HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

J. Biol. Chem., Vol. 275, Issue 28, 21539-21548, July 14, 2000

This Article Abstract Full Text (PDF) All Versions of this Article: 275/28/21539    most recent M002366200v1 Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Feinberg, H. Articles by Weis, W. I. Search for Related Content PubMed PubMed Citation Articles by Feinberg, H. Articles by Weis, W. I. Social Bookmarking                      What's this?

Structure of a C-type Carbohydrate Recognition Domain from the Macrophage Mannose Receptor^*

Hadar Feinberg, Shaun Park-Snyder, Anand R. Kolatkar, Charles T. Heise, Maureen E. Taylor, and William I. Weis^§

From the Department of Structural Biology, Stanford University School of Medicine, Stanford, California 94305 and  Glycobiology Institute, Department of Biochemistry, University of Oxford, Oxford OX1 3QU, United Kingdom

Received for publication, March 21, 2000, and in revised form, April 12, 2000

	ABSTRACT

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS AND DISCUSSION REFERENCES

The mannose receptor of macrophages and liver endothelium mediates clearance of pathogenic organisms and potentially harmful glycoconjugates. The extracellular portion of the receptor includes eight C-type carbohydrate recognition domains (CRDs), of which one, CRD-4, shows detectable binding to monosaccharide ligands. We have determined the crystal structure of CRD-4. Although the basic C-type lectin fold is preserved, a loop extends away from the core of the domain to form a domain-swapped dimer in the crystal. Of the two Ca²⁺ sites, only the principal site known to mediate carbohydrate binding in other C-type lectins is occupied. This site is altered in a way that makes sugar binding impossible in the mode observed in other C-type lectins. The structure is likely to represent an endosomal form of the domain formed when Ca²⁺ is lost from the auxiliary calcium site. The structure suggests a mechanism for endosomal ligand release in which the auxiliary calcium site serves as a pH sensor. Acid pH-induced removal of this Ca²⁺ results in conformational rearrangements of the receptor, rendering it unable to bind carbohydrate ligands.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS AND DISCUSSION REFERENCES

The mannose receptor of macrophages and liver endothelial cells acts as a molecular scavenger, binding to and internalizing a variety of pathogenic microorganisms and potentially harmful glycoproteins (1, 2). The receptor is well suited to this scavenging function, since it has an extracellular region containing multiple domains that allow recognition of a diverse range of glycoconjugate ligands. The ability of the receptor to mediate both endocytosis and phagocytosis also means that it can facilitate clearance of both particulate and soluble ligands. Once internalized, ligands are released from the receptor following endosomal or phagosomal acidification, and the receptor recycles to the cell surface.

The role of the mannose receptor in the innate immune response is well documented, with several clinically important pathogens including Mycobacterium tuberculosis and Pneumocystis carinii subject to opsonin-independent phagocytosis by the receptor. More recent studies indicate that the receptor may play a central role in coordinating the innate and adaptive immune responses by enhancing uptake and processing of soluble glycoconjugates released from pathogens for presentation to T cells by major histocompatibility complex class II molecules (1, 2). Levels of endogenous proteins bearing high mannose oligosaccharides such as lysosomal enzymes and tissue plasminogen activator are also regulated by the mannose receptor. These proteins are released from tissues into the blood in response to pathological events. The receptor also plays an important role in removing pituitary hormones such as lutropin and thyrotropin from the circulation after they have acted on their target cells (3, 4).

The primary structure of the mannose receptor (5) reflects its diverse carbohydrate specificity. An N-terminal cysteine-rich domain mediates recognition of sulfated N-acetylgalactosamine, which is the terminal sugar of the unusual oligosaccharides present on pituitary hormones (6). The cysteine-rich domain is followed by a fibronectin type II repeat, the function of which is unclear. The remaining extracellular portion of the receptor consists of eight tandemly repeated C-type (Ca²⁺-dependent) carbohydrate recognition domains (CRDs).¹ Several of the eight C-type CRDs, but not the cysteine-rich domain or the fibronectin type II repeat, are involved in Ca²⁺-dependent recognition of mannose, fucose, or N-acetylglucosamine (GlcNAc) residues on the surface of pathogens or at the termini of oligosaccharides of endogenous glycoproteins (7, 8). The mannose receptor is the prototype of a family of receptors sharing the same overall domain organization, although it is the only one known to bind carbohydrate ligands (9).

Of the eight C-type CRDs, CRDs 4-8 are required for binding and endocytosis of mannose/GlcNAc/fucose-terminated ligands, but only CRD-4 has demonstrable sugar binding activity in isolation. The requirement of multiple CRDs for high affinity ligand binding and endocytosis probably reflects the requirement for multivalent binding, which is a common feature of C-type and other lectins (10). CRD-4 forms part of a protease-resistant ligand binding core with CRD-5 and is central to sugar recognition by the receptor. Studies combining ligand binding assays, site-directed mutagenesis, and NMR have given a detailed picture of how CRD-4 binds to sugar and Ca²⁺ (11-13). CRD-4 of the mannose receptor has specificity for mannose, GlcNAc, and fucose like the C-type CRDs of rat serum and liver mannose-binding proteins (MBPs), which have been well characterized by crystallography, NMR, and mutagenesis (10). Mannose receptor CRD-4 binds to two Ca²⁺ as do MBPs. Some aspects of the mode of binding of sugar and Ca²⁺ by CRD-4 are similar to those of the MBP CRDs, but others are different (12), indicating that mannose binding by the mannose receptor probably evolved separately from mannose binding by other C-type lectins.

Understanding the molecular basis of cell surface ligand recognition and endosomal release by the mannose receptor requires information about how individual domains interact with sugars as well as the structural arrangement of the multiple domains. As CRD-4 is the central ligand binding domain of the receptor, detailed analysis of this domain represents a first step toward understanding the receptor as a whole. This paper describes structural analysis of CRD-4 of the mannose receptor by crystallography. The structure obtained appears to represent a non-sugar binding form of the domain that reveals a possible mechanism for release of sugar ligands from the receptor in the acid environment of the endosome. The structure also suggests ways in which multiple CRDs in the whole receptor might interact with each other.

	EXPERIMENTAL PROCEDURES

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS AND DISCUSSION REFERENCES

Protein Expression and Crystallization-- CRD-4 of the human mannose receptor was produced in Escherichia coli as described previously, except that DNA coding for CRD-4 in the OmpA expression vector was modified to remove the portion specifying 12 amino acids at the N-terminal end of the domain (11). These residues form part of the linker region between CRD-4 and CRD-3. The recombinant protein contains residues 626-769 of the mature mannose receptor (5), with the sequence Gly-Ile fused to its N terminus and an extra Asp at the C terminus. The protein was purified by affinity chromatography on mannose-Sepharose followed by reverse-phase HPLC as described previously (11). Lyophilized CRD-4 was dissolved in water to a concentration of approximately 40 mg/ml and neutralized with 100 mM NaOH to give a final pH of 8.0. The protein was diluted to a concentration of 12 mg/ml protein, and CaCl₂ was added to a final concentration of 3-5 mM. Crystals were grown at 21 °C by hanging drop vapor diffusion by mixing 1 µl of protein solution with 1 µl of reservoir solution containing 8-13% polyethylene glycol 10,000 (Aldrich), 100 mM Tris, pH 8.0, 100 mM NaCl, and 10 mM CaCl₂ (solution A). For the native crystal used in data collection, solution A also contained 45 mM MgCl₂, although Mg²⁺ is not required for crystal growth. Clusters or single thin plates (typical size 0.25 × 0.10 × 0.02 mm³) appeared within a week. The crystals belong to space group C2 and contain two molecules per asymmetric unit, corresponding to a solvent content of 40%. For data collection, crystals were transferred to 20 µl of solution A for 20 min, then transferred to 20 µl of solution A containing 20% 2-methyl-2,4-pentanediol for another 5 min and flash-cooled to 100 K in a nitrogen gas stream.

The crystal used for the methyl 3-O-(-D-mannopyranosyl)--D-mannopyranoside (Man(1,3)Man) disaccharide sugar soak was grown from a protein stock that also contained 100 mM -O-methyl mannoside (Sigma). This stock was used in an attempt to improve crystal quality, but the presence of the sugar proved to not affect crystal growth. The crystal was transferred to 20 µl of solution A containing 200 mM Man(1,3)Man (V-Labs, Inc.) for 40-60 min before transferring the crystals to 20 µl of the cryoprotectant solution, which also contained 200 mM Man(1,3)Man. It is likely that any monosaccharide present in the original crystal was removed by the soaking procedure.

Structure Determination and Refinement-- Diffraction data from native and Man(1,3)Man-soaked crystals were measured to maximum Bragg spacings of 2.3 Å on a MAR Image plate detector at the Stanford Synchrotron Radiation Laboratory (SSRL) beam line 9-1. Intensities were integrated, scaled, and merged with DENZO and SCALEPACK (14) (Table I).

The structure was solved by molecular replacement using the program AMORE (15) using the native CRD-4 data. Several different search models were tried, and a composite search model constructed by superimposing the CRDs from three C-type CRDs (residues 1 to 122 of human E-selectin (16) (Protein Data Bank (PDB) code 1esl), residues 14 to 144 of human lithostathine (17) (PDB code 1lit), and residues 45 to 181 of human tetranectin (18) (PDB code 1tn3)) gave the clearest results. Ca²⁺ and water molecules were removed from the search model, but all side chains and the refined individual atomic temperature factors were retained. Two strong peaks (6.1 and 5.7 S.D. over the mean ()) were found in a cross-rotation function that was calculated using reflections between 12 and 3.0 Å (next highest peak 4.2). The highest 15 rotation solutions were used for a translation search, and after rigid-body refinement, the best solution, corresponding to the highest rotation and translation function solutions, gave an R-value of 51.6% and a correlation coefficient of 35.6% (copy A). Another round of translation search was done while fixing the first molecule, using the next 14 sets of rotation angles found in the cross-rotation procedure. The highest correlation coefficient and lowest R-value corresponded to the second highest peak of the rotation function (copy B). After rigid-body refinement of the two molecules in the asymmetric unit, the R-value was 48.2%, and the correlation coefficient was 46.9%. The two molecules in the asymmetric unit are related by a 2-fold rotation axis almost parallel to the b axis, explaining why a self-rotation function did not show a 2-fold non-crystallographic symmetry axis. Subsequent phase improvement was done using DM (19). Electron density maps were calculated using data to 2.6Å after solvent flattening or solvent flipping and with or without 2-fold averaging. 10% of the data were omitted for cross-validation (20) in DM and, later, in refinement. An initial model was constructed using the molecular graphics program O (21). At this stage some amino acid residues at the N and C termini of the protein and in loop regions could not be interpreted, but most of the structure comprising the core of the protein could be determined clearly from the map.

The model was subjected to simulated annealing and iterative cycles of positional and temperature factor refinement followed by manual fitting and rebuilding. All refinement procedures employed the maximum likelihood amplitude target in CNS (22). An overall anisotropic temperature factor and a bulk solvent correction were applied throughout. Strict non-crystallographic symmetry constraints were applied in the first 10 rounds, and some connections in loop regions could be placed unambiguously. Subsequently, tight non-crystallographic symmetry restraints (force constant, 300 kcal mol¹) were applied to allow regions of significant difference to be refined properly. Extending the resolution to 2.3 Å made the identification of most amino acids possible. Three Ca²⁺ ions were identified, one in each protomer and one shared between copy A and copy B. Water molecules were assigned to peaks in F_o  F_c electron density maps at a contour greater than 3 that were within hydrogen bonding distance from a potential partner.

At this stage the first and last residues of a large loop region (residues 701-707 and 729-734) could not be placed in the maps. Attempts to build the missing regions in a manner that would connect the remainder of the molecule to produce a compact, monomeric CRD like those CRDs used in the molecular replacement calculations gave models with prohibited backbone torsion angles for those loop residues. The 2 F_o  F_c and F_o  F_c maps made from this model and also a simulated annealing omit map showed that instead of bending and turning back to form a monomeric CRD molecule, the chain in this region extends out to a symmetry-related molecule. This feature was present in both regions in copy B and one region (residues 729-734) in the other copy A (see Fig. 1). Therefore, a model of mannose receptor CRD-4 was built that consisted of two symmetry-related molecules that make a dimer by domain swapping (23) residues 702-734. Refinement of this model containing alanine residues where side chains could not be assigned from the map in the two ambiguous regions immediately dropped the R- and R-free values by 1%, and significant positive electron density could be assigned for the side chains in the loops (Fig. 1). Independent refinement of the model against the Man(1,3)Man soak data starting with a model lacking these regions also showed the domain-swapped structure. The good geometry in the dimeric model for these loop regions is further evidence of its existence in the crystal structure. Refinement was stopped when no further improvement to R-free was made. For the CRD-4 crystal that was soaked in Man(1,3)Man, no sugar could be detected in the electron density maps. Refinement statistics are given in Table II. The current CRD-4 native model contains residues 628-702 and 707-763 of copy A, residues 626-768 (residue 768 in the final model was assigned as Ala, because no side chain could be identified in the electron density maps) and the Ile preceding the first residue for copy B, 3 Ca²⁺ and, 145 water molecules. The Man(1,3)Man-soaked CRD-4 model contains residues 628-636, 639-702, and 707-762 for copy A, residues 626-767 and the N-terminal Ile for copy B, 3 Ca²⁺, and 115 water molecules.

View larger version (37K): [in this window] [in a new window]   Fig. 1.   Electron density maps (stereo) in the region I (residues 701-708)/II (residues 729-734) area showing domain swapping. Blue, 2 Fo  Fc map at the 1 level; green, +3 Fo  Fc map; red, -3 Fo  Fc map. In A and C, the model was build as a compact CRD akin to the MBP CRD; B and D depict the maps calculated using the domain-swapped model with the extended loop. A and B, region II, copy A. C and D, region I, copy B.

Analytical Ultracentrifugation-- Equilibrium sedimentation measurements were performed in a Beckman Optima XL-A analytical ultracentrifuge using an An60Ti rotor at 20,000 and 22,000 rpm and 20 °C. Equilibrium distributions from three different loading concentrations and the two different rotor speeds were analyzed simultaneously using the Nonlin curve fit program supplied with the instrument.

	RESULTS AND DISCUSSION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS AND DISCUSSION REFERENCES

Overall Structure of Mannose Receptor CRD-4 and Comparison to Other C-type Lectins-- The structure of mannose receptor CRD-4 was determined by molecular replacement using a search model composed of three superimposed C-type CRDs (see "Experimental Procedures" and Table III). The asymmetric unit of the crystal contains two molecules, which provides two independent views of the domain. In an attempt to obtain a ligand complex, diffraction data were also measured from a crystal soaked in a synthetic mother liquor containing 200 mM Man(1,3)Man. Independent refinement of a model against these data revealed no electron density for the sugar, and the structure obtained is identical to the model refined against data from the native crystal to within the experimental error. Copy A is modeled between residues 628 and 763, except for residues 703-706, whereas copy B is well defined with no chain breaks between residues 626 and 768. Although CRD-4 binds to two Ca²⁺ in solution, only the principal site is occupied in both copies. One other Ca²⁺ is present in the asymmetric unit of the crystal, but it shares ligands between copy A and copy B and is likely an adventitious site resulting from crystal packing.

View this table: [in this window] [in a new window]   Table III Comparison of CRD-4 with other C-type lectins Root-mean-square deviation (r.m.s.d.) between equivalent C atoms for CRD4-MF and other C-type lectin- or C-type-lectin-like domains. The percentage of sequence identity over equivalent positions is given in row 3. The last row shows the r.m.s.d. including only the basic core of the protein (residues 640-644, 653-664, 695-698, 735-739, 744-750, and 756-760 of mannose receptor CRD-4; 38 C positions).

The overall structure of CRD-4 (Fig. 2A) is similar to other C-type CRDs (Table III), containing two  helices and two small antiparallel  sheets. It is useful to compare CRD-4 to the only other known mannose binding C-type lectin structure, MBP (Fig. 2B). The core region of the CRD-4 domain, consisting of  strands 1-5 and the two  helices, superimposes on the equivalent residues of the rat MBP-A CRD with an r.m.s.d. of 1.5 Å. The principal difference resides in the position of helix 2, which is the most variable element of secondary structure among the known C-type lectin-like folds. In CRD-4, the helix is pushed outward relative to MBP and is reminiscent of the position of this helix in E-selectin (16). The variability of this region in the known C-type CRD structures may account for the fact that the clearest molecular replacement results were obtained using composite search models in which several different CRDs were superimposed.

View larger version (21K): [in this window] [in a new window]   Fig. 2.   Structure of the CRD-4 monomer and comparison to rat mannose-binding protein A. A, ribbon diagram of CRD-4. Disulfide bonds are shown in pink ball-and-stick representation, and the Ca2+ is shown as a blue-green sphere. The two segments that connect the extended loop to the core of the CRD, region I (residues 701-708) and region II (residues 729-734), are shown in yellow. B, ribbon diagram of the MBP-A CRD. Ca2+ site 1 is the auxiliary site, and Ca2+ site 2 is the principal site.

At the base of the molecule,  strands from the N and C termini pair to form the lower sheet and are similar to those in MBP. An additional antiparallel strand, 0, precedes 1; this strand is somewhat irregular and makes only a few hydrogen bonds with 1. The small sheet is stabilized by a disulfide bond formed between a cysteine in the loop preceding 0 and a cysteine in 1. This disulfide-stabilized, three-stranded antiparallel sheet has been observed in several other C-type lectin-like domains, including lithostathine (17), tetranectin (24), factor IX/X-binding protein (25), CD94 (26), and Ly49a (27). In addition to the known structures, CRD-2, -3, -6, and -8 of the mannose receptor as well as other C-type lectins such as the rat hepatic lectin (RHL) contain this pair of cysteines, suggesting that these domains will have a similar structure in this region. In CD94 (26) and Ly49a (27), this region participates in homodimer interfaces in which 0 from two molecules pair in a parallel orientation. It is possible that such a structure forms between some pairs of CRDs in the full mannose receptor.

Unlike the core of the domain, the "upper" portion of the polypeptide chain, which contains irregular loop regions, differs strikingly from that of MBP. The extended series of loops that follows 2, residues 701-734, extends away from the core of the domain (Fig. 2A), in contrast to MBPs, in which the equivalent region forms the upper part of the compact domain (Fig. 2B). In MBPs, this region is stabilized by two Ca²⁺ ions. The Ca²⁺ site 1 of MBPs, which is also found in tetranectin, is composed exclusively of residues from this region and is designated as the auxiliary Ca²⁺ site (12). Ca²⁺ site 2 of MBPs, which is conserved among all C-type lectin-like domains with sugar binding activity and participates in direct interactions with bound sugar ligands, is formed in part by residues from extended loops as well as residues from 4. This site is referred to as the principal site. Although CRD-4 binds to two Ca²⁺ in solution (11), only the principal Ca²⁺ is observed in the CRD-4 structure described here. The implications of this observation will be discussed below.

Superposition of the core of the domain from the two copies of CRD-4 in the asymmetric unit of the crystal reveals a significantly different disposition of the extended loop with respect to the core of the protein (Fig. 3). The difference can be described as a rotation of the loop by 43° with respect to the core. This observation suggests that the loop is flexibly disposed with respect to the core of the domain. The general shape and disposition of this region is similar to that of the coagulation factor IX/X-binding protein (IX/X-BP) (Fig. 3).

View larger version (42K): [in this window] [in a new window]   Fig. 3.   Flexibility of the extended loop region. The CRD cores of the two independent copies of mannose receptor CRD-4 (brown, copy A; blue, copy B) have been superimposed along with the cores of the two subunits of the IX/X-BP (yellow and cyan).

Domain-swapped Dimer Structure-- Extensive interactions occur between the extended loops formed by residues 701-734 of two CRD-4 molecules related by a crystallographic 2-fold rotational symmetry axis (Fig. 4, A and B). Most strikingly, the most distal portion of the loop (residues 708 to 728) forms the upper portion of the Ca²⁺- and sugar-binding site of the partner molecule. This domain swapping (23) produces a dimer in which each end, comprised of residues 625-700 and 735-768 from one polypeptide chain and residues 708-728 from a crystal symmetry-related molecule, has the compact fold typical of C-type CRDs (except IX/X-BP) (Fig. 4C). This "hybrid molecule" will be referred to as CRD4 monomer-like (CRD4-M). Comparison of CRD4-M with MBP (Fig. 2B) and E-selectin as well as the other C-type CRDs reveals a remarkable similarity in the structure of these loops. The two crossover segments that connect residues 708-728 to the core of the CRD, residues 701-707 and residues 729-734, will be referred to as regions I and II, respectively. Regions I and II are examples of "hinge loops," which are segments of polypeptide that link the swapped domain to the remainder of the molecule and which have different conformations in the monomer and dimer (23). The IX/X-BP, which is a heterodimer of C-type lectin-like domains, serves as a precedent for the domain-swapped architecture observed here. In addition to the non-covalent interactions between the extended loop and the partner molecule such as that seen in CRD-4, the protomers of IX/X-BP are linked by a disulfide bond in the region that corresponds to region I of CRD-4.

View larger version (33K): [in this window] [in a new window]   Fig. 4.   Domain-swapped dimer structure. A, two molecules of copy A of CRD-4, related by a crystallographic 2-fold rotational symmetry axis in the lattice. B, crystallographic dimer of copy B. The two protomers are shown in blue and red. C, CRD-4-MF. The distal part of the extended loop of the partner protomer in the dimer, which forms part of the Ca2+-binding site, is shown in red, next to the blue core of the other protomer. Note the similarity to the MBP-A structure in Fig. 1B. D, hydrophobic stabilization of the domain-swapped dimer. The two protomers of copy B are shown in red and blue, with the side chains of aromatic and certain other hydrophobic residues that stabilize the interchanged loop structure indicated.

The extended loop structure extending from region I to region II is anchored by a series of aromatic and other hydrophobic residues extending between the cores of the two CRDs in a dimer. Phe-708, Trp-710, Trp-721, Pro-726 as well as Trp-746 in  strand 4 participate in interactions (28) similar to those in other C-type lectin and C-type lectin-like domains (Fig. 4D). Moreover, three other residues that are part of the region I and region II crossover segments, Tyr-701, Pro-704, and Tyr-729, interact with these other residues to create a continuous hydrophobic ladder between the protomers. In addition to these hydrophobic packing interactions, residues from both protomers form the principal Ca²⁺-binding site. Glu-725, Asn-727, and Asn-728 present in the loop ligate the Ca²⁺, whose other ligands are Asn-747 and Asp-748 of the partner protomer. Thus, the loop penetrates into the partner molecule and mimics not only the overall backbone path but the detailed hydrophobic and Ca²⁺ ligation interactions observed in MBP, E-selectin, and other members of the superfamily.

Previous characterization of CRD-4 by gel filtration and chemical cross-linking suggested that it is a monomer in solution at micromolar concentrations (11). Given the extensive interactions observed in the crystallographic domain-swapped dimer, we reinvestigated the oligomeric state of the domain by sedimentation equilibrium analytical ultracentrifugation, which provides accurate molecular weight measurements independent of molecular shape. Measurements were carried out under a variety of buffer, divalent cation, and protein concentration conditions. Since acid pH and reverse-phase HPLC purification have been shown to induce formation of a strand-swapped dimer in cyanovirin-N (29), measurements were carried out at pH 4.5 and 6.0 as well as 7.8, and protein purified by HPLC was compared with protein not further purified after elution from mannose-Sepharose. In all conditions tested, the domain behaved as a monomer with molecular weight estimates close to the predicted value of 16,781 (data not shown). At conditions closest to those used for crystal growth (25 mM Tris HCl, pH 7.8, 0.5 M NaCl, with a highest starting protein concentration of 7 mg/ml), the molecular weight obtained was 17,771 ± 113 in the absence of divalent cations and 17,682 ± 253 in the presence of 5 mM CaCl₂ and 20 mM MgCl₂. The values given represent the mean and S.D. of estimates derived from runs at 20,000 and 22,000 rpm, each of which was obtained from simultaneous fitting of data from three cells containing different protein concentrations. In the absence of any evidence for dimer formation in solution, it must be concluded that the domain swapping is a consequence of lattice formation.

As in other lectins, the relative spatial arrangement of the mannose receptor CRDs is central to high avidity, specific multivalent ligand recognition. Moreover, the pH sensitivity of Ca²⁺ binding differs between isolated CRD-4 and larger fragments of the receptor, which implies that inter-CRD interactions in the intact mannose receptor can tune the pH sensitivity of ligand binding (11). Although we do not believe that the domain-swapped dimer observed in the CRD-4 crystal has any physiological significance, it is interesting to speculate that this mode of association might represent a mode of inter-CRD interaction between at least some of the domains in the intact receptor. Interestingly, the exchanged region of CRD-4, residues 707-729, is well conserved with the equivalent region of CRD-5, residues 854-875.

Ca²⁺ Coordination-- Measurements of the Ca²⁺ dependence of sugar binding and resistance to proteolysis indicate that CRD-4 binds to two Ca²⁺ in solution (11). Sequence alignments reveal that the residues in the principal Ca²⁺ site that participate directly in sugar binding in MBPs are identical in mannose receptor CRD-4. In contrast, only some of the residues that form the auxiliary Ca²⁺ site in MBPs are conserved in CRD-4. Mutagenesis of CRD-4 suggested that the principal Ca²⁺ site is similarly structured in MBPs and CRD-4, whereas the ligation of the auxiliary Ca²⁺ is likely to be different (12).

In the present crystal structure only the principal Ca²⁺ site is occupied. The site is formed by residues from both protomers of the domain swapped dimer (Fig. 5A). Surprisingly, the site differs in several respects from that of MBPs (Fig. 5A). The Ca²⁺ is eight-coordinated, with the five amino acids contributing seven ligands and a water molecule occupying the eighth position. The side chain of Glu-725 is shifted downward with respect to the homologous Glu-185 of MBP-A such that both of its carboxylate oxygens bind to the Ca²⁺. Moreover, the carbonyl oxygen of the Asn-728 side chain occupies a similar coordination position to that of Glu-193 of MBP-A. In MBP-A, Asp-188, the residue equivalent to Asn-728 of CRD-4, ligates the auxiliary Ca²⁺; conversely, Glu-733 of CRD-4, which is equivalent to Glu-193 of MBP-A, forms part of the region II crossover segment and does not participate in Ca²⁺ binding. This "swap" of Asn-728 for the expected Glu-733 is reminiscent of the E-selectin Ca²⁺-binding site, where Asn-83 (equivalent to Asn-728 of CRD-4) serves indirectly in Ca²⁺ binding by positioning a water molecule in the coordination shell, and Glu-88 (equivalent to Glu-733 of CRD-4) is swung out of the site (Fig. 5B).

View larger version (33K): [in this window] [in a new window]   Fig. 5.   Comparison of Ca2+ binding in the principal site between mannose receptor CRD-4 (gray bonds), MBP-A (brown bonds), and E-selectin (yellow bonds). Red, blue, and green spheres represent oxygen, nitrogen, and calcium, respectively. Ca2+ coordination bonds are shown as dashed lines and hydrogen bonds are shown as thin solid lines. In CRD-4, residues 725, 727, and 728 come from one protomer of the dimer, and 747 and 748 come from the other protomer. In A, and B, the residue numbers are from the mannose receptor, with the equivalent MBP-A (A) or E-selectin (B) residue numbers shown in parentheses. In C, MBP-A numbers are shown with equivalent residue numbers of E-selectin shown in parentheses.

The lack of the expected auxiliary Ca²⁺ in the CRD-4 structure is correlated with the unexpected loop extension and domain swapping. In MBPs and tetranectin the auxiliary Ca²⁺ site is formed in part by residues equivalent to regions I and II of CRD4; MBP-A residues Asp-161 and Glu-165 are equivalent to Tyr-701 and Glu-706 in region I of CRD4, and Glu-193 (whose main-chain carbonyl oxygen is a ligand for the auxiliary Ca²⁺) and Asp-194 are equivalent to CRD-4 region II residues Glu-733 and Tyr-734. As two of these acid side chains in MBP-A, 161 and 194, are not conserved in CRD-4, it was expected that the auxiliary site would be different. Indeed, mutation of CRD-4 residues Tyr-701 and Tyr-734 to phenylalanine and Glu-706 to alanine produced no change in Ca²⁺ binding (12). Nonetheless, mutagenesis of all potential Ca²⁺-ligating residues that lie within 10 Å of the equivalent MBP-A auxiliary site identified this general area as forming the site. However, since the present structure lacks the auxiliary Ca²⁺ and appears to have an altered conformation as a consequence, we cannot interpret these results beyond the model provided by Mullin et al. (12).

Sugar Binding-- In an attempt to visualize directly the mode of sugar binding to CRD-4, diffraction data were measured from a crystal soaked in a solution containing 200 mM Man(1,3)Man, which is well above the K_D of 2.4 mM measured for -methyl mannoside (12). No electron density was observed for the sugar. Inspection of the lattice contacts suggests that there would be ample room for the ligand to bind in the mode observed in MBPs (30, 31). There is no evidence for conformational changes upon binding in any of the known C-type lectin complex structures (30-33). If conformational changes were induced upon or required for binding, the crystals might be expected to crack or dissolve when soaked in this compound, which did not occur; also, the crystal used in this experiment was grown in a solution containing 100 mM -methyl mannoside, yet it was isomorphous with the native crystal (Table I). Therefore, it appears that the conformation of CRD-4 present in these crystals is not able to bind to sugars.

Comparison of the Structure with Mutagenesis Data-- The inability to observe a bound ligand in the crystals and the fact that one of the two Ca²⁺ ions known to be required for sugar binding by CRD-4 is not present led us to examine earlier site-directed mutagenesis results (12) in the context of the present structure. These studies were based on the homology between MBPs and mannose receptor CRD-4 and provided evidence that the mode of sugar recognition by CRD-4 is similar to that of MBPs. The essential aspect of sugar binding to MBPs is the coordination of the principal Ca²⁺ and formation of hydrogen bonds with amino acids that also serve as Ca²⁺ ligands by the equatorial 3- and 4-OH groups of mannose (30, 31) (Fig. 5A). A lone pair of electrons from each OH forms a coordination bond with Ca²⁺; the other lone pair acts as a hydrogen bond acceptor from an NH₂ group of asparagine residues, and the protons of these hydroxyl groups serve as hydrogen bond donors to acidic oxygen atoms (Fig. 5A).

The full complement of coordination ligands and hydrogen bond donors and acceptors appears to be an absolute requirement for carbohydrate binding by C-type lectins. For example, in MBP-A the change Glu-185  Gln (34) leaves Ca²⁺ binding unaffected but eliminates sugar binding because the NH₂ of the Gln side chain cannot act as a hydrogen bond acceptor for a sugar OH group. Similarly, the change Asn-187  Asp in MBP-A has the same phenotype (30). The change Glu-193  Gln in MBP-A would be predicted to retain Ca²⁺ binding but eliminate the required hydrogen bond acceptor provided by the carboxylate side chain. The analogous mutation in mannose receptor CRD-4, Glu-733  Gln, displays these characteristics (12). In the present structure, however, Glu-733 is not involved with Ca²⁺ binding and is "replaced" by Asn-728 in the Ca²⁺ coordination shell. The NH₂ group of Asn-728 would not be able to act as a hydrogen bond acceptor, so it is not clear how this structure would bind to sugar ligands. Also, the mutant Asn-728  Ala binds to mannose-bovine serum albumin (albeit more weakly than wild-type CRD-4) and shows first-order concentration dependence in Ca²⁺ binding (12). The phenotype of the Asn-728  Ala mutant can only be explained if Asn-728, like the equivalent residue Asp-188 in MBP-A, is a ligand for the auxiliary Ca²⁺ but not the principal Ca²⁺ (12).

His-189 of MBP-A and the equivalent Val194 of MBP-C are located in the sugar-binding site and form van der Waals contacts with the bound ligand. The homologous residue of CRD-4, Tyr-729, has been shown to contribute to sugar binding. NMR spectra of bound ligands display upfield ring current shifts consistent with their interaction with an aromatic ring. These shifts disappear in Tyr-729  Ala or Tyr-729  Leu mutants but are present in Tyr-729  Phe (12). These observations provide strong support for a CRD-4 Ca²⁺ and sugar-binding site very similar to that of MBPs. In the present structure Tyr-729 is located in the region II hinge loop and occupies a position underneath the Ca²⁺ site that would preclude interaction with a sugar bound in the orientations observed in MBP-A and MBP-C.

The comparison of mutant phenotypes and the present crystal structure leads to the conclusion that the conformation observed here is not competent to bind the full complement of Ca²⁺ or to sugar ligands. It seems likely that the principal Ca²⁺ binding site is rearranged relative to the fully bound, two Ca²⁺ state that has been analyzed in solution. The possibility that the sugar binding mode of mannose receptor CRD-4 is fundamentally different from other C-type lectins cannot be excluded, but the available evidence suggests that it is indeed very similar.

As noted above, Glu-88 of E-selectin is swung out of the Ca²⁺ site and replaced by Asn-83 (Fig. 5C), which positions a water molecule for interaction with the Ca²⁺ (16). Since a water molecule has both hydrogen bond donor and acceptor potential, this structure might support fucose binding, with the water molecule acting as a hydrogen bond acceptor, analogous to the role of the second carboxylate oxygen of Glu-193 in MBP (Fig. 5C). The crystal structure of E-selectin was obtained under high salt conditions² known to inhibit binding of sialyl Lewis^x and related ligands to the protein (35), which most likely explains why soaking these crystals in the sugar did not reveal any sugar density (16). The present results, however, raise the possibility that the structure observed in the crystal structure of E-selectin is not competent to bind to sugar; Glu-88, the equivalent of MBP-A Glu-193 and CRD-4 Glu-733, would have to swing into the site to provide the correct set of interactions with fucose.

A Possible Intermediate in Ca²⁺ Binding-- It is likely that the structure reported here corresponds to an intermediate state between Ca²⁺-free CRD4 and a state fully bound by Ca²⁺ that is able to bind carbohydrate ligands. Such an intermediate state must exist physiologically, as the Ca²⁺ dependence of sugar binding is exploited by this and other endocytic receptors in order to deliver endocytosed ligands to intracellular compartments. The mannose receptor binds its ligands in the high Ca²⁺, neutral pH environment of the extracellular milieu. The endocytic vesicle containing the receptor-ligand complex becomes progressively more acidic, resulting in titration of Ca²⁺ off the CRD and loss of the ligand. The ligand and receptor sort from one another, with the receptor returning to the cell surface for another round of ligand binding. C-type CRDs generally show this reversible, pH-dependent Ca²⁺ and sugar binding, with the pH optimum of Ca²⁺ binding dependent on the particular protein. For those C-type lectins that function as recyclable receptors, the pH optimum is likely to be tuned to release ligand in the appropriate target compartment.

Isolated mannose receptor CRD-4 displays reversible, pH-dependent Ca²⁺ and sugar binding. The midpoint of the transition between Ca²⁺ bound and free forms is approximately pH 5.0 (11). In the crystals studied here, the pH is 8.0, which should support Ca²⁺ binding. Formation of crystals, however, was observed to be exquisitely sensitive to the concentration of Ca²⁺; crystals only grew in a very narrow range of 6.5-7.5 mM. The apparent K_Ca, pH 7.8, obtained by measuring Ca²⁺ dependence of sugar binding is 0.26 mM, and the Ca²⁺ dependence of resistance to proteolysis (the domain becomes very sensitive to proteases in the absence of Ca²⁺) gives a K_Ca of 1.2 mM (11). The Ca²⁺ concentration under which the crystals grow is, therefore, not very far over the K_Ca. Moreover, the effective Ca²⁺ concentration is probably somewhat lower than the nominal concentration because the precipitating agent, polyethylene glycol, has many vicinal hydroxyl groups that might interact with free Ca²⁺. These observations are consistent with the notion that the crystal structure corresponds to a one-ion form of the protein present in solution. Because the effect of low pH is to reduce the amount of bound Ca²⁺ by lowering the affinity, information about Ca²⁺ free or partially bound states at neutral pH, obtained in limiting Ca²⁺ concentrations, is likely to be relevant to a description of the structure at acid pH.

Ca²⁺-dependent conformational changes in C-type CRDs have been studied structurally and kinetically in MBPs (36, 37). Crystal structures of Ca²⁺-free MBP and a form in which only the auxiliary Ca²⁺ site is occupied show that loss of the principal Ca²⁺ results in a cis  trans isomerization of a conserved proline that must be in the cis conformation in order to form a Ca²⁺ binding site. The cis  trans isomerization results in major structural rearrangements in the CRD. Ca²⁺ binding or release occurs in two distinct kinetic phases, one of which involves rapid ion binding or release, and the other, a slow cis-trans isomerization. Modeling of the kinetic data indicates that a state containing a cis-proline in the principal Ca²⁺ site, but with only one ion bound to the protein, is likely to exist in solution; the models do not distinguish which site is occupied in this state. Curiously, the one-ion structure of MBP-A shows that the principal Ca²⁺ site can be rearranged without effect on the auxiliary site, although the physiological significance of this state is unclear.

In the present structure of CRD-4, the fact that only the principal Ca²⁺ site is occupied and that the extended loop between  strands 2 and 3 domain appears to be flexibly linked to the core suggests that in solution, under low Ca²⁺ conditions or endosomal pH, this region is flexible and does not interact strongly with the rest of the domain. There are probably many conformational states of this loop with roughly equal energies; the fact that two different dimer conformations are observed in the crystal (Fig. 3) supports the notion that a flexibly linked loop is fluctuating among various conformations in solution. The aromatic ladder that forms the dimer interface is composed largely of tyrosine residues and so is not especially hydrophobic, suggesting that this loop can adopt the observed extended conformation in solution. It seems likely that the lattice has trapped an extended conformation from the ensemble of solution conformations, such that the distal part of the loop can interact with a partner molecule in a manner similar to how it likely interacts with the core of the domain in the monomeric form. This interaction probably reflects the inherent ability of this portion of the polypeptide to form part of the Ca²⁺-binding site.

A Mechanism for Endosomal Release of Ligand-- The general resemblance of the principal Ca²⁺-binding site formed by the domain-swapped dimer to that of MBPs suggests that the conformation of this region of the structure, although it is interacting with a partner protomer rather than forming a compact CRD, largely resembles the monomeric, solution structure of CRD-4 in this region. However, detailed differences in structure (Fig. 5A) and mutagenesis data implicating Asn-728 and Glu-733 in binding the auxiliary and principal Ca²⁺ suggest that the loss of the auxiliary Ca²⁺ produces rearrangements in the structure that prevents sugar from binding at the principal Ca²⁺ site. In mannose receptor CRD-4, the auxiliary Ca²⁺ may serve as a pH sensor in the endosome, with its loss coupled to changes in the principal Ca²⁺ site that make the latter unable to bind carbohydrate ligands, despite the presence of Ca²⁺.

In addition to the mannose receptor, other C-type lectins that serve as endocytic receptors, such as the mammalian and avian hepatic lectins, require two Ca²⁺ ions for sugar binding (38). Like CRD-4 of the mannose receptor, the CRDs of both RHL and the chicken hepatic lectin undergo pH-dependent conformational changes that result in decreased affinity for Ca²⁺ and loss of sugar binding activity (38, 39). However, CRD-4 retains ligand binding activity at lower pH than the CRDs of RHL and the chicken hepatic lectin, indicating that there must be differences in the exact mechanism of pH-induced ligand release from mannose receptor CRD-4 (11). These differences are likely to be related to the fact that CRD-4 must operate in the context of multiple different CRDs within the whole mannose receptor.

Recent mutagenesis studies of the RHL CRD have identified a histidine residue (His-256), conserved in the chicken hepatic lectin and the macrophage asialoglycoprotein receptor but not in mannose receptor CRD-4, that is thought to form part of a pH-sensitive switch for Ca²⁺ and sugar binding (39). Mutation of RHL His-256 results in an increase in Ca²⁺ affinity at physiological pH and a downward shift in the pH at which Ca²⁺ is released (39). A similar phenotype is observed upon mutation of Asp-266, whereas the opposite behavior is obtained upon mutation of Arg-270. From a model of the RHL CRD based on the crystal structure of a galactose binding mutant of MBP, it is predicted that His-256, Asp-266, and Arg-270 are close to the principal Ca²⁺ site. In particular, His-256 is likely to form a hydrogen bond with the amide moiety of Asn-264, whose side chain carbonyl oxygen ligates the principal Ca²⁺. Protonation of His-256 would disrupt this hydrogen bond and directly affect the position of Asn-264, thereby decreasing the affinity for the principal Ca²⁺ (39). This mechanism is in contrast to that proposed here for mannose receptor CRD-4, in which pH-induced loss of Ca²⁺ from the auxiliary site alters the principal Ca²⁺ site. In each of these CRDs, however, it is likely that ligation of the auxiliary Ca²⁺ contributes to establishing the affinity of Ca²⁺ binding at the principal site. Sequence and genomic evidence suggest that the mannose receptor is very divergent from RHL, so it is perhaps not surprising that the two proteins have evolved different mechanisms for release of ligand (40).

In MBPs, which are not endocytic receptors, the loss of the Ca²⁺ in the principal site results in a kinetic block to rebinding due to the need to reisomerize the proline to the cis configuration, a process that takes several minutes (36, 37). Given the high conservation of the principal Ca²⁺ site, it is likely that this behavior will be true of other C-type lectins. In at least some C-type lectins that are true endocytic receptors, a kinetic block to rebinding Ca²⁺ may not be desirable. Moreover, the strict requirements of the principal site for the arrangement of Ca²⁺ coordination ligands and hydrogen bond donors and acceptors may make it difficult to tune its pH sensitivity for release in the appropriate intracellular compartment. Therefore, the conformation of the principal Ca²⁺ site may be coupled to the auxiliary site so that the basic sugar binding mechanism can be retained while differences in pH sensitivity can be encoded in a second site with less stringent structural requirements.

	ACKNOWLEDGEMENTS

We thank Kurt Drickamer for discussions and comments on the manuscript and Russell Wallis for help with analytical ultracentrifugation.

	FOOTNOTES

^* This work was supported by Wellcome Trust Grants 054508 and 041845, the Glycobiology Institute Endowment (to M. E. T.), and National Institutes of Health Grant GM50565 (to W. I. W.).The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

The atomic coordinates and structure factors (codes 1EGG for native and 1EGI for Man(1,3)Man-soaked CRD-4) have been deposited in the Protein Data Bank, Research Collaboratory for Structural Bioinformatics, Rutgers University, New Brunswick, NJ (http://www.rcsb.org/).

^§ To whom correspondence should be addressed: Dept. of Structural Biology; Fairchild Bldg., Stanford University School of Medicine, 299 Campus Dr. W. Stanford, CA 94305. Tel.: 650-725-4623; Fax: 650-723-8464; E-mail: bill.weis@stanford.edu.

Published, JBC Papers in Press, April 21, 2000, DOI 10.1074/jbc.M002366200

² B. Graves, personal communication.

	ABBREVIATIONS

The abbreviations used are: CRD, carbohydrate recognition domain; CRD4-M, CRD4 monomer-like; GlcNAc, N-acetylglucosamine; IX/X-BP, coagulation factor IX/X-binding protein; Man, Mannose; Man(1, 3)Man, methyl 3-O-(-D-mannopyranosyl)--D-mannopyranoside; RHL, rat hepatic lectin; HPLC, high performance liquid chromatography; MBP, mannose-binding protein; r.m.s.d., root mean square deviation.

	REFERENCES

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS AND DISCUSSION REFERENCES

CiteULike

Complore

Connotea

Del.icio.us

Digg

Reddit

Technorati

This article has been cited by other articles:

				A. Bugarcic, K. Hitchens, A. G Beckhouse, C. A Wells, R. B Ashman, and H. Blanchard Human and mouse macrophage-inducible C-type lectin (Mincle) bind Candida albicans Glycobiology, September 1, 2008; 18(9): 679 - 685. [Abstract] [Full Text] [PDF]

				J. Sturge, S. K. Todd, G. Kogianni, A. McCarthy, and C. M. Isacke Mannose receptor regulation of macrophage cell migration J. Leukoc. Biol., September 1, 2007; 82(3): 585 - 593. [Abstract] [Full Text] [PDF]

				E. P. McGreal, M. Rosas, G. D. Brown, S. Zamze, S. Y.C. Wong, S. Gordon, L. Martinez-Pomares, and P. R. Taylor The carbohydrate-recognition domain of Dectin-2 is a C-type lectin with specificity for high mannose Glycobiology, May 1, 2006; 16(5): 422 - 430. [Abstract] [Full Text] [PDF]

				J. Boskovic, J. N. Arnold, R. Stilion, S. Gordon, R. B. Sim, A. Rivera-Calzada, D. Wienke, C. M. Isacke, L. Martinez-Pomares, and O. Llorca Structural Model for the Mannose Receptor Family Uncovered by Electron Microscopy of Endo180 and the Mannose Receptor J. Biol. Chem., March 31, 2006; 281(13): 8780 - 8787. [Abstract] [Full Text] [PDF]

				S. A. Mah, W. J. Swanson, and V. D. Vacquier Positive Selection in the Carbohydrate Recognition Domains of Sea Urchin Sperm Receptor for Egg Jelly (suREJ) Proteins Mol. Biol. Evol., March 1, 2005; 22(3): 533 - 541. [Abstract] [Full Text] [PDF]

P. Paaventhan, C. Kong, J. S. Joseph, M. C.M. Chung, and P. R. Kolatkar Structure of rhodocetin reveals noncovalently bound heterodimer interface Protein Sci., January 1, 2005; 14(1): 169 - 175. [Abstract]