contribute significantly to the binding energy. Mutation of Tyr 729 to Phe has no effect on ␣-Me-Man binding, indicating that the hydroxyl group of Tyr 729 does not play a role in sugar binding. Mutation of Tyr 729 to His causes less than 2-fold decrease in affinity for ␣-Me-Man, indicating that the imidazole ring of histidine can interact with bound sugar in a similar manner to the aromatic ring of tyrosine and phenylalanine. With leucine at position 729, the affinity for ␣-Me-Man is decreased somewhat more, indicating that a hydrophobic, nonaromatic side chain can also partially compensate for the loss of the aromatic ring of tyrosine.
Inspection of the spectrum of ␣-Me-Man in the presence of the mutated domains confirms that interaction of the sugar with Tyr 729 causes the ring current shifts observed in the  3 Leu mutants (Fig. 10, A and B). important for binding to sugars other than mannose. Values of K i obtained for mannose and GlcNAc are shown in Table IV. Comparison of the absolute values of K i obtained for wild type CRD-4 and the domain with alanine at position 729 is not meaningful, since the affinity of the mutant domain for the reporter ligand is greatly reduced. Comparison of the relative inhibitory potencies of mannose and GlcNAc, given by K i, Glc -NAc/K i, mannose , gives similar values for the wild type and mutated domains. Since the NMR assay showed that mutation of Tyr 729 to Ala decreases the affinity for mannose by a factor of 4 when compared with wild type CRD-4, the affinity for GlcNAc must also decrease by a factor of 4 for the value of K i, GlcNAc / K i, mannose to be the same as that of wild type. These data, combined with the fact that ring current shifts are observed in the spectra of other sugars in the presence of CRD-4, make it possible to conclude that interaction with Tyr 729 is an important factor in binding of GlcNAc and fucose as well as of mannose to  ⌬G values for binding of ␣-Me-Man to wild type and mutated CRD-4 can be calculated from the values for the dissociation constant obtained in this study. Comparison of ⌬G values for wild type CRD-4 (Ϫ15.1 KJ/mol) and for the Tyr 729 3 Ala mutant (Ϫ11.5 KJ/mol) shows that interaction of the sugar with the aromatic ring of Tyr 729 provides 3.6 KJ/mol (24%) of the total free energy of binding. The contribution to free binding energy by this interaction in CRD-4 is similar to that of the ␤-carbon of His 189 to binding of mannose by MBP-A (15). Since mutation of Tyr 729 to Gly resulted in a domain that could not be purified, it is not possible to say whether the ␤-carbon of Tyr 729 is also important for sugar binding to CRD-4. Lack of binding to mannose-Sepharose could be due to loss of affinity because interactions with the ␤-carbon as well as the aromatic ring contribute to sugar binding, but it could also be due to misfolding of the domain.

Localization of Ca 2ϩ Binding Sites in CRD-4
Conserved Ca 2ϩ Sites in C-type Lectins-Most other C-type CRDs for which Ca 2ϩ -dependent sugar binding activity has been demonstrated contain the five residues (two glutamic acids, two asparagines, and an aspartic acid) present at the conserved Ca 2ϩ site in MBP-A, E-selectin, and CRD-4, except that in CRDs that bind galactose rather than mannose, one glutamate and one asparagine are changed to glutamine and aspartate, respectively (4,24). Although each of the CRDs will probably have additional contacts with the bound sugar, such as that seen between the ␤-carbon of His 189 of MBP-A and mannose (15), it is likely that in each case, the main interaction with sugar will be via direct ligation to Ca 2ϩ at this conserved site. Either the MBP-A pattern of ligation or the alternative ligation pattern seen in E-selectin could be utilized. However, the presence of the ligands for the conserved Ca 2ϩ in a CRD  does not necessarily mean that this CRD will bind sugar. The type II antifreeze protein of herring, which is homologous to C-type CRDs, contains all of the ligands for the conserved Ca 2ϩ and does bind one Ca 2ϩ , probably at this site, but it does not bind carbohydrate (25). Equally, some of the more divergent groups of C-type lectins, including the type II transmembrane proteins found on natural killer cells, may utilize a completely different mechanism for binding sugar and/or Ca 2ϩ , since these proteins do not contain the ligands for the conserved Ca 2ϩ (4,26,27).

Position of Proposed Auxiliary Ca 2ϩ Binding Site in CRD-4 -
The positions of potential Ca 2ϩ ligands mutated in this study are shown on the molecular model of CRD-4 in Fig. 11. Four residues, Asn 728 , Asn 731 , Glu 737 , and Asn 750 , are predicted, based on the mutagenesis results, to be involved in ligation of the auxiliary Ca 2ϩ (Fig. 11A). Mutagenesis of nine other potential Ca 2ϩ ligands (Fig. 11B), including three of the four residues at positions equivalent to the Ca 2ϩ 1 ligands in MBP-A, has no effect on Ca 2ϩ binding. The results of the mutagenesis of potential Ca 2ϩ ligands in CRD-4 indicate that the auxiliary Ca 2ϩ of CRD-4 is not ligated at the same site as in MBP-A or in the same way.
Three of the four residues predicted to be ligands for the auxiliary Ca 2ϩ , Asn 728 , Asn 731 , and Asn 750 , are located close to each other in the model structure. In this orientation, it would be possible for them to form a Ca 2ϩ binding site. The fourth residue, Glu 737 , located on ␤-strand 3, is too far away from the other three residues for it to be part of a Ca 2ϩ binding site involving all four residues. Although Fig. 11 is only a model of CRD-4, based on the structure of a protein with which it shares 28% sequence identity, it is difficult to imagine sufficient rearrangement of ␤-strand 3 to bring Glu 737 close enough to take part in direct Ca 2ϩ ligation with the other three residues.
If Glu 737 is not coordinated directly to the auxiliary Ca 2ϩ , then other possibilities must be considered to explain the phenotype caused by mutation of this residue to alanine. In the CRDs of E-and P-selectin, glutamate is also found at the position equivalent to Glu 737 . In the crystal structure of Eselectin, this glutamate residue (Glu 92 ) forms an ion pair with a lysine residue (Lys 113 ) and a hydrogen bond with the amide group of Asn 105 , which is a ligand for the conserved Ca 2ϩ (7). Mutation of this glutamate to alanine in both E-and P-selectin reduces their ability to bind to sugar ligands (28,29). It seems likely that this reduction in binding ability is due to loss of the hydrogen bond between Glu 92 and Asn 105 and that this affects the ability of the domain to ligate Ca 2ϩ at the conserved site. By analogy, it might be proposed that the Glu 737 3 Ala change in CRD-4 indirectly affects the ability of the domain to bind the conserved Ca 2ϩ and not the auxiliary Ca 2ϩ . However, it is difficult to rationalize the phenotype of the Glu 737 3 Ala mutation with an effect on binding of the conserved Ca 2ϩ . This mutant domain shows first order dependence on Ca 2ϩ for sugar binding, indicating that it is binding only one Ca 2ϩ , and its  sugar binding activity is reduced. Assuming that sugar binds to the conserved Ca 2ϩ , the fact that the domain retains the ability to bind sugar, albeit weakly, indicates that it must be the conserved Ca 2ϩ that is present and the auxiliary Ca 2ϩ that is absent. Thus, it is most likely that mutation of Glu 737 affects binding of the auxiliary Ca 2ϩ . It is possible that the interaction between Glu 737 and the auxiliary Ca 2ϩ is not direct but involves bridging via a water molecule as has been seen in a number of Ca 2ϩ -binding proteins including E-selectin (7). A modest change in the side chain conformation predicted by the model would suffice to bring Glu 737 into a position where it could be involved in ligation of the auxiliary Ca 2ϩ through an intermediary water molecule. Alternatively, it is possible that Glu 737 is involved in an interaction that orients a backbone carbonyl oxygen atom that interacts directly with the auxiliary Ca 2ϩ . In MBP-A, backbone carbonyl oxygens are seen to be involved in the ligation of both Ca 2ϩ .
The phenotypes caused by mutation of the other three proposed auxiliary Ca 2ϩ ligands could also be explained by indirect effects. However, all other side chains with the potential to ligate Ca 2ϩ located in the top half of the molecule have been mutated with no effect on Ca 2ϩ binding. Thus, it seems likely that at least some of the four amino acids proposed to form the auxiliary Ca 2ϩ binding site must coordinate directly to Ca 2ϩ . 7-Fold co-ordination of Ca 2ϩ is usually seen in Ca 2ϩ -protein complexes (30). In MBP-A, the shell of ligands for the auxiliary Ca 2ϩ is supplied by the two carboxylate oxygens of the side chain of Glu 165 , one carboxylate oxygen from the side chains of each of the three aspartate residues, the backbone carbonyl oxygen of Glu 193 , and the oxygen of a water molecule (5). It is possible that an eighth ligand is supplied by the other carboxylate oxygen of Asp 161 . Assuming that the three asparagine residues proposed to be involved in ligation of the auxiliary Ca 2ϩ to CRD-4 make direct contact with the Ca 2ϩ but that Glu 737 does not, three ligands for the Ca 2ϩ would be supplied. The oxygen atom from the carbonyl group of each of the three asparagine residues could form bonds to the Ca 2ϩ . This means that four bonds must be supplied by either backbone carbonyl oxygens or water molecules.
Function of the Auxiliary Ca 2ϩ in CRD-4 -The phenotypes produced by mutation of residues at this site in CRD-4 suggest that, as in MBP-A, the auxiliary Ca 2ϩ in CRD-4 is required for correct positioning of loops around the sugar binding site but is not involved in direct ligation of sugar. The different modes of ligation of the auxiliary Ca 2ϩ seen in CRD-4 of the mannose receptor and MBP-A might be related to the different functions of the two proteins. The mannose receptor is an endocytic receptor, which must release its ligands in the acidic environment of the endosome. Since stability of Ca 2ϩ binding to CRD-4 is pH-dependent, loss of Ca 2ϩ binding at low pH probably contributes to release of glycoconjugates by the mannose receptor in the endosomes (13). It is interesting to note that the magnitude of the increase (approximately 10-fold) in K Ca seen when binding of the auxiliary Ca 2ϩ to CRD-4 is disrupted by mutation is the same as the increase in K Ca seen when the pH is reduced from 7.8 to 5.0 (13). Thus, it can be imagined that a change in conformation in CRD-4 at low pH causes loss of Ca 2ϩ from the auxiliary site, which leads to a decrease in affinity for Ca 2ϩ , so that the conserved Ca 2ϩ and its bound sugar are released. In contrast, MBP-A, which is a soluble extracellular protein that initiates fixation of complement after binding to carbohydrates on the surface of pathogens, is not known to require pH-dependent release of Ca 2ϩ or sugar (1).
Comparison with Other C-type CRDs-These studies increase to three the different modes of Ca 2ϩ ligation to C-type CRDs. The structure of E-selectin reveals that the conserved Ca 2ϩ alone can support sugar binding, while the mannosebinding proteins and CRD-4 of the mannose receptor demonstrate that there are at least two alternative ways in which an auxiliary site can be formed.
Two other CRDs, those of the chicken hepatic lectin and the rat asialoglycoprotein receptor, have been shown to bind two Ca 2ϩ in solution but have not yet been crystallized (31,32). These CRDs have the same residues as the MBPs at both Ca 2ϩ binding sites (4). Examination of the sequences of C-type CRDs for which Ca 2ϩ -dependent sugar binding has been demonstrated but the stoichiometry of the requirement for Ca 2ϩ has not been determined indicates that the ligands for the auxiliary Ca 2ϩ of the MBPs are present in most of them (4). Of these CRDs, only those of the placental mannose receptor and the bovine lectin CL-43 contain more than two of the ligands for the auxiliary Ca 2ϩ in CRD-4, and these only contain three (33,34). Therefore, it is probable that if these proteins do bind a second Ca 2ϩ , it will be ligated in a similar manner to the auxiliary Ca 2ϩ of the MBPs.
So far, only the CRDs of E-selectin and of the type II antifreeze protein of herring have been shown to bind only one Ca 2ϩ (7,25). Solution binding studies on another member of A, side chains predicted, as a result of mutagenesis, to be involved in ligation of Ca 2ϩ at the auxiliary site. B, side chains mutated but found to have no effect on Ca 2ϩ binding. The black sphere shows the position of Ca 2ϩ at the conserved site. the selectin family, P-selectin, have shown that this protein binds two Ca 2ϩ (35). However, this study was carried out on the whole protein, not just the CRD, so it is possible that one of the Ca 2ϩ is bound to the epidermal growth factor-like domain or to one of the complement-regulatory domains. Residues found at positions equivalent to the auxiliary Ca 2ϩ binding site of MBP-A are Lys, Asn, Asn, and Asp in E-selectin and P-selectin, and Lys, Gly, Asn, and Asp in L-selectin. Only two of the four proposed ligands for the auxiliary Ca 2ϩ in CRD-4 are present in each of the selectins (4). Thus, it seems unlikely that any of the selectins will bind a second Ca 2ϩ at either the auxiliary site of MBP-A or the proposed auxiliary Ca 2ϩ site of CRD-4.
Ca 2ϩ -dependent sugar binding has not yet been demonstrated for many of the more divergent members of the C-type lectin family, including other CRDs of the mannose receptor. Of the eight CRDs of the mannose receptor, apart from CRD-4, only CRD-5 contains all of the conserved Ca 2ϩ binding site residues of MBP-A, indicating that Ca 2ϩ and sugar could be ligated at this site. Sugar binding activity was not detected for CRD-5 when it was expressed alone, but binding studies with different combinations of expressed mannose receptor CRDs indicate that CRD-5, as well as CRDs 6, 7, and 8 must have the ability to interact with sugar (11,12). Neither the auxiliary Ca 2ϩ binding site of the MBPs or the proposed auxiliary site of CRD-4 is conserved in CRD-5 or in the other six mannose receptor CRDs. Thus, the proposed method of ligation of the auxiliary Ca 2ϩ in CRD-4 is not a general mechanism for Ca 2ϩ ligation at this site in the other mannose receptor CRDs. However, the residues identified here as ligands for the auxiliary Ca 2ϩ in CRD-4 of the human mannose receptor are all conserved in CRD-4 of the mouse mannose receptor (36).
It is also not yet clear whether the multiple CRDs of the phospholipase A 2 receptor and of DEC-205 are capable of Ca 2ϩdependent sugar binding. No studies of ligand binding to DEC-205 have been reported. The phospholipase A 2 receptor has been shown to bind neoglycoprotein ligands, but this binding is not Ca 2ϩ -dependent (8). Since the secretory phospholipases that the phospholipase A 2 receptor binds with high affinity are not glycosylated, the physiological relevance of sugar binding by this receptor is unclear (8,9). Few of the ligands for either the conserved Ca 2ϩ or the auxiliary Ca 2ϩ of the MBPs or CRD-4 are present in the CRDs of the phospholipase A 2 receptor or DEC-205 (8 -10). If the CRDs of these two proteins do bind sugar and/or Ca 2ϩ , it must be by a different mechanism from that seen in MBP-A or CRD-4 of the mannose receptor.
From the comparisons above, it can be seen that the proposed binding site for the auxiliary Ca 2ϩ in CRD-4 is not conserved in other CRDs, indicating that this is not a general mechanism for Ca 2ϩ binding to CRDs that do not contain the ligands for the auxiliary Ca 2ϩ of MBP-A. Clearly, sequence comparisons alone are insufficient for determining the Ca 2ϩ binding characteristics of a C-type CRD, and physical techniques will have to be applied.

Monosaccharide-Protein Interactions Stacking Interactions in Protein-Carbohydrate Complexes-
The stacking interaction between the aromatic ring of Tyr 729 and bound sugar in CRD-4 is a common feature of proteincarbohydrate complexes (2,37). In all structures of proteins in complex with galactose, including plant lectins, bacterial toxins, animal S-type lectins, and the galactose-binding mutant of MBP-A, the nonpolar B-face of galactose stacks against tryptophan or phenylalanine residues (2). However, stacking of mannose against an aromatic residue, as shown here for CRD-4, is not always seen.
In conconavalin A, C-6 and C-5 of mannose stack against a tyrosine residue, as predicted for binding of mannose to CRD-4 (38). In other mannose-specific legume lectins, C-6 and C-5 stack against a phenylalanine residue (39,40). Stacking of mannose is not seen in the snowdrop lectin or in either MBP-A or MBP-C (5,6,41). MBP-C has been co-crystallized with GlcNAc and fucose as well as mannose. These sugars bind at the same site as mannose, and neither of them stack against an aromatic ring. Only one other GlcNAc-binding lectin, wheat germ agglutinin, has been crystallized with GlcNAc, and in this case stacking of GlcNAc against tyrosine is seen (42). Thus, while there is precedent for stacking of mannose and GlcNAc in plant lectins, CRD-4 of the mannose receptor provides the first example of this type of interaction with these sugars in an animal lectin.
Sequence comparisons of C-type CRDs allows prediction of whether stacking interactions with bound sugar will be seen in those that have not yet been analyzed, either by crystallography or by NMR binding studies. All C-type CRDs shown to bind galactose have an aromatic residue at the position equivalent to Tyr 729 of CRD-4 and His 189 of MBP-A (4,26), and tryptophan at this position is seen to stack against galactose in the crystal structure of the MBP-A galactose-binding mutant (43). Thus, stacking of galactose is likely to be a common feature of all galactose-binding C-type CRDs.
Of the mannose/fucose/GlcNAc-specific C-type CRDs, none apart from MBP-A has an aromatic residue at the position equivalent to Tyr 729 (4,26). His 189 at this position in MBP-A, however, is splayed away from bound mannose and makes only an edgewise contact with the 2-hydroxyl group (5). The only C-type lectin to show strong preference for GlcNAc rather than mannose, the chicken hepatic lectin, also lacks an aromatic residue at this position (26). It is also interesting to note the absence of aromatic residues in the corresponding position of the other CRDs of the mannose receptor (3). Therefore, CRD-4 of the mannose receptor appears to be the only known C-type CRD in which hydrophobic stacking against an aromatic residue is important for binding of mannose, GlcNAc, or fucose.
Position of Tyr 729 in CRD-4 -The data presented here show that replacing Tyr 729 of CRD-4 with histidine has only a small effect on the ability of the domain to bind mannose and that C-5 and C-6 can interact with the imidazole ring in the same way as with the aromatic ring of tyrosine. It is therefore probable that the exact position of Tyr 729 in CRD-4 is different from that of His 189 , the equivalent residue in MBP-A, bringing it closer to the bound sugar. In galactose-binding C-type CRDs, a glycinerich loop is inserted adjacent to the tryptophan that stacks against galactose. This glycine-rich loop, which is not present in the mannose/fucose/GlcNAc-specific CRDs, including CRD-4, locks the tryptophan residue into the optimal position for stacking against galactose (23,43). It seems that a difference between CRD-4 and MBP-A in the arrangement of the loops in this area of the molecule must be responsible for positioning Tyr 729 so that it can stack against bound sugars. This difference in the arrangement of loops around the sugar binding site could be brought about by the different modes of ligation of the auxiliary Ca 2ϩ in the two proteins.
Orientation of Monosaccharides Bound to CRD-4 -Observation of which protons of each sugar are affected by the ring current of Tyr 729 allows some conclusions to be made about the orientation of monosaccharides bound to CRD-4. When ␣-Me-Man is bound, C-5 and C-6 stack against Tyr 729 , whereas it is C-2 and the O-methyl group that interact with Tyr 729 when ␣-Me-GlcNAc is bound. Assuming that each of these two sugars is ligated to the conserved Ca 2ϩ via hydroxyls 3 and 4, as is the case in the MBPs, then bound GlcNAc must be rotated 180°r elative to bound mannose.
The orientations of ␣and ␤-Me-Fuc, which would ligate to Ca 2ϩ via hydroxyls 2 and 3, would also be predicted to differ by 180°with respect to each other, since C-1 and the O-methyl group of ␣-Me-Fuc interact with Tyr 729 , whereas no ring current effect is apparent when ␤-Me-Fuc binds. Thus, in bound ␣-Me-Man and ␤-Me-Fuc, C-5 and C-6 would be positioned near Tyr 729 , while C-1 and C-2 of ␣-Me-GlcNAc and ␣-Me-Fuc would be closest to the tyrosine. The difference in bound orientation of ␣and ␤-methyl fucosides and the consequent lack of interaction of ␤-Me-Fuc with Tyr 729 probably account for the 10-fold tighter binding of ␣-Me-Fuc.
Although both MBP-A and MBP-C bind GlcNAc and fucose as well as mannose, only MBP-C has been crystallized in complex with all three sugars (6). Unlike what appears to be the case in CRD-4, mannose and GlcNAc and the two anomers of fucose bind in the same orientation to MBP-C. However, mannose bound to MBP-A is rotated 180°with respect to mannose and GlcNAc bound to MBP-C (5). In MBP-C, C-5 and C-6 of these two sugars are closest to Val 194 , the residue equivalent to Tyr 729 , whereas in MBP-A, C-1 and C-2 of mannose are closest to His 189 . Since the position of Tyr 729 is likely to be shifted slightly compared with that of the equivalent residue in each of the MBPs, it is not possible to predict with certainty whether the absolute orientation of mannose, with respect to the binding of hydroxyls 3 and 4, in CRD-4 is the same as in MBP-A or as in MBP-C. Nevertheless, it is clear that the relative orientations of various sugars bound to CRD-4 must be different from either MBP.
Conclusions-While the major interaction between sugars and CRD-4 is via direct ligation to a conserved Ca 2ϩ , as is seen in the mannose-binding proteins, other aspects of sugar and Ca 2ϩ binding to CRD-4 are different. CRD-4 ligates a second Ca 2ϩ in a unique way, and this difference in the mode of ligation of the auxiliary Ca 2ϩ is likely to cause a difference in the position of loops around the sugar binding site when compared with other C-type CRDs. A stacking interaction with Tyr 729 provides approximately 25% of the binding energy for mannose binding to CRD-4, and interaction with this residue is also important for binding to GlcNAc and fucose. Among C-type CRDs, CRD-4 seems to have evolved a unique way of binding to mannose, GlcNAc, and fucose.